{ "name": "AI Agents for Manufacturing", "description": "Practical, sourced AI use cases for manufacturing companies. Figures are as reported by each case's sources (often vendor or self-reported); see https://aiagentsformanufacturing.com/evidence/ for evidence levels.", "url": "https://aiagentsformanufacturing.com/", "last_updated": "2026-10-29", "count": 84, "departments": [ { "name": "Procurement", "url": "https://aiagentsformanufacturing.com/departments/procurement/", "count": 12 }, { "name": "Production", "url": "https://aiagentsformanufacturing.com/departments/production/", "count": 12 }, { "name": "Quality", "url": "https://aiagentsformanufacturing.com/departments/quality/", "count": 12 }, { "name": "Maintenance", "url": "https://aiagentsformanufacturing.com/departments/maintenance/", "count": 12 }, { "name": "Logistics / Supply Chain", "url": "https://aiagentsformanufacturing.com/departments/logistics/", "count": 12 }, { "name": "Engineering", "url": "https://aiagentsformanufacturing.com/departments/engineering/", "count": 12 }, { "name": "Office & management", "url": "https://aiagentsformanufacturing.com/departments/office/", "count": 12 } ], "feeds": { "rss": "https://aiagentsformanufacturing.com/feed.xml", "llms_txt": "https://aiagentsformanufacturing.com/llms.txt", "llms_full_txt": "https://aiagentsformanufacturing.com/llms-full.txt" }, "use_cases": [ { "title": "AI normalizes supplier quotes into one comparison", "url": "https://aiagentsformanufacturing.com/use-cases/supplier-quote-comparison/", "markdown": "https://aiagentsformanufacturing.com/use-cases/supplier-quote-comparison.md", "department": "Procurement", "department_url": "https://aiagentsformanufacturing.com/departments/procurement/", "evidence": "Verified", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Purchasing", "problem": "Quotes come back as PDFs, spreadsheets and emails, each with its own layout, units, price breaks, MOQs, freight and payment terms. Someone re-keys them into a comparison sheet line by line. At OFS (furniture manufacturer, Huntingburg, Indiana), comparing just 10–15 line items could take more than two hours, and manual sourcing work took 20–25 hours a week. At Hyphen (robotic food-assembly machines, ~900 made-to-spec parts per order), normalizing quotes took about 5 hours per sourcing event.", "how_ai_helps": "An agent reads each supplier reply in whatever format it arrives, extracts the prices, MOQs, volume tiers, lead times, freight, tariffs and terms into one model, calculates landed cost at the quantity you choose, flags missing lines and proposes an award. The buyer checks the extracted figures against the source and makes the award.", "result": "Hyphen, Syntro agent: time spent on quote normalization went from about 5 hours to about 5 minutes; a supplier that quoted only 7 of 8 parts was flagged in about 30 seconds instead of a 30-minute line-by-line check; because the team could quote 5 times as many suppliers, made-to-spec part costs fell about 10%; sourcing a full machine (RFQ to PO) went from 60 to 6 hours (Syntro customer story). OFS, Lumari agents: RFQ processing 10x faster and landed-cost comparison 2x faster; one sourcing specialist handled what had been a team-scale workload; implemented in 2 weeks and suppliers kept quoting by email, Excel or PDF; OFS still weighs tooling, MOQ, lead time and supplier capability before awarding (Lumari case study). Both are vendor-published.", "companies": [ { "name": "Hyphen", "url": "https://aiagentsformanufacturing.com/companies/#hyphen" }, { "name": "OFS", "url": "https://aiagentsformanufacturing.com/companies/#ofs" } ], "tools_deployed": [ { "name": "Lumari Forge", "url": "https://aiagentsformanufacturing.com/tools/#lumari-forge" }, { "name": "Syntro", "url": "https://aiagentsformanufacturing.com/tools/#syntro" } ], "tools": [ "Syntro (agent connected to the buyer's inbox, PLM and ERP): used at Hyphen", "Lumari Forge agents: used at OFS", "Other tools to evaluate: Keelvar, Arkestro, Fairmarkit, Luminovo (electronics BOMs), sourcing modules of SAP Ariba or Coupa", "Do-it-yourself: ChatGPT, Claude or Copilot to extract quotes into a table, with landed cost calculated in a spreadsheet" ], "sources": [ "https://www.usesyntro.com/customers/hyphen", "https://lumari.ai/case-studies/ofs" ], "date_added": "2026-10-12", "last_updated": "2026-10-14" }, { "title": "AI builds RFQ packages and finds suppliers for hard-to-source parts", "url": "https://aiagentsformanufacturing.com/use-cases/rfq-generation-and-supplier-discovery/", "markdown": "https://aiagentsformanufacturing.com/use-cases/rfq-generation-and-supplier-discovery.md", "department": "Procurement", "department_url": "https://aiagentsformanufacturing.com/departments/procurement/", "evidence": "Verified", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Purchasing", "problem": "Every RFQ means pulling drawings one at a time, writing a package and a cover sheet, choosing suppliers from memory, and chasing replies. Watlow (industrial heaters, 13 plants) saw RFQ responses stretch to 45–60 days for typical packages and had 235 'unsourceable' parts sitting idle because it could not see which suppliers could make them. DMG MORI's purchasing team found manual supplier scouting from references and web searches slow and unsatisfying.", "how_ai_helps": "AI extracts material, process and features from 2D and 3D files, finds similar parts already bought and the suppliers who made them, and builds the RFQ package with a cover sheet. For new categories, a supplier search engine screens external data (trade data, HS codes, websites) against a requirement profile and returns a long list to qualify. The buyer approves the supplier list and the package before anything is sent.", "result": "Watlow, Paperless Parts Radar: a strategic sourcing action that would have taken 4–5 weeks of engineering and procurement time was done in 2–3 days (80%+ less RFQ time); all 235 previously 'unsourceable' parts categorized with potential sources from the existing approved supplier list; 35% direct cost savings on Wave 1 parts quoted through the tool; 4 of 5 suppliers polled said the packages were easier to quote (Paperless Parts case study). DMG MORI (machine tools), Matchory supplier search: first used for cast-iron suppliers in India, China and Eastern Europe; 317 global suppliers identified across 11 raw-material categories, 9 matches in 3 months, 3 months from kickoff to outreach (Matchory success story; its 'up to 85% faster scouting' is stated as the goal, not a measured figure). Both are vendor-published.", "companies": [ { "name": "Watlow", "url": "https://aiagentsformanufacturing.com/companies/#watlow" } ], "tools_deployed": [ { "name": "Matchory", "url": "https://aiagentsformanufacturing.com/tools/#matchory" }, { "name": "Paperless Parts", "url": "https://aiagentsformanufacturing.com/tools/#paperless-parts" } ], "tools": [ "Paperless Parts Radar (part data extraction, similarity matching, RFQ package generation): used at Watlow", "Matchory (AI supplier search): used at DMG MORI", "Other tools to evaluate: Lumari, Syntro, Thomasnet and other supplier databases, sourcing modules of SAP Ariba or Coupa", "Do-it-yourself: an LLM to draft the RFQ sheet and emails from your part data, with suppliers chosen by the buyer" ], "sources": [ "https://radar.paperlessparts.com/resources/case-studies/watlow-centralizes-sourcing-intelligence-to-accelerate-rfqs-by-80-and-unlock-cost-savings-with-radar", "https://www.matchory.com/en/success-stories/smart-supplier-search-at-dmg-mori-with-matchory" ], "date_added": "2026-10-09", "last_updated": "2026-10-14" }, { "title": "AI replenishment recommendations so reorders are not missed", "url": "https://aiagentsformanufacturing.com/use-cases/ai-replenishment-recommendations/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-replenishment-recommendations.md", "department": "Procurement", "department_url": "https://aiagentsformanufacturing.com/departments/procurement/", "evidence": "Verified", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Purchasing", "problem": "Reorders are decided from spreadsheets or static min/max settings. With thousands of raw materials and long, unstable lead times, buyers miss items that are about to run out and overbuy others, so production stops for a missing part while cash sits in excess stock. AAI (US fire-sensor maker) managed about 3,000 raw materials and 550 products on spreadsheets, with frequent stock-outs; GRPI (US inks and coatings maker) had a three-month lead time on raw materials and planned in spreadsheets with little visibility across thousands of SKUs.", "how_ai_helps": "A planning tool connected to the ERP forecasts usage per item, recalculates safety stock and reorder points from real lead times, and every day lists the items that need an order now, with a suggested quantity. The buyer reviews the list, groups items by supplier and places the POs; nothing is ordered without the buyer.", "result": "AAI (Sage 100cloud ERP, Netstock): daily order recommendations and item classification by value and velocity; the company reports fewer stock-outs and a 99% fill rate in its Residential division; Netstock also recommended higher safety stock where lead times exceed 52 weeks (Netstock case study). GRPI (Fishbowl ERP, Netstock Opportunity Engine, an AI feature that surfaces daily replenishment and purchasing opportunities): found about $1.5 million of excess inventory it could not see in spreadsheets, and reports reduced stock-outs and better control of raw-material planning and safety stock; no stock-out figure is given (Netstock case study). Both are vendor-published. The AI here is forecasting and recommendation; the purchasing decision and the PO stay with the buyer.", "companies": [ { "name": "GRPI", "url": "https://aiagentsformanufacturing.com/companies/#grpi" } ], "tools_deployed": [ { "name": "Netstock", "url": "https://aiagentsformanufacturing.com/tools/#netstock" } ], "tools": [ "Netstock integrated with Sage 100cloud: used at AAI", "Netstock Opportunity Engine integrated with Fishbowl: used at GRPI", "Other tools to evaluate: Streamline (GMDH), Prediko, Inventory Planner, and the MRP / replenishment modules of your ERP", "Do-it-yourself: a spreadsheet or Python script for the reorder-point maths, plus an LLM to draft the buyer's summary and supplier emails" ], "sources": [ "https://www.netstock.com/case-studies/grpi/", "https://www.casestudies.com/company/netstock/case-study/netstock-ignites-inventory-processes-at-aai" ], "date_added": "2026-10-10", "last_updated": "2026-10-14" }, { "title": "Find existing parts before designing or buying a new one (AI shape search)", "url": "https://aiagentsformanufacturing.com/use-cases/find-existing-parts-ai-shape-search/", "markdown": "https://aiagentsformanufacturing.com/use-cases/find-existing-parts-ai-shape-search.md", "department": "Procurement", "department_url": "https://aiagentsformanufacturing.com/departments/procurement/", "evidence": "Verified", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Engineering", "problem": "Engineers lose time searching supplier catalogues and the company's own ERP and PDM for a part that fits, and when they cannot find it they design a new one, even if an identical part already exists under another name at another site. At Wanzl (shopping carts), finding a component and its CAD data often took more than two hours and sometimes ended with the part being modelled by hand. At Starlinger (machines for woven plastic sacks, 550 employees), parts were searched by text only in the ERP and PDM.", "how_ai_helps": "Shape-based (geometric) search compares the 3D geometry of a part with every part in the company library and supplier catalogues, and shows matches with their ERP and PLM data in seconds, so the engineer can reuse an existing or catalogue part. Newer tools use machine-learning embeddings of the geometry to match parts across sites regardless of their names. An engineer still checks fit, form, function and specs before reusing.", "result": "Starlinger, CADENAS PARTsolutions with geometric similarity search (GEOsearch), in use since 2011 (the CADENAS story is undated, so the figures may describe an earlier period): searches take 3–10 seconds; every engineer saves at least 15 minutes a day, which the company values at €112,000 a year at €50/hour; creation of new design parts fell by more than 10% despite more engineers and projects (CADENAS success story). Wanzl, PARTsolutions: central access to manufacturer catalogues with 3D CAD data at four sites; the company reports more than two hours saved per searched purchased part (CADENAS success story, undated). A global multi-site manufacturer (anonymous), Physna geometric AI: duplicate parts found across facilities under different names and 'up to 5x' engineering productivity, according to Physna; no other figures (AWS blog, Aug 2026). Note: CADENAS's geometric search is similarity matching, not a chatbot; all figures are vendor-published.", "companies": [ { "name": "Starlinger", "url": "https://aiagentsformanufacturing.com/companies/#starlinger" } ], "tools_deployed": [ { "name": "CADENAS PARTsolutions", "url": "https://aiagentsformanufacturing.com/tools/#cadenas-partsolutions" }, { "name": "Physna", "url": "https://aiagentsformanufacturing.com/tools/#physna" } ], "tools": [ "CADENAS PARTsolutions (GEOsearch geometric similarity search, 3D supplier catalogues): used at Starlinger and Wanzl", "Physna geometric AI: used at the anonymous global manufacturer", "Other tools to evaluate: CADSEEK, the geometric search in your PDM/PLM (e.g. SOLIDWORKS PDM, Siemens Teamcenter, PTC Windchill), 3Dfindit, Octopart for electronic components", "A general LLM to turn a requirement into search terms and a datasheet checklist (it does not know your library or live catalogues)" ], "sources": [ "https://partsolutions.com/starlinger-co-employ-partsolutions-standard-parts-management-system-to-streamline-design/", "https://www.cadenas.de/files/cadenas/Downloads/PDF/Referenzen/EN/PARTsolutions_Wanzl_Success_Story_EN.pdf", "https://aws.amazon.com/blogs/physical-ai/how-physna-and-aws-use-geometric-intelligence-to-bridge-engineering-design-and-procurement/", "https://www.reddit.com/r/manufacturing/comments/1n8kror/how_ai_is_helping_in_manufacturing_projects/" ], "date_added": "2026-10-07", "last_updated": "2026-10-14" }, { "title": "Supplier PO follow-ups and order confirmations by AI agent", "url": "https://aiagentsformanufacturing.com/use-cases/supplier-emails-and-follow-ups/", "markdown": "https://aiagentsformanufacturing.com/use-cases/supplier-emails-and-follow-ups.md", "department": "Procurement", "department_url": "https://aiagentsformanufacturing.com/departments/procurement/", "evidence": "Verified", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Purchasing", "problem": "Buyers spend much of their day chasing suppliers: sending POs, waiting for acknowledgements, asking again, checking that confirmed prices, quantities and dates match the PO, and typing the answers into the ERP. The conversation is scattered across email threads, follow-ups are inconsistent, and a missed reply shows up later as a late delivery. At Footprint, the VP of Procurement says the team was 'spending half our time creating POs'.", "how_ai_helps": "AI agents create POs from MRP output in the ERP, send them to suppliers, track acknowledgements, send polite reminders when nothing comes back, read the replies and record confirmed dates and prices in the ERP. When a supplier answers with a different price, quantity or payment term, the agent opens a task for the buyer with the whole context instead of updating anything itself.", "result": "Footprint (plant-based fiber packaging manufacturer), Didero agents on Sage X3: 2,600+ purchase orders a year processed, 5,800+ supplier follow-ups automated and 4,700+ MRO and spare-parts line items tracked (these are volumes, with no time saved or delivery improvement given); live about six weeks after the decision to go (Didero customer story, a vendor claim). An unnamed mid-size manufacturer with 80+ suppliers in 12 countries (Oido Studio case study, vendor claim): agents send POs by each supplier's preferred channel, remind after 24 hours and flag at 48 hours; procurement hours went from 40+ to about 8 a week of exception handling, supplier confirmation rate from 65% to 98% and procurement cycle time from 8 to 3.5 days; these cover several agents and the whole purchasing process (PO creation too), not only the follow-ups.", "companies": [ { "name": "Footprint", "url": "https://aiagentsformanufacturing.com/companies/#footprint" } ], "tools_deployed": [ { "name": "Didero", "url": "https://aiagentsformanufacturing.com/tools/#didero" } ], "tools": [ "Didero (procurement agents on Sage X3 and other ERPs): used at Footprint", "Oido Studio agents (custom, ERP via an MCP server)", "Order-confirmation tools: Tacto, Zalion, Workist", "Microsoft Copilot or Gemini in Outlook/Gmail with the prompt below for a manual version" ], "sources": [ "https://www.didero.ai/case-studies/footprint", "https://oidostudio.com/case-studies/manufacturing-supply-chain" ], "date_added": "2026-10-13", "last_updated": "2026-10-26" }, { "title": "AI shift handover summaries from voice and typed notes", "url": "https://aiagentsformanufacturing.com/use-cases/shift-handover-summaries/", "markdown": "https://aiagentsformanufacturing.com/use-cases/shift-handover-summaries.md", "department": "Production", "department_url": "https://aiagentsformanufacturing.com/departments/production/", "evidence": "Pilot", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Production", "problem": "Handovers happen verbally or on paper at the end of a tiring shift. Details get lost: a machine running hot, a quality flag on a batch, a temporary setting change, a late delivery. At Apollo Foods (New Zealand beverage maker, 25 people, 3 shifts a day), handovers were paper-based with 30-minute overlaps three times a day, and the team had no way to search past fixes. At an automotive parts plant in Saudi Arabia (350 people, 24/7), handovers were 15–20 minute verbal meetings with no written record, and the plant manager estimated 30% of night and early-morning production issues came from incomplete handover information. Demand is real: in September 2026 Volvo Cars Ghent published an open challenge asking for AI-generated shift briefings (a request for solutions, not a result).", "how_ai_helps": "People record the handover the way they already do it (voice, typing, photos). AI transcribes it, writes a short structured summary, pulls out every action item and open issue, and makes the history searchable so the next shift can ask 'what changed on line 2 overnight?'. A supervisor reads and corrects the summary before relying on it.", "result": "Apollo Foods, RossOps 30-day pilot (July 2025): 61 shift handovers completed with use on every shift from day one; 3.2x more actionable content captured, all readable; 40,000 characters logged in one month; six months of past paper handovers loaded so they could be searched. One example: an engineer found that an overnight temporary settings change caused a morning performance issue and reverted it, turning hours of mechanical troubleshooting into about 20 minutes. No time-saving per handover was measured (RossOps case study). Automotive parts plant, Saudi Arabia, anonymous (CubePeaks case study): tablets at each of the three handover stations record and transcribe on the device, and a local LLM writes the summary, action items and a score against the plant manager's checklist (safety, machine status, quality flags, deliveries); nothing leaves the building. The figures it publishes, including 15 minutes saved per shift for the incoming supervisor and 70% less information loss, are labelled PROJECTED results, not measured ones.", "companies": [ { "name": "Apollo Foods", "url": "https://aiagentsformanufacturing.com/companies/#apollo-foods" } ], "tools_deployed": [ { "name": "CubePeaks", "url": "https://aiagentsformanufacturing.com/tools/#cubepeaks" }, { "name": "RossOps", "url": "https://aiagentsformanufacturing.com/tools/#rossops" } ], "tools": [ "RossOps (voice, text and photo handovers with AI summaries and search): used at Apollo Foods", "CubePeaks on-device transcription and local LLM on tablets: used at the anonymous automotive plant", "Other tools to evaluate: Tulip Shift Summary Reporter, eschbach Shiftconnector, digital shift-log modules of MES/CMMS platforms", "Do-it-yourself: Microsoft Copilot, ChatGPT or Claude with the prompt below, on notes you paste in" ], "sources": [ "https://rossops.com/docs/apollo.pdf", "https://cube-peaks.com/case-studies/closing-shift-handover-gap/", "https://www.agoria.be/system/files/documents/2026-09/Partnering4Success%202026%20-%20BUSINESS%20CHALLENGE%201%20Volvo%20Cars%20-%20AI%20Shift%20Handover.pdf" ], "date_added": "2026-10-09", "last_updated": "2026-10-26" }, { "title": "AI-drafted SOPs and work instructions from existing documents", "url": "https://aiagentsformanufacturing.com/use-cases/ai-sop-work-instructions-from-existing-documents/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-sop-work-instructions-from-existing-documents.md", "department": "Production", "department_url": "https://aiagentsformanufacturing.com/departments/production/", "evidence": "Verified", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Production", "problem": "Procedures live in old PowerPoints, binders and office folders, not at the machine. Writing a proper step-by-step work instruction by hand means capturing photos or video, editing and typing every step, so documents fall behind the process. When an audit comes, quality teams scramble: a precision tooling maker faced an ISO recertification audit with only weeks to prepare and night-shift operators who could not quickly find procedures.", "how_ai_helps": "AI turns existing material (slides, manuals, OEM documents, videos, work-order history) into draft work instructions in a standard format, with steps, warnings and images. A subject-matter expert checks and approves the draft, and the result is published on tablets at the point of use and versioned for audits.", "result": "Multi-site precision tooling manufacturer (Minnesota and South Dakota), Dozuki: AI converted legacy PowerPoint training into guides that arrived about 90% complete, cutting authoring time by about 60% (roughly 30 minutes saved per guide), with tablets on every production floor before an ISO recertification audit (Dozuki customer story, a vendor claim). Maximal Group (small cardboard packaging maker, Romania), Tacit AI: from 42 OEM documents and 1,193 work orders, AI drafted work instructions and FMEAs 12 times faster, reviewed and implemented by engineers in days (Tacit AI case study, a vendor claim).", "companies": [], "tools_deployed": [ { "name": "Dozuki CreatorPro AI", "url": "https://aiagentsformanufacturing.com/tools/#dozuki-creatorpro-ai" }, { "name": "Tacit AI", "url": "https://aiagentsformanufacturing.com/tools/#tacit-ai" } ], "tools": [ "Dozuki (CreatorPro AI): used at the precision tooling maker", "Tacit AI: used at Maximal Group", "Other work-instruction platforms with AI authoring: Azumuta, SwipeGuide, Augmentir, Poka, Tulip", "Any LLM (ChatGPT, Claude, Copilot) with the prompt below for a first draft from text" ], "sources": [ "https://www.dozuki.com/customers/precision-tooling-manufacturer", "https://tacitx.ai/case-study/maximal-group/" ], "date_added": "2026-10-12", "last_updated": "2026-10-14" }, { "title": "Vision defect inspection on the edge", "url": "https://aiagentsformanufacturing.com/use-cases/vision-defect-inspection-on-the-edge/", "markdown": "https://aiagentsformanufacturing.com/use-cases/vision-defect-inspection-on-the-edge.md", "department": "Quality", "department_url": "https://aiagentsformanufacturing.com/departments/quality/", "evidence": "Verified", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Quality", "problem": "Manual or rule-based visual inspection lets defects slip through. People get tired, shifts judge differently, and classic image-processing rules break when parts, lighting or surfaces vary. Every escape becomes a customer complaint, a return or a warranty claim, and inspectors become the bottleneck as line speed rises.", "how_ai_helps": "A camera on the line feeds a deep-learning model that runs on a small edge computer next to the machine, so decisions come back in milliseconds without the cloud. The model is trained on labeled images of good and bad parts from your own line, flags or rejects suspect parts, and sends results to the MES or PLC. Inspectors review the uncertain cases and the new defect images are used to retrain the model.", "result": "Seagate (hard-drive slider production, its own published case): replaced conventional image processing with machine learning on 34 automated vision machines, running on an in-house edge device chosen over workstations for cost. Outgoing defective parts per million fell 37%, from 243 to 153 by January 2025, and the downstream lot acceptance rate rose from a baseline of 86.8% to 97% across 35 shipments, with no loss of throughput. An unnamed mid-size Czech automotive-parts maker (12,000+ parts a day, two lines): a custom model trained on 50,000 labeled images, running on an NVIDIA Jetson module at 15 ms per part and linked to the MES, cut the defect escape rate from 4% to 0.3%; the €95K project saved about €180K a year, payback in about 6 months (integrator GK Servis's case study, so a vendor claim).", "companies": [ { "name": "Seagate", "url": "https://aiagentsformanufacturing.com/companies/#seagate" } ], "tools_deployed": [], "tools": [ "Edge computers: NVIDIA Jetson / IGX Orin modules, or an industrial PC with a GPU", "Industrial cameras and lighting (GigE Vision cameras from Cognex, Basler, Keyence and others)", "No-code vision platforms with edge deployment: Landing AI LandingLens, Cognex ViDi / In-Sight, Siemens Industrial Edge, Elementary", "Open-source path: Ultralytics YOLO or similar models with a labeling tool such as CVAT or Label Studio", "Integrators such as GK Servis for custom builds" ], "sources": [ "https://www.seagate.com/resources/seagate-boosts-automated-vision-system-with-ai/", "https://www.gkservis.com/case-studies/quality-inspection-automation.html" ], "date_added": "2026-10-12", "last_updated": "2026-10-26" }, { "title": "Anomaly detection on machine and process data to cut scrap", "url": "https://aiagentsformanufacturing.com/use-cases/anomaly-detection-to-cut-cost-of-quality/", "markdown": "https://aiagentsformanufacturing.com/use-cases/anomaly-detection-to-cut-cost-of-quality.md", "department": "Quality", "department_url": "https://aiagentsformanufacturing.com/departments/quality/", "evidence": "Verified", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Quality", "problem": "Scrap and rework often come from slow drifts nobody sees in time: a tool wearing out, a process parameter creeping, a test result that is borderline. Manual downtime logs and paper quality tests vary with who fills them in and hide root causes. In high-volume machining, a broken drill or worn tap is often noticed only after a batch of bad parts has been made. At Rosenberger, a high-frequency connector maker, some quality issues went undetected during manual inspection until machine-learning models were run on streaming line data.", "how_ai_helps": "Models learn the normal pattern of machine and process signals (spindle load, torque, temperatures, pressures, test results) and raise an alert when the pattern changes, before parts go out of tolerance. Alerts go to the operator or engineer with the signal that moved, and the history helps find root causes. People decide whether to stop, adjust or change a tool.", "result": "NYX (automotive injection molder using Milacron's large machines), Milacron M-Powered by ei3: online collection of quality tests and AI cause-and-effect pattern detection on machine data; scrap rate down 23% (fewer than 1 part in 100 out of standard), unplanned downtime down 10% and routine maintenance spending down 8–12%, with paper tracking for process verification abandoned (ei3 case study, a vendor claim). ORBIS (plastic packaging manufacturer), Spur Insights diagnostics (control-signal analysis, vibration sensors and energy data) on 22 injection molding machines: scrap down by half, from 2% to 1%, over 7,000 hours of unplanned downtime eliminated, 18,000 kWh saved per machine a year, and up to two months' warning of hydraulic pump failure; the scrap figure covers the whole diagnostics programme, not one alert type (Spur Insights case study, a vendor claim with a quote from ORBIS's VP of manufacturing). Schabmüller Automobiltechnik (Bavarian automotive supplier), Xplain Data causal analysis of production data with SW Machines: in a pilot on one line the scrap and rework rate fell from over 10% to just under 1.6% after the analysis showed the leak-test result depended on part temperature (so the wait between washing and testing was shortened) and found quality differences between casting nests in the supplied blanks; the supplier's managing director says the approach will be used on all production lines (SW Machines case study, April 2022, a vendor claim; this is root-cause analysis of recorded data, not live alerting). Rosenberger (high-frequency connection technology): machine-learning anomaly detection on streaming data from up to 120 data types per production line, which found quality issues earlier than manual inspection and, per the integrator, significantly reduced scrap; no figure is published (Device Insight case study, vendor claim). Bosch (own plants), Bosch Center for AI platform on Nexeed MES: sensor data from machines is analysed to detect anomalies early and reduce rejects, with associates deciding the action; pilot plants save one to two million euros a year, and at Hildesheim line cycle times fell 15% after process disruptions were found and removed; no scrap figure is published, and the rollout covered about 50 powertrain plants and 800+ lines in 2021 (Bosch press release, 3 March 2021, company-reported). Global automotive precision-machining supplier (name withheld), MachineMetrics Tool Anomaly Detection: high-frequency load, torque and spindle-speed data from the machine control, no added sensors; the company estimates each machine saves thousands of dollars a month in scrap, downtime and maintenance (MachineMetrics case study, vendor claim, no exact figure).", "companies": [ { "name": "Bosch", "url": "https://aiagentsformanufacturing.com/companies/#bosch" }, { "name": "NYX", "url": "https://aiagentsformanufacturing.com/companies/#nyx" }, { "name": "ORBIS", "url": "https://aiagentsformanufacturing.com/companies/#orbis" }, { "name": "Rosenberger", "url": "https://aiagentsformanufacturing.com/companies/#rosenberger" }, { "name": "Schabmüller Automobiltechnik", "url": "https://aiagentsformanufacturing.com/companies/#schabmuller-automobiltechnik" } ], "tools_deployed": [ { "name": "Device Insight (with Softing DataFEED)", "url": "https://aiagentsformanufacturing.com/tools/#device-insight-with-softing-datafeed" }, { "name": "MachineMetrics", "url": "https://aiagentsformanufacturing.com/tools/#machinemetrics" }, { "name": "Milacron M-Powered (ei3)", "url": "https://aiagentsformanufacturing.com/tools/#milacron-m-powered-ei3" }, { "name": "Spur Insights", "url": "https://aiagentsformanufacturing.com/tools/#spur-insights" }, { "name": "Xplain Data", "url": "https://aiagentsformanufacturing.com/tools/#xplain-data" } ], "tools": [ "Milacron M-Powered by ei3: used at NYX", "Spur Insights (iPID, wave field analysis, multi-modal diagnostics): used at ORBIS", "Xplain Data (causal analysis of production data), integrated by SW Machines: used at Schabmüller Automobiltechnik", "Device Insight data streaming with Softing DataFEED, Azure and Databricks machine-learning models: used at Rosenberger", "Bosch Center for AI manufacturing analytics on Bosch Nexeed MES: used at Bosch powertrain plants", "MachineMetrics Tool Anomaly Detection: used at the automotive supplier", "Other options: Sight Machine, Seeq, AVEVA PI with analytics, Siemens Insights Hub, Braincube", "Do-it-yourself: Python with an anomaly-detection library on historian data, or Excel control charts as a starting point" ], "sources": [ "https://ei3.com/case-studies/nyx-scrap-rate-ei3-ai-optimization", "https://www.machinemetrics.com/blog/tool-anomaly-detection-case-study", "https://spurinsights.com/case_studies/orbis/", "https://xplain-data.de/wp-content/uploads/2022/07/SW_CaseStudy_Final_EN_220722.pdf", "https://device-insight.com/wp-content/uploads/2025/02/Data-Driven-Factory_Use-Case_Rosenberger_EN.pdf", "https://www.bosch-presse.de/pressportal/de/en/on-the-path-toward-zero-defect-production-with-bosch-ai-225472.html" ], "date_added": "2026-10-09", "last_updated": "2026-10-26" }, { "title": "8D / NCR drafting and root-cause hypotheses", "url": "https://aiagentsformanufacturing.com/use-cases/8d-ncr-drafting-and-root-cause-hypotheses/", "markdown": "https://aiagentsformanufacturing.com/use-cases/8d-ncr-drafting-and-root-cause-hypotheses.md", "department": "Quality", "department_url": "https://aiagentsformanufacturing.com/departments/quality/", "evidence": "Pilot", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Quality", "problem": "A customer or internal nonconformance needs an 8D fast, but D2 (problem definition) and D4 (root cause analysis) take hours and depend on experience. Much of that time is just searching: in a 2026 pilot, engineers spent on average 47 minutes per case finding old reports, data and notes (PDFs, old Excel sheets, calling colleagues) before the analysis could even start. Knowledge also walks out the door: at one pilot partner, a Tier-2 CNC machining supplier, a team spent three months looking for the cause of a dimensional deviation that had been documented in an 8D two years earlier; the engineer who wrote it had left and nobody knew the report existed.", "how_ai_helps": "An AI tool finds similar past cases in seconds, drafts the D2 problem description, and proposes possible root causes with their sources. Engineers decide which cause is real and which test proves it.", "result": "Pilot by Galileon (maker of the SolveR tool), Jan–Mar 2026: 8 Slovenian manufacturers (70 to 1,700 employees; Tier-1/Tier-2 automotive suppliers, aluminium die casting and CNC machining, electrical engineering, polymer injection moulding, machine building) each solved real 8D cases twice, first manually, then with AI, same engineers, timed with a stopwatch. 10 cases; defects: casting porosity, weld defects, missing components, incomplete assembly. Measured averages: D2 went from 43.5 min (range 27.0–55.5) to 4.6 min (3.7–7.1), 9.6x faster as Galileon states it (the rounded averages shown give about 9.5x); D4 went from 239 min (171–295) to 31.7 min (21.4–40.1), 7.5x faster; D2+D4 together went from 283 to 36.3 min, 7.8x faster. The AI proposed 8 to 16 possible causes per case; engineers kept 20–30%, and the real root cause was always among those kept. Calculated, not measured: Galileon's model puts a full 8D at 43.1 engineering hours today vs 22.5 with AI (48% less, stated as 45%), assuming a 5-person team and no savings outside D2/D4. Limits of the study: only 10 cases; the manual run always came first, so engineers already knew each case when they used the AI; all partners are in Slovenia; the study's author co-founded Galileon and discloses it.", "companies": [], "tools_deployed": [ { "name": "SolveR (Galileon)", "url": "https://aiagentsformanufacturing.com/tools/#solver-galileon" } ], "tools": [ "SolveR by Galileon (used in the pilot)", "Any general LLM (ChatGPT, Claude, Copilot) for drafting with the prompt below", "Your archive of past 8D/NCR reports" ], "sources": [ "https://galileon.si/results", "https://ieomsociety.org/proceedings/milwaukee2026/191.pdf" ], "date_added": "2026-10-11", "last_updated": "2026-10-26" }, { "title": "AI-assisted deviation investigations, root cause and CAPA", "url": "https://aiagentsformanufacturing.com/use-cases/ai-deviation-investigation-root-cause-capa/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-deviation-investigation-root-cause-capa.md", "department": "Quality", "department_url": "https://aiagentsformanufacturing.com/departments/quality/", "evidence": "Pilot", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Quality", "problem": "Every deviation or nonconformance needs an investigation: assess product impact, find the root cause, look for trends and write a CAPA (corrective and preventive action) that will hold up with customers and auditors. The knowledge to do it well exists in years of past investigations, but investigators cannot search it quickly, classifications can differ between sites, and sites that work in languages other than English still owe customers a clear investigation report. Catalent, a contract drug manufacturer (CDMO) with nearly 40 sites, describes investigations as among its most critical and resource-intensive processes.", "how_ai_helps": "An assistant searches the company's own history of deviations and CAPAs (retrieval over its quality records), shows similar past events and what fixed them, suggests a severity classification, supports root-cause analysis with evidence from past events, and helps produce a first investigation draft; in the Catalent case the drafts are in English for sites that work in other languages (the source does not mention other target languages). It has read-only access and never creates or changes quality records; a quality expert validates every output.", "result": "Catalent (contract drug manufacturer, nearly 40 sites on four continents), Qai built on Azure OpenAI (GPT-4 and GPT-4o) with Microsoft (Sept 2026): retrieval over the network's deviation and CAPA history, consistent minor/major/critical classification across sites, first drafts in English for non-English sites, human validation documented for every output, read-only and no creation or modification of GMP (good manufacturing practice) records. Rollout was nearly complete, with early signs of productivity gains in deviation management; Catalent says the figures are still being measured, and audit management is next. No percentage has been published yet (Microsoft customer story).", "companies": [ { "name": "Catalent", "url": "https://aiagentsformanufacturing.com/companies/#catalent" } ], "tools_deployed": [], "tools": [ "Catalent Qai on Microsoft Azure OpenAI / Foundry, Azure App Service, Microsoft Fabric", "QMS platforms adding AI: ETQ, MasterControl, Veeva, Sparta TrackWise, Intellect, Qualio", "Any enterprise LLM with access to your past 8D/CAPA reports for a pilot" ], "sources": [ "https://www.microsoft.com/en/customers/story/27171-catalent-azure-openai-in-foundry-models" ], "date_added": "2026-10-13", "last_updated": "2026-10-26" }, { "title": "Predictive maintenance from vibration sensors, planned in the CMMS", "url": "https://aiagentsformanufacturing.com/use-cases/predictive-maintenance-vibration-sensors-cmms/", "markdown": "https://aiagentsformanufacturing.com/use-cases/predictive-maintenance-vibration-sensors-cmms.md", "department": "Maintenance", "department_url": "https://aiagentsformanufacturing.com/departments/maintenance/", "evidence": "Verified", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Maintenance & reliability", "problem": "Rotating equipment (pumps, gearboxes, motors, fans) often fails without obvious warning, and an unplanned stop on an extrusion or process line costs hours of repair plus hours to bring the line back to temperature. Route-based checks miss problems between visits, and maintenance teams end up reacting at night or on weekends. Fiberon estimated one avoided melt-pump failure alone at $56,000 and 16 hours of downtime.", "how_ai_helps": "Wireless sensors measure vibration, temperature and sometimes magnetic field or oil condition. AI compares the readings with known failure patterns and, often with a reliability expert, issues an alert naming the likely fault (bearing wear, looseness, misalignment). The alert becomes a work order in the CMMS, and the repair is planned in the next shutdown or low-impact window.", "result": "Fiberon (decking maker, Fortune Brands; New London, NC plant), Augury Machine Health on 40 machines, Dec 2023–Aug 2024: $274,000 in estimated savings, 178 hours of downtime avoided and 2.5x return in eight months, with a 96% response rate to alerts. One alert in March 2024 on the L20 melt-pump gearbox (looseness and bearing wear) let the team replace the motor and gearbox during a planned shutdown. Fortune Brands then began a 12-month roll-out (from October 2024) targeting 1,000+ machines across 16 sites in the U.S., UK and Mexico; at the halfway point about two-thirds of the machines were deployed, and the article (Plant Services, Aug 2025) gives no result for the roll-out itself. A Plante Moran example: a model spotted small speed drops about two weeks before bearing failures, enough time to get parts and schedule repairs in planned downtime.", "companies": [ { "name": "Fiberon", "url": "https://aiagentsformanufacturing.com/companies/#fiberon" } ], "tools_deployed": [ { "name": "Augury", "url": "https://aiagentsformanufacturing.com/tools/#augury" } ], "tools": [ "Augury Machine Health: used at Fiberon", "Other condition-monitoring platforms: AssetWatch, Tractian, Waites, Siemens Senseye, SKF", "CMMS to receive work orders: SAP PM, IBM Maximo, Fiix, UpKeep, MaintainX, Limble" ], "sources": [ "https://www.plantservices.com/predictive-maintenance/article/55307059/ai-powered-maintenance-cuts-downtime-and-costs-for-fiberons-manufacturing-operations", "https://www.plantemoran.com/explore-our-thinking/insight/2026/06/ai-use-cases-in-manufacturing" ], "date_added": "2026-10-12", "last_updated": "2026-10-26" }, { "title": "Agentic work-order drafting and repair-window scheduling", "url": "https://aiagentsformanufacturing.com/use-cases/agentic-work-order-drafting-and-repair-window-scheduling/", "markdown": "https://aiagentsformanufacturing.com/use-cases/agentic-work-order-drafting-and-repair-window-scheduling.md", "department": "Maintenance", "department_url": "https://aiagentsformanufacturing.com/departments/maintenance/", "evidence": "Pilot", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Maintenance & reliability", "problem": "The data to predict failures is usually already there, but in separate systems: live sensor readings in SCADA, repair history and spare parts in the ERP, the production schedule in the MES. Nobody can see on one screen which machine is about to fail, so maintenance follows the calendar until something breaks. Then a technician is pulled off another job, the repair is rushed, and the order behind it misses its ship date.", "how_ai_helps": "An anomaly model watches equipment data. When it flags a problem, a first agent checks the ERP for open work orders and parts, pulls the service instructions and the technician roster, and drafts a work order naming the likely failure and the parts needed. A second agent checks the MES for open orders, deadlines and spare capacity on parallel lines and recommends the repair window that least disrupts customer commitments. Agents can only read and recommend; an engineer approves every work order and schedule change, and each step is logged.", "result": "European industrial-pump manufacturer (3 plants, 14 lines, 2,500+ pump configurations), built by GroupBWT on SAP ERP, Siemens Opcenter MES, SCADA and existing IoT sensors (2026): over six months, unplanned downtime hours fell 31% versus the previous six months, emergency repairs fell 27% and maintenance spend fell 18% as condition-based maintenance replaced part of the fixed calendar. The client then extended the system to quality and production planning at all three plants. These are the integrator's published figures for an unnamed client.", "companies": [], "tools_deployed": [], "tools": [ "Built by GroupBWT on SAP ERP (repair history, parts, work orders), Siemens Opcenter MES (schedule) and SCADA/PLC sensor data", "Condition-monitoring platforms that can feed it: Augury, Siemens Senseye, AssetWatch, Tractian", "CMMS with APIs for work-order drafts: SAP PM, IBM Maximo, Fiix, UpKeep, MaintainX", "For a manual first pass: any LLM with the prompt below, fed an alert, the asset history and next week's schedule" ], "sources": [ "https://groupbwt.com/case/ai-agents-for-predictive-maintenance-manufacturer/" ], "date_added": "2026-10-07", "last_updated": "2026-10-14" }, { "title": "Troubleshooting copilot over manuals and work-order history", "url": "https://aiagentsformanufacturing.com/use-cases/troubleshooting-copilot-over-manuals/", "markdown": "https://aiagentsformanufacturing.com/use-cases/troubleshooting-copilot-over-manuals.md", "department": "Maintenance", "department_url": "https://aiagentsformanufacturing.com/departments/maintenance/", "evidence": "Verified", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Maintenance & reliability", "problem": "When a machine stops, technicians lose time finding the right manual, drawing or past repair, sometimes leaving the floor to locate, translate and print documents. At ArcelorMittal Hunedoara this took 43 minutes per search. Experienced technicians retire, newer ones lack the know-how, and the answers sit in thousands of pages and old work orders nobody can search.", "how_ai_helps": "An assistant indexes the plant's manuals, procedures, drawings, spare-parts lists and work-order history. A technician describes the symptom in plain language (on a phone, in their own language) and gets likely causes, step-by-step checks and the parts involved, each linked to the source page so they can verify it.", "result": "ArcelorMittal Hunedoara (rolling mills, Romania), Tacit AI: 2,000+ manuals and 3,000+ CAD drawings connected; a 43-minute document workflow now takes under a minute, 1,000+ engineering hours recovered a year, and answer accuracy improved from 73% to over 96% with feedback from mechanics. Maximal Group (cardboard packaging, Romania), Tacit AI: 59% faster repair cycles (both Tacit AI case studies). A leading tire manufacturer, C3 Generative AI over 1,000+ documents plus live SAP work orders and spare parts: average repair time cut by up to 50% (C3 AI case study). All vendor-published.", "companies": [ { "name": "ArcelorMittal", "url": "https://aiagentsformanufacturing.com/companies/#arcelormittal" } ], "tools_deployed": [ { "name": "C3 Generative AI", "url": "https://aiagentsformanufacturing.com/tools/#c3-generative-ai" }, { "name": "Tacit AI", "url": "https://aiagentsformanufacturing.com/tools/#tacit-ai" } ], "tools": [ "Tacit AI: used at ArcelorMittal Hunedoara and Maximal Group", "C3 Generative AI for Equipment Troubleshooting: used at the tire manufacturer", "Others: Siemens Industrial Copilot for operations, Aquant, MaintainX and Fiix AI assistants, Microsoft Copilot Studio over SharePoint documents" ], "sources": [ "https://tacitx.ai/case-study/arcelormittal/", "https://tacitx.ai/case-study/maximal-group/", "https://c3.ai/customers/leading-tire-manufacturer-reduces-equipment-troubleshooting-time-with-c3-generative-ai" ], "date_added": "2026-10-09", "last_updated": "2026-10-14" }, { "title": "Demand forecasting with external signals such as weather and local events (Bimbo Bakeries USA, Paulaner, Cranswick)", "url": "https://aiagentsformanufacturing.com/use-cases/demand-forecasting-with-external-signals/", "markdown": "https://aiagentsformanufacturing.com/use-cases/demand-forecasting-with-external-signals.md", "department": "Logistics / Supply Chain", "department_url": "https://aiagentsformanufacturing.com/departments/logistics/", "evidence": "Verified", "summary": "AI demand forecasting with weather and local events: Bimbo Bakeries USA cut forecast errors up to 30%; Paulaner reached 90% short-term accuracy.", "specialized": false, "who_this_is_for": null, "role": "Logistics / supply chain (demand planning and S&OP)", "problem": "Demand for many products swings with things your own sales history does not show: weather, local events, promotions, availability on the shelf, inflation and consumer confidence. When a producer plans only from past orders, it overproduces in some weeks and runs short in others. For fresh or short-life products that means waste and lost sales; for the producer upstream it means surplus raw material, as Cranswick saw with a glut of pigs while demand fell. Bimbo Bakeries USA, with 59 bakeries and 11,000 delivery routes, had to walk a fine line between waste from overstock and lost sales from out-of-stocks, with local demand made highly variable by weather, events and store operations, and with planning still done partly with desktop calculators and spreadsheets. Paulaner's brewery in central Munich has so little storage space that it served customers with as little as one day of inventory, so its forecasts had to be accurate down to the package level, and it had never used a forecasting system before.", "how_ai_helps": "A forecasting tool combines the producer's own data (shipments, orders, promotions) and, where customers share it, their sales and availability, with outside signals such as weather, temperature, holidays and local events. Machine learning or statistical models measure how much each signal moves demand and forecast at the level where decisions are made: SKU, store and week at Bimbo, package and day at Paulaner. Some tools add supply sensing and propose planning actions to close the gap between supply and demand. Planners and sales people review and adjust the forecast before it drives production and orders.", "result": "Bimbo Bakeries USA (largest US bakery company, 59 bakeries, over 20,000 associates), antuit.ai (now part of Zebra Technologies) inside its in-house 'Ion' program, with weather, local events, store stocking constraints and point-of-sale data: up to 30% reduction in forecast errors, forecast accuracy above 80% achieved and maintained over five years that included the pandemic (MAPE over six-week intervals), forecasting scaled from a sample to 11,000 routes within months, and changes that once needed four months made in as little as three weeks (vendor success story with Bimbo's VP of DSD Center of Excellence). Paulaner (Munich brewery, 3 brands, 20 beers, up to 30 packaging types), Eyelit Technologies Collaborative Demand Planning with weather, temperature, season and holiday factors chosen with the company's most experienced people: live within four months, short-term forecast accuracy (6-week lag) of 90% and long-term accuracy (18-month lag) of 85%, with better customer service despite the storage limit (vendor case study; no baseline given). Cranswick (UK pork producer) with Sainsbury's and the start-up Singular Intelligence, in a funded proof of concept under the Made Smarter Innovation Digital Supply Chain Hub, on Sainsbury's Taste the Difference pork range: Cranswick estimated that the oversupply of pork sent to the retailer fell by 15.25%; for Sainsbury's, availability improved by 2-5% and food waste fell by 50-70% on some products only (not across the range). This was a pilot on one product range.", "companies": [ { "name": "Bimbo Bakeries USA", "url": "https://aiagentsformanufacturing.com/companies/#bimbo-bakeries-usa" }, { "name": "Cranswick", "url": "https://aiagentsformanufacturing.com/companies/#cranswick" }, { "name": "Paulaner", "url": "https://aiagentsformanufacturing.com/companies/#paulaner" } ], "tools_deployed": [ { "name": "antuit.ai (Zebra Technologies)", "url": "https://aiagentsformanufacturing.com/tools/#antuit-ai-zebra-technologies" }, { "name": "CatmanAI (Singular Intelligence)", "url": "https://aiagentsformanufacturing.com/tools/#catmanai-singular-intelligence" }, { "name": "Eyelit Technologies", "url": "https://aiagentsformanufacturing.com/tools/#eyelit-technologies" } ], "tools": [ "antuit.ai, now part of Zebra Technologies (AI demand forecasting and predictive ordering): used at Bimbo Bakeries USA", "Eyelit Technologies Collaborative Demand Planning (statistical forecast with weather and other factors, Excel interface for sales): used at Paulaner", "CatmanAI by Singular Intelligence: used in the Cranswick / Sainsbury's pilot", "Other demand-planning tools with external signals: ToolsGroup, Kinaxis, o9, RELEX, Pecan, and the forecasting modules of most ERPs", "Free weather and calendar data: national weather services, public-holiday calendars", "Do-it-yourself: Excel (FORECAST.ETS) or Python (pandas + statsmodels or Prophet) to build and backtest the forecast; an LLM only to help write the code and explain results" ], "sources": [ "https://www.zebra.com/content/dam/zebra_dam/en/success-story/manufacturing-success-story-bimbo-bakeries-usa-en-us.pdf", "https://eyelit.ai/paulaner-brewery-weather-forecast-accuracy-demand-planning-software-case-study/", "https://www.madesmarter.uk/media/1nrnycg5/cranswick-case-study.pdf" ], "date_added": "2026-10-08", "last_updated": "2026-10-26" }, { "title": "Bulk HS-code classification for your whole item master", "url": "https://aiagentsformanufacturing.com/use-cases/bulk-hs-code-classification-item-master/", "markdown": "https://aiagentsformanufacturing.com/use-cases/bulk-hs-code-classification-item-master.md", "department": "Logistics / Supply Chain", "department_url": "https://aiagentsformanufacturing.com/departments/logistics/", "evidence": "Verified", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Logistics & trade compliance", "problem": "Hundreds or thousands of part numbers in the ERP have no HS code, or an old one nobody can justify. Every urgent export or import then waits while someone digs through datasheets and the tariff, and two people often pick two different codes for the same item. It is slow work: at Ingram Micro's German branch, classifying one new item by hand could take up to 30 minutes, for about 1,250 new items a month. A wrong code means wrong duties, lost CUSMA/USMCA preference, delays, and penalties or reassessments later.", "how_ai_helps": "Export the item master, let AI propose a 6-digit HS code plus the national 8/10-digit tariff line (Canada or US) for every item, with its reasoning under the General Rules of Interpretation (GRI) and a confidence level. High-confidence items go to a fast review, low-confidence or ambiguous items go to a customs specialist or broker, and the approved codes are written back to the ERP with the justification kept as an audit trail.", "result": "ZwickRoell (materials-testing machines, 1,800 employees, Germany): in production since July 2025 for single and bulk classification, over 3,000 products classified, 84% of AI suggestions adopted, each with a full justification and source references (traide AI case study, July 2026). MINDA Industrieanlagen (conveyor and plant engineering, 500–999 employees): 50% less time spent on daily classification; bulk classification of master data fed into the ERP (traide AI, June 2025). Ingram Micro Distribution (Germany, an IT distributor/importer rather than a manufacturer): about 1,250 new items a month; classification time down from about 30 minutes to 5–8 minutes per item; up to 80% of AI suggestions accepted without rework; savings of about half a full-time employee (traide AI, Oct 2025). Kaemingk (seasonal decoration products, selling to 90+ countries): about 9,000 products to classify every six months; Squadra's PowerClass.ai now classifies about 18,000 products a year in batches of about 3,000, with about 90% accuracy (the source writes '±90%') and an average of 675 hours of manual classification saved per batch (case published on the PowerSuite.ai blog, 21 Jan 2025; PowerClass.ai is a Squadra / PowerSuite product). Reality check: in an independent 2025 benchmark built from 200 US CBP binding rulings, the best general model (GPT-5-Thinking) got only 25% of 10-digit codes fully right and 55.5% at 6 digits; a fine-tuned model reached 40% and 57.5%. Without your own specs and a human review step, raw chatbot codes are not good enough. All company figures come from the vendors' published case studies.", "companies": [ { "name": "ZwickRoell", "url": "https://aiagentsformanufacturing.com/companies/#zwickroell" } ], "tools_deployed": [ { "name": "PowerClass.ai", "url": "https://aiagentsformanufacturing.com/tools/#powerclass-ai" }, { "name": "traide AI", "url": "https://aiagentsformanufacturing.com/tools/#traide-ai" } ], "tools": [ "traide AI (traideCompliance): used at ZwickRoell, MINDA and Ingram Micro", "PowerClass.ai (Squadra / PowerSuite.ai): used at Kaemingk", "Other AI classification tools: Gaia Dynamics (Canadian 10-digit), iCustoms.ai, and the classification modules of global trade platforms", "Any general LLM (ChatGPT, Claude, Copilot) with the prompt below for a first pass on a small batch", "Free official references: CBSA Customs Tariff (https://www.cbsa-asfc.gc.ca/trade-commerce/tariff-tarif/menu-eng.html), USITC HTS search (https://hts.usitc.gov/), CBP CROSS rulings database (https://rulings.cbp.gov/home)" ], "sources": [ "https://traide.ai/en/success-stories/proaktiv-statt-reaktiv-wie-zwickroell-ki-gest%C3%BCtzte-zolltarifierung-einf%C3%BChrt-bevor-der-druck-entsteht", "https://traide.ai/en/success-stories/effiziente-und-rechtssichere-tarifierung-bei-minda-mit-traide-ai", "https://traide.ai/en/success-stories/standardisiert-statt-subjektiv-wie-ingram-micro-mit-traide-ai-die-produktklassifizierung-vereinheitlichte", "https://powersuite.ai/blog/streamlining-hs-code-classification-with-ai", "https://arxiv.org/abs/2509.18400", "https://www.cbsa-asfc.gc.ca/trade-commerce/tariff-tarif/menu-eng.html", "https://hts.usitc.gov/", "https://rulings.cbp.gov/home" ], "date_added": "2026-10-10", "last_updated": "2026-10-14" }, { "title": "AI quoting from drawings and RFQ emails", "url": "https://aiagentsformanufacturing.com/use-cases/ai-quoting-from-drawings-and-rfq-emails/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-quoting-from-drawings-and-rfq-emails.md", "department": "Engineering", "department_url": "https://aiagentsformanufacturing.com/departments/engineering/", "evidence": "Verified", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Sales & customer service", "problem": "In many job shops one or two experienced people do all the quoting: they read the drawing, estimate material, machine or press time and tooling, set the margin and type the proposal. At a 45-person metal stamper with about $12M revenue, the CEO spent 15–20 hours a week quoting, answers to prospects were slow, and RFQs were turned down in busy periods because nobody else could estimate. The pricing logic lived in his head, and finding similar past jobs to reuse could take 30 minutes per search.", "how_ai_helps": "AI reads the RFQ email, the drawing and the specs and pulls out the facts that drive cost: material, dimensions, tolerances, threads, finishes, quantities and operations. It finds similar parts you have already made and what they really cost. A rule-based cost model, written with your estimator, calculates material, machine or press time, tooling and secondary operations, and a person reviews and approves the quote before it goes out.", "result": "Metal stamping company (~45 people, ~$12M revenue), system built by DMexec: AI extraction from drawings, prints and RFQ emails plus a deterministic cost model and an approval queue; about 90% faster quote setup and 65% faster quote delivery, and junior staff now prepare quotes the CEO used to do (DMexec case study). Thermoforming manufacturer (100+ employees, Germany), built by SimplifieD Solutions in six weeks: time per quote from 2–4 hours to under 30 minutes, with similar past projects found automatically and anyone in sales able to quote (March 2025 case study). Laserhub (online manufacturing platform for sheet-metal and machined parts) with Werk24 drawing extraction: engineers had spent 30–90 minutes per drawing; extraction now takes seconds, quotes are 30% faster and accuracy on price-driving features is 98% (Werk24 case study). A high-precision metal and plastic parts maker using CADDi: searching past drawings and quotes went from up to 30 minutes to 1–2 minutes (CADDi, 2026). All figures are vendor-published.", "companies": [], "tools_deployed": [ { "name": "Werk24", "url": "https://aiagentsformanufacturing.com/tools/#werk24" } ], "tools": [ "DMexec (custom quoting system for a stamper)", "SimplifieD Solutions (custom AI quoting for a thermoformer)", "Werk24 (API that reads 2D drawings into structured data): used at Laserhub", "CADDi Drawer (drawing similarity search across past jobs)", "Quoting platforms for job shops: Paperless Parts, Toolpath, and the estimating modules of JobBOSS, ProShop, E2 / Shoptech", "Any LLM with vision (ChatGPT, Claude, Copilot) with the prompt below to test drawing extraction" ], "sources": [ "https://dmexec.com/use-cases/ai-quoting-metal-manufacturing-founder-bottleneck", "https://www.simplified-solutions.de/en/case-studies/ai-quote-generation-thermoforming/", "https://werk24.io/case-studies/laserhub", "https://us.caddi.com/case-studies/a-leading-manufacturer" ], "date_added": "2026-10-07", "last_updated": "2026-10-14" }, { "title": "PLC code drafting and explanation (structured text yes, ladder with caution)", "url": "https://aiagentsformanufacturing.com/use-cases/plc-code-drafting-and-explanation/", "markdown": "https://aiagentsformanufacturing.com/use-cases/plc-code-drafting-and-explanation.md", "department": "Engineering", "department_url": "https://aiagentsformanufacturing.com/departments/engineering/", "evidence": "Method", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Engineering", "problem": "Controls engineers are scarce and spend much of their time on repetitive work: boilerplate function blocks, I/O mapping, alarm lists, HMI screens and documentation. Old programs written by someone who left are hard to read. General chatbots can write PLC code that looks correct but often does not compile or behaves wrongly on a real machine, because the details that matter (vendor dialect, timer behaviour, scan order, addressing) are invisible when you just read the code.", "how_ai_helps": "An AI assistant drafts structured text (SCL/ST) or ladder logic from a plain-language description with the exact I/O list and wiring, explains existing code rung by rung, and generates comments, alarm texts and HMI screens. Integrated tools such as Siemens' copilot work inside the engineering software and can create code and visualisations directly in the project. The engineer compiles, simulates and tests every output before it touches a machine.", "result": "Siemens Industrial Copilot for TIA Portal (available since July 2024): Siemens says more than 100 companies were using it by late 2024, including Schaeffler and thyssenkrupp Automation Engineering, which planned a global roll-out from early 2025 to write SCL code, integrate it into TIA Portal projects and generate WinCC Unified screens. Siemens' stated figures: generated code needs only about 20% adaptation, and panel visualisations take about 30 seconds. The two Siemens products are separate: the 100+ companies, Schaeffler and thyssenkrupp above belong to Industrial Copilot. For its newer Eigen Engineering Agent, Siemens reports (Thailand launch release, 27 Aug 2026) use at 100+ companies in 19 countries, including ANDRITZ Metals, CASMT and Prism Systems, and engineering workflows 2–5 times faster than manual, across tasks that include PLC programming, HMI and device configuration, not only PLC code (all vendor figures; in the sources we read, no customer publishes its own measured savings). Independent test (one small task): ACC Automation gave ChatGPT and Claude the same motor start/stop task with a pre-start buzzer and normally-closed stop and overload wiring for a CLICK PLC; both produced logic judged correct and handled the normally-closed trap, because the prompt gave the exact I/O and wiring. The author then tested one of the programs on a connected PLC (CLICK has no offline simulator) and still advises a hardwired overload backstop. One motor example says nothing about success rates on larger programs. LADX's engineers explain why general models still fail: mixed vendor dialects, wrong timer types, ignoring scan order, and wrong address formats, which is why every output must be compiled and checked.", "companies": [ { "name": "Schaeffler", "url": "https://aiagentsformanufacturing.com/companies/#schaeffler" }, { "name": "thyssenkrupp", "url": "https://aiagentsformanufacturing.com/companies/#thyssenkrupp" } ], "tools_deployed": [ { "name": "Siemens Industrial Copilot", "url": "https://aiagentsformanufacturing.com/tools/#siemens-industrial-copilot" } ], "tools": [ "Siemens Industrial Copilot for TIA Portal: used at Schaeffler and thyssenkrupp Automation Engineering (the newer Eigen Engineering Agent is a separate Siemens product, cited at ANDRITZ Metals, CASMT and Prism Systems)", "Rockwell FactoryTalk Design Studio Copilot, Beckhoff TwinCAT Chat, CODESYS AI assistants", "LADX (PLC code generation with a compile-and-check loop)", "General LLMs (ChatGPT, Claude, Copilot) for structured text, explanations and documentation", "To test before hardware, use a simulator for your PLC family: for example PLCSIM for Siemens (check the licence for your version) or the CODESYS simulation. Note that the free CLICK programming software has no offline simulator: ACC Automation tested on a connected CLICK PLC, after checking the logic shape in a free online ladder simulator" ], "sources": [ "https://press.siemens.com/jp/en/pressrelease/pr-20241024", "https://www.automationworld.com/factory/digital-transformation/news/55248289/thyssenkrupp-adopts-siemens-industrial-copilot", "https://news.siemens.com/en-us/eigen-launch-thailand/", "https://accautomation.ca/can-ai-actually-write-plc-code-i-put-it-to-the-test/", "https://ladx.ai/resources/why-chatgpt-ladder-logic-fails" ], "date_added": "2026-10-10", "last_updated": "2026-10-26" }, { "title": "AI accounts payable invoice processing", "url": "https://aiagentsformanufacturing.com/use-cases/ai-accounts-payable-invoice-processing/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-accounts-payable-invoice-processing.md", "department": "Office & management", "department_url": "https://aiagentsformanufacturing.com/departments/office/", "evidence": "Verified", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Finance & AP", "problem": "Every supplier sends a differently formatted invoice. Someone has to open each PDF, read it, find the matching purchase order, check quantities, prices and tax, and key it into the ERP. Template-based OCR breaks on the variety. At Aralab, a three-person finance team spent years doing this for more than 2,000 invoices a month instead of cash-flow forecasting or supplier negotiation; Nordell, an 80-person plastics moulder, saw the same paperwork load as a brake on growth.", "how_ai_helps": "AI reads each invoice whatever the layout, extracts supplier, invoice number, lines, quantities, prices and tax, and matches them to the purchase order and receipt in the ERP. Rules then recalculate every line total and tax figure independently. Invoices that match go straight through; mismatches and low-confidence lines are routed to a person with the reason shown. Every step is logged so finance can audit what the system did.", "result": "Measured (vendor-reported): Aralab (manufacturer, Portugal), agent built by Twistag with Anthropic Claude for reading and deterministic code for the arithmetic: more than 2,000 supplier invoices a month; about 80% need no human touch and the ambiguous 20% go to a review screen; the three finance staff moved from data entry to forecasting, negotiation and planning; built by one engineer and live in six weeks (Twistag case study). Expected, not yet measured: Nordell (plastic injection moulder, West Sussex, 80+ staff): a £34,000 project, half funded by a £17,000 Made Smarter grant, for Demi AP integrated with its DELMIAWorks ERP; it captures invoice data, checks it against purchase orders and flags discrepancies; it is expected to save more than 60 hours of administration a month (forecast announced July 2026). See also the fastener maker case for a touchless-posting example.", "companies": [ { "name": "Aralab", "url": "https://aiagentsformanufacturing.com/companies/#aralab" } ], "tools_deployed": [ { "name": "Demi AP", "url": "https://aiagentsformanufacturing.com/tools/#demi-ap" } ], "tools": [ "Demi AP integrated with DELMIAWorks ERP: used at Nordell", "Custom agent on Anthropic Claude, Google Cloud and LangFuse (for tracing each AI decision), built by Twistag: used at Aralab", "Other AP automation tools: Esker, Rossum, Medius, Basware, Tipalti, Stampli, and the AP modules of many ERPs", "Any LLM with the prompt below to test extraction on a handful of invoices" ], "sources": [ "https://twistag.com/case-studies/ai-invoice-automation-aralab-manufacturing", "https://www.madesmarter.uk/resources/case-study-nordell/" ], "date_added": "2026-10-07", "last_updated": "2026-10-14" }, { "title": "Customer PO and quote emails → sales orders in the ERP (drafts or automated entry) (RUKO, Gems Setra, Stelrad, Siemens Healthineers)", "url": "https://aiagentsformanufacturing.com/use-cases/customer-po-emails-to-erp-sales-orders/", "markdown": "https://aiagentsformanufacturing.com/use-cases/customer-po-emails-to-erp-sales-orders.md", "department": "Office & management", "department_url": "https://aiagentsformanufacturing.com/departments/office/", "evidence": "Verified", "summary": "AI turns customer PO emails into ERP sales orders: RUKO saved 87% of entry time, Stelrad 64% touchless, Siemens Healthineers 8% to 85% automated.", "specialized": false, "who_this_is_for": null, "role": "Sales & customer service", "problem": "Customer purchase orders arrive by email as PDFs, spreadsheets or scans in every layout and language, and customer service re-keys each one into the ERP by hand. Customers' part numbers rarely match yours, so every line has to be looked up, and the time spent typing is time not spent answering customers. At Siemens Healthineers Italy, a 12-person customer service team was manually entering 27,000 orders a year, representing 300,000 line items in very different formats, filling every SAP field by hand. At Gems Setra, the average order took more than 7 hours to process, including engineering validation and pricing approval, against a goal of four business hours.", "how_ai_helps": "Two levels. (1) Simple: an LLM reads each order email or PDF and prepares a draft sales order that a customer service rep checks and keys or approves. (2) Integrated: a document-AI tool such as Workist, Conexiom or Esker reads the order, learns each customer's format and product codes, checks it against master data (customers, part numbers, price lists, VAT numbers) and creates the order in the ERP automatically, sending only exceptions to a rep. Customers keep sending orders the way they always have. The results below come from level 2.", "result": "RUKO Präzisionswerkzeuge (over 250 employees, drill and countersink maker, Germany), Workist: from 70 full working days of manual order entry per quarter to 9 days, an 87% time saving, including an automatic check of EU VAT numbers (vendor success story). Gems Setra (sensing and monitoring products; part of Ralliant since 2025, while Conexiom's undated case study still says Fortive), Conexiom: average order processing from over 7 hours to under 2.5 hours (65% faster), 5.5 to 2.9 minutes of handling per order, 36% of orders touchless, customer service caseload from 600 to 400 a day with nobody replaced, and more same-day shipments; the company reports higher NPS and voice-of-customer scores (vendor case study with the director of customer operations). Stelrad (UK radiator maker, three manufacturing lines in South Yorkshire), Esker: 64% of orders processed touchless and average order entry time down to 55 seconds (vendor blog). Siemens Healthineers Italy (laboratory diagnostics), Esker with SAP: automated orders from 8% to 85%, with 93% of order fields correctly captured, and customers' product codes mapped to Siemens codes automatically (vendor case study).", "companies": [ { "name": "Gems Setra", "url": "https://aiagentsformanufacturing.com/companies/#gems-setra" }, { "name": "RUKO", "url": "https://aiagentsformanufacturing.com/companies/#ruko" }, { "name": "Siemens Healthineers", "url": "https://aiagentsformanufacturing.com/companies/#siemens-healthineers" }, { "name": "Stelrad", "url": "https://aiagentsformanufacturing.com/companies/#stelrad" } ], "tools_deployed": [ { "name": "Conexiom", "url": "https://aiagentsformanufacturing.com/tools/#conexiom" }, { "name": "Esker", "url": "https://aiagentsformanufacturing.com/tools/#esker" }, { "name": "Workist", "url": "https://aiagentsformanufacturing.com/tools/#workist" } ], "tools": [ "Workist (automated order entry into the ERP): used at RUKO", "Conexiom (AI sales order automation, no templates): used at Gems Setra", "Esker Order Management with Synergy AI: used at Stelrad and Siemens Healthineers", "Other tools to evaluate: Rossum and the order-capture modules of your ERP", "Any LLM (ChatGPT, Claude, Copilot) with the prompt below for level 1 drafts" ], "sources": [ "https://www.workist.com/en/success-stories/ruko", "https://conexiom.com/gems-setra", "https://www.esker.com/en-gb/blog/order-cash/ai-powered-b2b-customer-service-inspiring-success-stories/", "https://www.esker.com/company/customer-success-stories/siemens-healthineers-customer-story/" ], "date_added": "2026-10-11", "last_updated": "2026-10-26" }, { "title": "Multilingual operator training", "url": "https://aiagentsformanufacturing.com/use-cases/multilingual-operator-training/", "markdown": "https://aiagentsformanufacturing.com/use-cases/multilingual-operator-training.md", "department": "Office & management", "department_url": "https://aiagentsformanufacturing.com/departments/office/", "evidence": "Verified", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Training & HR", "problem": "Plants run on many languages and three shifts, with high turnover. Classroom training in one language, on paper or slides, does not reach everyone and cannot show who learned what. Outsourced training videos are slow and costly to produce and translate: at Novelis each course took four to six weeks, was delivered only in English and in person, and translating into eight languages was costing close to $1,000,000.", "how_ai_helps": "Training teams write one course and use AI to produce it as a video or an avatar-led lesson, then generate voice-overs and subtitles in each worker's language. Updates are made in-house in hours instead of going back to an agency. Some tools add a quiz and a record per person, so supervisors and auditors can see who completed which procedure, in which language, with what result.", "result": "Novelis (aluminium manufacturer, about 13,500 employees): moved operations training for the Novelis Operating System in-house with Synthesia AI video, in ten languages, through a digital learning academy open to all sites. Localization costs fell about 85%, course production time fell 83% (from 1.5 months to 1 week), and more than 3,000 training videos were produced in one quarter (Synthesia does not say how many are unique lessons versus language versions of the same lesson). Use spread to EHS, finance, procurement and HR (Synthesia case study). Grupo Jorge (Spanish manufacturer with 12 production sites), DEPLOYMENT AND REACH ONLY, no savings reported: NAIMA, an AI avatar trainer built by Imascono, delivers training to more than 5,000 operators at 12 production sites, with 7+ mother tongues on the floor and three shifts, in 40+ languages, online or offline, with a built-in exam and an auditable record per person; the project runs within the European RISE SME programme (Imascono case study). Both are vendor-published stories; only Novelis reports measured time and cost savings.", "companies": [ { "name": "Novelis", "url": "https://aiagentsformanufacturing.com/companies/#novelis" } ], "tools_deployed": [ { "name": "NAIMA (Imascono)", "url": "https://aiagentsformanufacturing.com/tools/#naima-imascono" }, { "name": "Synthesia", "url": "https://aiagentsformanufacturing.com/tools/#synthesia" } ], "tools": [ "Synthesia (AI video with avatars and voice in many languages): used at Novelis", "NAIMA by Imascono (avatar trainer app with exams and per-person records): used at Grupo Jorge", "Other options: HeyGen, Colossyan, Microsoft Copilot or DeepL for text translation, and connected-worker platforms with AI translation such as Augmentir, Dozuki, SwipeGuide", "Your LMS or training matrix to record completions" ], "sources": [ "https://www.synthesia.io/case-studies/novelis", "https://imascono.com/en/success-story/grupo-jorge/" ], "date_added": "2026-10-08", "last_updated": "2026-10-14" }, { "title": "Agent confirms supplier PO acknowledgements in the ERP", "url": "https://aiagentsformanufacturing.com/use-cases/agent-po-confirmation-erp/", "markdown": "https://aiagentsformanufacturing.com/use-cases/agent-po-confirmation-erp.md", "department": "Procurement", "department_url": "https://aiagentsformanufacturing.com/departments/procurement/", "evidence": "Verified", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Purchasing", "problem": "Suppliers confirm purchase orders by email and PDF in every format imaginable; someone has to open each one, compare it field by field with the PO in the ERP and post the confirmation, and many POs never get confirmed at all, so delivery dates are a guess.", "how_ai_helps": "OCR and GPT models extract PO number, quantities, prices and dates from each acknowledgement, business rules check them against the ERP purchase order, clean matches are confirmed automatically and exceptions go to the buyer with a drafted supplier email.", "result": "Sigmoid reports, for an anonymized personal and home care products manufacturer with 90+ plants (~500K POs a year): ~80% of PO confirmations processed autonomously, ~65% shorter order-confirmation cycle time and ~$200K annual operating cost savings (Sigmoid case study, vendor-reported, company anonymized; these are Sigmoid's figures and we have no independent confirmation). Note: Sigmoid's own illustration shows missing-date acknowledgements processed automatically 'with a flag', so the ~80% autonomous figure may include confirmations that this page recommends treating as exceptions.", "companies": [], "tools_deployed": [], "tools": [ "OCR + GPT models for document extraction (used in the Sigmoid case)", "SAP S/4HANA and ECC/R3 via APIs (used)", "Microsoft Teams for exception routing (used)", "Built by Sigmoid (implementation partner)", "Other tools to evaluate: Tacto, Didero, Workist, Zalion" ], "sources": [ "https://www.sigmoid.com/case-studies/agentic-ai-transforms-procurement-operations-with-80-autonomous-purchase-order-processing/" ], "date_added": "2026-10-14", "last_updated": "2026-10-26" }, { "title": "Agent checks supplier invoices against contract terms to stop leakage", "url": "https://aiagentsformanufacturing.com/use-cases/invoice-contract-compliance-agent/", "markdown": "https://aiagentsformanufacturing.com/use-cases/invoice-contract-compliance-agent.md", "department": "Procurement", "department_url": "https://aiagentsformanufacturing.com/departments/procurement/", "evidence": "Analyst-reported", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Purchasing", "problem": "Invoices get paid even when prices, surcharges or delivery performance don't match what was negotiated in the contract, so the value of the deal slowly leaks away.", "how_ai_helps": "The source says one pharmaceutical company 'has deployed AI agents to enforce invoice-to-contract compliance... Its agents track supplier delivery performance and automatically check invoices and POs against contract terms.' It does not describe the tools, data, or how deviations are handled. The steps below are a practical method you can apply; they are not a description of what the company did.", "result": "McKinsey (article 'Redefining procurement performance in the era of agentic AI', 5 Feb 2026) gives this case one sentence: an anonymized pharmaceutical company 'has deployed AI agents to enforce invoice-to-contract compliance', and 'the new approach has cut the value lost through “leakage” by 4 percent.' The article does not say 4 percent of what (spend, invoices checked, or the previous leakage amount), gives no baseline or time frame, and does not say how much is due to the agents, so do not read it as the size of the benefit. The article's other figures (e.g. 10–15% savings from AI-guided negotiations at a telco) concern other companies and other use cases, not invoice checking. This page is mainly a practical method; the evidence for its result is thin.", "companies": [], "tools_deployed": [], "tools": [ "Not named in the source", "Other tools to evaluate: contract-compliance and spend-recovery features of AP automation and source-to-pay suites (e.g. Coupa, SAP Ariba, Medius, Basware), contract AI tools (e.g. Icertis)", "Do-it-yourself: an LLM to extract contract terms and compare them with invoice lines, with amounts recalculated in a spreadsheet" ], "sources": [ "https://www.mckinsey.com/capabilities/operations/our-insights/redefining-procurement-performance-in-the-era-of-agentic-ai" ], "date_added": "2026-10-14", "last_updated": "2026-10-26" }, { "title": "Agent sets order quantities and inventory from the production plan", "url": "https://aiagentsformanufacturing.com/use-cases/production-plan-driven-ordering-agent/", "markdown": "https://aiagentsformanufacturing.com/use-cases/production-plan-driven-ordering-agent.md", "department": "Logistics / Supply Chain", "department_url": "https://aiagentsformanufacturing.com/departments/logistics/", "evidence": "Analyst-reported", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Planning & scheduling", "problem": "Material orders and stock levels are set by hand or by old min/max settings that don't follow the actual production plan, so the plant carries too much inventory of some parts and runs short of others.", "how_ai_helps": "The source says only this: an aircraft OEM 'is using agents to automate order execution and inventory levels based on production planning data.' It does not say what the agents are (language models, optimization software or rules), which ERP/MRP they work with, what the baseline was, or over what period. The steps below are a standard production-plan-driven (MRP) method you can apply yourself; they are not a description of what the OEM did.", "result": "McKinsey (article 'Redefining procurement performance in the era of agentic AI', Feb 2026) reports, in two sentences and with the company anonymized, that the approach 'helped the company cut active inventory by 30 percent, boosting EBIT by around $700 million.' McKinsey gives no baseline, time frame, method or breakdown of how much came from the agents versus classic planning and inventory work, so treat this as an attributed claim, not a measured AI result. The dollar figure reflects a very large manufacturer.", "companies": [], "tools_deployed": [], "tools": [ "Not named in the source", "Other tools to evaluate: the MRP module of your ERP (SAP, Microsoft Dynamics, Epicor, Infor), Netstock, Kinaxis, o9", "Do-it-yourself: a spreadsheet or Python script for net requirements, plus an LLM only to summarise exceptions and draft supplier messages" ], "sources": [ "https://www.mckinsey.com/capabilities/operations/our-insights/redefining-procurement-performance-in-the-era-of-agentic-ai" ], "date_added": "2026-10-14", "last_updated": "2026-10-14" }, { "title": "AI generates CNC toolpaths so programming stops being the bottleneck", "url": "https://aiagentsformanufacturing.com/use-cases/ai-cnc-toolpath-programming-cam-assist/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-cnc-toolpath-programming-cam-assist.md", "department": "Engineering", "department_url": "https://aiagentsformanufacturing.com/departments/engineering/", "evidence": "Verified", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Engineering", "problem": "CNC programming by hand is the bottleneck in a busy job shop, especially when there are many similar parts, and new programmers take a long time to get productive. High-mix, low-volume shops feel it most: Arcamed, a contract maker of surgical cases and trays in Indianapolis, says its parts 'are similar but never the same', new products from customers tend to arrive at the same time, and each wave of new parts used to need a lot more manpower to program. Its quoting of machined parts was also manual and took a time-constrained person's time.", "how_ai_helps": "An AI add-in to the CAM software (CloudNC CAM Assist in Autodesk Fusion) generates the machining strategy and toolpaths for the chosen machine and setup, and a companion feature (Cutting Parameters AI) suggests cutting speeds and feeds; the programmer then reviews and fine-tunes the program. The owner says complex 3+2 toolpaths still need some tweaking.", "result": "Total Manufacturing Solutions, a Florida job shop (since 2016, about 14,000 sq ft, 14 CNC machines: 3 three-axis, the rest 3+2, 4- or 5-axis; batches from one-offs to 200 parts) says that on a job with a few hundred similar part variants, programs that took about 45 minutes to an hour to write by hand took 5 to 10 minutes. The owner says the AI does about 80% of the work and less experienced staff get up to speed faster; integrating tool databases and speeds and feeds slowed them down at first (customer interview on CloudNC's site, a vendor claim; no cost or licence figure). A second shop, Arcamed (contract medical manufacturer, Indianapolis; 5-axis and 3-axis mills cutting plastic components that hold implants and screws during sterilization), started using CAM Assist in August 2024; its CNC programmer says the tool currently completes 75% of the process, creating the operations and selecting and chaining the geometry while people work on other tasks, and that they are starting to use it for more accurate cycle times when quoting (customer interview on CloudNC's site, a vendor claim; no time, cost or licence figures).", "companies": [ { "name": "Total Manufacturing Solutions", "url": "https://aiagentsformanufacturing.com/companies/#total-manufacturing-solutions" } ], "tools_deployed": [ { "name": "CloudNC CAM Assist", "url": "https://aiagentsformanufacturing.com/tools/#cloudnc-cam-assist" } ], "tools": [ "CloudNC CAM Assist: used at Total Manufacturing Solutions (Autodesk Fusion add-in) and at Arcamed", "Cutting Parameters AI (CloudNC): used at Total Manufacturing Solutions", "Other tools to evaluate: CAM Assist for Mastercam, Siemens NX, GibbsCAM and SolidCAM; feature-recognition and template tools in your CAM; Toolpath" ], "sources": [ "https://www.cloudnc.com/case-studies/total-manufacturing-solutions-cutting-programming-time-with-cam-assist", "https://www.cloudnc.com/case-studies/arcamed-applying-cam-assist-to-accelerate-surgical-case-and-tray-production" ], "date_added": "2026-10-14", "last_updated": "2026-10-16" }, { "title": "AI pulls BOM data from supplier documents and checks it against stock and specs", "url": "https://aiagentsformanufacturing.com/use-cases/ai-bom-processing-supplier-documents/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-bom-processing-supplier-documents.md", "department": "Engineering", "department_url": "https://aiagentsformanufacturing.com/departments/engineering/", "evidence": "Verified", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Engineering", "problem": "At RM Components, as Interloom describes it, every production run starts from a bill of materials, and engineers copied component data out of supplier PDFs by hand, cross-checked it against inventory records and design tolerances, and flagged discrepancies for purchasing. A single complex BOM could take half a day, and errors caught late in the cycle caused costly production delays.", "how_ai_helps": "AI extracts the component data from supplier documents, then fixed rules check each component against inventory levels and design specifications, and any mismatch goes to the responsible engineer with the full context.", "result": "RM Components (described by Interloom as a precision engineering firm making complex assemblies for industrial clients; other companies share this name and we could not independently confirm which one this is) reports BOM processing time down 61%, about 90% 'automatic field extraction accuracy' on standard supplier formats (the story does not define how accuracy is measured), and an average of 3.5 hours a week saved per engineer (Interloom customer story, 20 Oct 2025, a vendor claim). The story says a single complex BOM could take half a day before, but does not say whether the 61% applies to that case. Discrepancies found by the deterministic checks (stock, tolerances, specs) go to the responsible engineer, and the team refines extraction rules for new document types; the story does not say what happens to extraction errors that still pass the checks, or how many there are.", "companies": [], "tools_deployed": [ { "name": "Interloom", "url": "https://aiagentsformanufacturing.com/tools/#interloom" } ], "tools": [ "Interloom (AI-assisted extraction plus deterministic validation): used at RM Components", "Other tools to evaluate: document-AI platforms (Rossum, ABBYY, Azure Document Intelligence), BOM tools in PLM/ERP (Arena, Propel, Siemens Teamcenter), Luminovo for electronics BOMs", "Any LLM with the prompt below to test extraction on a few documents" ], "sources": [ "https://interloom.com/en/customer-stories/2025-10-rm-components-bills-of-materials/" ], "date_added": "2026-10-14", "last_updated": "2026-10-26" }, { "title": "Clean up duplicate and incomplete spare-parts records in the material master", "url": "https://aiagentsformanufacturing.com/use-cases/spare-parts-master-data-cleanup/", "markdown": "https://aiagentsformanufacturing.com/use-cases/spare-parts-master-data-cleanup.md", "department": "Maintenance", "department_url": "https://aiagentsformanufacturing.com/departments/maintenance/", "evidence": "Verified", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Maintenance & reliability", "problem": "The spare-parts master data is full of duplicates and incomplete records, so parts are hard to find, stock is higher than needed, the same part gets bought twice, and obsolete parts are only discovered at reorder, which lengthens machine downtime.", "how_ai_helps": "software with matching algorithms compares the plant's spare-parts records with a large reference database of manufacturer part data (SPARETECH says millions of records from thousands of manufacturers) to find duplicates, fill in missing fields and block new duplicates when parts are created. SPARETECH calls these 'intelligent matching algorithms'; its page does not say whether any of this is machine learning or generative AI, so treat this case as master-data matching against a reference database. An LLM or fuzzy matching on your own export can do a first-pass grouping, but it has no reference database behind it.", "result": "ElringKlinger (about 9,500 employees, automotive supplier): in three years with SPARETECH, its maintenance team adjusted over 9,000 records and cleaned up over 2,200 duplicates, and completeness of relevant material master fields went from 72% to 98% (vendor case study, SPARETECH). SPARETECH also headlines '277% faster material creation'; the page does not give the old and new time per material or say how the percentage is calculated, so it cannot be converted into minutes or checked. The page says ElringKlinger avoided unnecessary procurement costs and optimized inventory but gives no figure for either, and it does not separate what the software's matching did from the maintenance team's own work. This case is about cleaning and standardizing the material master; for finding excess and interchangeable stock across plants, see the MRO Verusen case.", "companies": [ { "name": "ElringKlinger", "url": "https://aiagentsformanufacturing.com/companies/#elringklinger" } ], "tools_deployed": [ { "name": "SPARETECH", "url": "https://aiagentsformanufacturing.com/tools/#sparetech" } ], "tools": [ "SPARETECH: used at ElringKlinger", "Other tools to evaluate: Verusen, GenAlpha, Prometheus Group, SAP Master Data Governance, and an LLM or Excel fuzzy matching for a first pass" ], "sources": [ "https://sparetech.io/customers/elringklinger" ], "date_added": "2026-10-14", "last_updated": "2026-10-26" }, { "title": "Stop paying freight charges you don't owe: AI flags surcharges, detention and billing errors before payment", "url": "https://aiagentsformanufacturing.com/use-cases/freight-invoice-audit-surcharges-demurrage/", "markdown": "https://aiagentsformanufacturing.com/use-cases/freight-invoice-audit-surcharges-demurrage.md", "department": "Logistics / Supply Chain", "department_url": "https://aiagentsformanufacturing.com/departments/logistics/", "evidence": "Verified", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Logistics & trade compliance", "problem": "Carrier and forwarder invoices usually go straight to finance, and finance pays what is on the invoice: the amount matches a supplier and a PO, so it looks fine. The logistics technician who booked the truck or container often never sees the invoice, so they never learn that a detention fee, a dry-run fee or a fuel surcharge was added, and can't dispute it, even though they are the only one who knows whether the truck waited, whether the pickup was cancelled in time, or what fuel table was agreed. The most common unjustified charges are: (1) late or delay fees: detention (truck or container kept too long) and demurrage (container left at the port or terminal too long), billed even when free time was not exceeded or the delay was the carrier's; (2) missed-pickup fees: 'dry run' or 'truck ordered not used', billed when the pickup was cancelled in time or the truck never showed up; (3) fuel surcharges that were not announced in advance or do not match the agreed index and table; (4) charges billed to the wrong party, for example export/import document preparation fees that are outside the carrier's scope, or that the customer refused so the carrier rebills the seller, although the sale terms (the Incoterm) put that cost on the buyer (the Incoterm splits costs between buyer and seller; it does not by itself decide what you owe the carrier); plus duplicate invoices and charges not in the contract. Dispute windows are short: in US ocean shipping, the FMC rule (46 CFR Part 541, effective May 2024) gives the billed party at least 30 days from the invoice date to request a waiver or refund of demurrage or detention, and a carrier that invoices more than 30 days after the charge was last incurred cannot require payment (an NVOCC, a carrier that sells ocean shipping without owning ships, has 30 days from the invoice it received instead; 46 CFR 541.7). Once finance has paid, recovering money is much harder.", "how_ai_helps": "Every freight invoice is checked line by line before it is approved for payment: base rate against the contract, each surcharge against the agreed accessorial list, fuel against the contracted fuel table and the published index for the shipment date, and detention, demurrage and dry-run fees against the shipment's own evidence (appointment and gate timestamps, free time, pickup confirmations and cancellations), and every charge against the Incoterm and agreed responsibilities to check who should pay it. A script or spreadsheet does the calculations; the AI reads the invoices and documents, matches them, explains each flag and drafts the dispute. Flagged lines are held and routed to the logistics person who handled the shipment, with the dispute deadline, before finance pays.", "result": "US maker of aluminum extrusions and forgings (anonymous; 12 plants, 9 warehouses, more than $5M a year in freight, 1,200 ocean shipments a year), Pando 'Pi' AI agent: before, manual 4-way invoice matching let overpayments and duplicate charges reach more than $200K a year; the agent now matches invoices against contracts, shipment details and delivery confirmations, flags issues such as incorrect ocean detention charges or rail fuel surcharges, reports 98% exception-free invoice processing, and over $200K of annual overpayments eliminated; detention and demurrage penalties fell 25%, which Pando attributes to proactive shipment monitoring and automated carrier communications (Pando case study). Leading North American appliance manufacturer (anonymous; $40M+ freight, about $4.5M a year in accessorials), ITS Traffic Systems: analysis of recurring accessorials (36,000+ beyond charges, 6,300+ weight verification charges, 2,600+ liftgate charges a year) led to automated audit rules, carrier-vs-shipment weight matching and address validation before dispatch; $4.185M documented process savings over 2024–2026, which ITS calls about 10% of total accessorial spend (ITS case study; the 10% does not fit with $4.5M a year over a few years and ITS gives no base for it, so treat it as an unreconciled figure; this is rules and analytics, not an LLM). Global electrical and power-management manufacturer (anonymous), Intelligent Audit freight audit with AI anomaly detection: $15.4M in annualized savings and a stated 1,906% ROI, of which $10.4M audit and recovery ($9.2M freight, $1.2M parcel), $1.74M from AI anomaly detections, and $3.02M from optimization (service downgrades, consolidations, address corrections) that is not invoice checking; those three items add up to $15.16M, $0.24M less than the $15.4M headline, and the page does not explain the gap, the ROI method or the period (Intelligent Audit case study, undated). All figures are vendor-published.", "companies": [], "tools_deployed": [ { "name": "Intelligent Audit", "url": "https://aiagentsformanufacturing.com/tools/#intelligent-audit" }, { "name": "ITS Traffic Systems", "url": "https://aiagentsformanufacturing.com/tools/#its-traffic-systems" }, { "name": "Pando", "url": "https://aiagentsformanufacturing.com/tools/#pando" } ], "tools": [ "Pando Pi (AI freight audit & pay agent): used at the aluminum extrusion and forging maker", "ITS Traffic Systems (freight audit, accessorial analytics and rules): used at the appliance manufacturer", "Intelligent Audit (freight audit, recovery, AI anomaly detection): used at the electrical manufacturer", "Other tools to evaluate: Cass Information Systems, nVision Global, Trax, Freehand, CTSI-Global, and the freight-settlement module of your TMS", "Do-it-yourself: a spreadsheet or Python script for expected charges, plus an LLM to read invoices, match evidence and draft disputes (prompt below)" ], "sources": [ "https://pando.ai/multi-industry-parts-manufacturer-cuts-200k-usd-in-freight-costs-with-ai-agents", "https://www.itstraffic.com/case-study-4-how-its-generated-4-million", "https://www.intelligentaudit.com/case-studies/large-global-manufacturer-achieves-15-4m-in-annualized-savings-and-an-roi-of-1-906", "https://www.ecfr.gov/current/title-46/chapter-IV/subchapter-B/part-541", "https://www.fmc.gov/articles/fmc-publishes-final-rule-on-detention-and-demurrage-billing-practices/" ], "date_added": "2026-10-14", "last_updated": "2026-10-26" }, { "title": "AI drafts and checks PFMEA / DFMEA from past FMEAs and the process flow (truck chassis welding study, custom seating maker)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-drafts-and-checks-pfmea-dfmea/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-drafts-and-checks-pfmea-dfmea.md", "department": "Quality", "department_url": "https://aiagentsformanufacturing.com/departments/quality/", "evidence": "Pilot", "summary": "AI drafts PFMEAs from past FMEAs: a truck chassis welding study cut FMEA time from 2–5 days to under 1 hour; a seating maker from weeks to 30 min.", "specialized": false, "who_this_is_for": null, "role": "Quality (APQP and launch: process and design FMEA)", "problem": "FMEAs are slow to build and easy to get wrong. Teams copy old spreadsheets, operation numbers drift from the process flow and control plan, severity ratings go missing, and design and process FMEAs are not linked, so risk analysis is incomplete and launches slip. Past FMEAs hold the knowledge, but nobody can search them well: at a custom seating manufacturer, building a P-FMEA and control plan meant collecting failure modes and remedies from stakeholders in Excel over several weeks, without reusing previous analyses; in a truck chassis welding study, the classic process of searching an FMEA database by keyword, collecting data by email and paper, and validating in a team meeting took 2 to 5 days per FMEA.", "how_ai_helps": "An LLM drafts functions, failure modes, effects and causes from a process flow diagram or design specs, suggests Severity / Occurrence / Detection ratings for the team to challenge, checks the PFMEA against the control plan (operation numbers, prevention and detection controls), and flags gaps against the AIAG-VDA 7-step structure (e.g. missing severity, causes placed at the wrong level, actions with no owner).", "result": "Truck chassis welding at an automotive manufacturer, experimental LLM-based PFMEA system (Poznan University of Technology and Politehnica Bucharest, Applied Sciences, December 2025, peer-reviewed): experts from design, production engineering, operations and quality enter structure, function, failure, risk and optimization data in web forms; the system adds similar past FMEAs (8 welding processes on bus chassis and 12 on concrete-mixer chassis) and asks OpenAI's model to draft the PFMEA, restricted to entries in the knowledge base. After four invalid first attempts used to calibrate the input, PFMEAs for the next three welding processes were generated and validated by the quality manager; full-cycle time to generate an FMEA went from 2 to 5 days with the classic method to under one hour. Custom seating manufacturer (anonymous), Retrocausal Kaizen Copilot fed with all its existing P-FMEAs: P-FMEA and control plan preparation from several weeks to 30 minutes, which sped up PPAP (vendor case study). Omnex Systems (anonymous global manufacturer) describes an LLM/ML-based FMEA agent with 200+ pre-trained validation rules and reports qualitative gains only (faster approvals, better classification of failure modes and causes, less rework). A 2025 Design Science study (El Hassani et al.) proposes an LLM framework for FMEA with human validation and compares GPT-4, GPT-4o and Gemini on a case study.", "companies": [], "tools_deployed": [ { "name": "Retrocausal Kaizen Copilot", "url": "https://aiagentsformanufacturing.com/tools/#retrocausal-kaizen-copilot" } ], "tools": [ "General LLM (ChatGPT, Claude, Copilot, Gemini) for drafting, with the prompt below", "Custom web app with OpenAI's API and a database of past FMEAs: used in the truck chassis welding study", "Retrocausal Kaizen Copilot (P-FMEA and control plans from past analyses): used at the custom seating manufacturer", "Omnex Systems FMEA AI Agent (validation rules and generation from core/master FMEAs): vendor case", "Your FMEA software or spreadsheet for the final, approved document (for example APIS IQ, PLATO, Relyence, Siemens Teamcenter)" ], "sources": [ "https://www.mdpi.com/2076-3417/15/24/13199", "https://retrocausal.ai/case-study/custom-seating-manufacturer-reduces-process-analysis-time/", "https://www.omnexsystems.com/case-study/fmea-ai-agent-for-a-global-manufacturer", "https://doi.org/10.1017/dsj.2025.7" ], "date_added": "2026-10-14", "last_updated": "2026-10-22" }, { "title": "AP invoices that match the PO post without anyone touching them (Birmingham Fastener, Bolletje, Duni)", "url": "https://aiagentsformanufacturing.com/use-cases/ap-touchless-invoice-posting-birmingham-fastener/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ap-touchless-invoice-posting-birmingham-fastener.md", "department": "Office & management", "department_url": "https://aiagentsformanufacturing.com/departments/office/", "evidence": "Verified", "summary": "Touchless AP for manufacturers: Bolletje posts 86% of invoices without a person, Duni went from 32% to 70%, Birmingham Fastener 74.3% of PO invoices.", "specialized": false, "who_this_is_for": null, "role": "Finance & AP", "problem": "A growing manufacturer cannot hire accounts payable (AP) clerks as fast as invoice volume grows. Invoices arrive by email in many layouts, and someone must key them in, match them to the purchase order (PO) and the goods receipt, chase approvals and post them, while early-payment discounts expire. Birmingham Fastener tripled in size in nine years; even with 80% of invoices arriving by email, matching them to POs, fixing problems and tracking status stayed manual, communication between departments was disjointed and metrics were hard to track. At Bolletje, a Dutch bakery with two production sites, finance, purchasing and warehouse staff had little visibility into invoice status and did three-way matching by hand. Duni, a Swedish maker of napkins, table covers and take-away packaging, processes about 100,000 invoices a year from a shared service centre and needed more automation.", "how_ai_helps": "Software reads each invoice, whatever the layout, and learns from corrections (machine learning). It matches the invoice to the PO (two-way match) or to the PO and the goods receipt (three-way match), suggests the general ledger (GL) coding, and posts the invoices that meet the rules without a person: that is a 'touchless' invoice. Invoices that do not match, such as a price above the PO or a missing receipt, go to the AP team or the approver with the reason shown and an audit trail. Deadline alerts help pay on time to capture early-payment discounts.", "result": "Birmingham Fastener (T-head bolt maker and fastener distributor, 748 employees, 20 locations in the US and Mexico, Epicor P21 ERP), Esker AP with Enlighten.Net: one year after go-live (the Esker customer story is undated and does not give the start date), 74.3% of PO invoices post automatically; payment discounts captured rose from 85% to 98%; the AP team absorbs growing volume without new hires (Esker customer story, a vendor claim). Bolletje (crackers, gingerbread and cereals, 500 employees, two production sites, Microsoft Dynamics AX), Medius AP Automation with three-way matching: 86% touchless rate on more than 20,000 invoices a year, after a three-month integration (Medius case study). Duni (table-top and take-away packaging, Sweden, about 100,000 invoices a year, Dynamics AX and SAP), Medius with two-way matching of invoice to PO: touchless rate up from 32% to 70% and 97% of invoices routed automatically to the right approver (Medius case study).", "companies": [ { "name": "Birmingham Fastener", "url": "https://aiagentsformanufacturing.com/companies/#birmingham-fastener" }, { "name": "Bolletje", "url": "https://aiagentsformanufacturing.com/companies/#bolletje" }, { "name": "Duni", "url": "https://aiagentsformanufacturing.com/companies/#duni" } ], "tools_deployed": [ { "name": "Esker", "url": "https://aiagentsformanufacturing.com/tools/#esker" }, { "name": "Medius AP Automation", "url": "https://aiagentsformanufacturing.com/tools/#medius-ap-automation" } ], "tools": [ "Esker Accounts Payable with Enlighten.Net ENet Docs, on Epicor P21: used at Birmingham Fastener", "Medius AP Automation: used at Bolletje and Duni", "Other AP tools: Rossum, Basware, Stampli, Tipalti, Yooz, and the AP modules of your ERP", "For a first test: any LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below, on one invoice, its PO and its receipt, with no confidential supplier data in a public chatbot" ], "sources": [ "https://cloud.esker.com/fm/others/034-esker-accounts-payable-customer-story-birmingham-fastener-en.pdf", "https://www.medius.com/resources/case-studies/bolletje/", "https://www.medius.com/resources/case-studies/duni-ap-automation/" ], "date_added": "2026-10-14", "last_updated": "2026-10-27" }, { "title": "APS planning board (INFORM FELIOS) for on-time delivery in a precision job shop", "url": "https://aiagentsformanufacturing.com/use-cases/aps-planning-board-on-time-delivery-capaul/", "markdown": "https://aiagentsformanufacturing.com/use-cases/aps-planning-board-on-time-delivery-capaul.md", "department": "Production", "department_url": "https://aiagentsformanufacturing.com/departments/production/", "evidence": "Verified", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Planning & scheduling", "problem": "In a precision job shop with many small orders, it is hard to see which orders are slipping and which machines are overloaded, so promised lead times are guesses. At CAPAUL, manual coordination of the plan took about eight hours a week, and its planning lead says that warning a customer of a delay four months ahead is far better than a week ahead.", "how_ai_helps": "An advanced planning and scheduling (APS) system, INFORM FELIOS, sits on top of the ERP, plans operations against real machine capacity and shows on a digital planning board which operations are late, which resources are overloaded and where to act; when a machine breaks down or an order changes, the plan is updated and bottlenecks are flagged early. INFORM calls FELIOS 'AI-driven', but the story does not say which functions use AI and which are classic finite-capacity scheduling, so treat this as an APS case.", "result": "CAPAUL (precision parts for aerospace, medical, energy and rail, Eupen, Belgium), using INFORM FELIOS since 2021: weekly planning coordination down from about eight hours to about two; on-time delivery up 'around 25 percent' (the page gives no starting or ending rate and does not say whether this is percentage points or a relative gain); inventory: the page says both that stock levels 'have not increased' despite significant growth and, quoting the CEO, that inventory 'remains constant relative to revenue growth', which are not the same thing (flat in absolute terms versus flat as a ratio to revenue), and it gives no figures (INFORM success story, undated web page, a vendor claim; no baseline year is given).", "companies": [ { "name": "CAPAUL", "url": "https://aiagentsformanufacturing.com/companies/#capaul" } ], "tools_deployed": [ { "name": "INFORM FELIOS", "url": "https://aiagentsformanufacturing.com/tools/#inform-felios" } ], "tools": [ "INFORM FELIOS APS: used at CAPAUL", "Other APS tools: Asprova, Siemens Opcenter APS, PlanetTogether, DELMIA Ortems, Fastems, and ERP planning modules" ], "sources": [ "https://www.inform-software.com/en/success-stories/capaul" ], "date_added": "2026-10-14", "last_updated": "2026-10-26" }, { "title": "AI scheduling for high-mix job shops: group similar parts, pick the best machine, plan machines and people together (J/E Bearing, Yeager, Three Sigma)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-job-shop-scheduling-je-bearing/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-job-shop-scheduling-je-bearing.md", "department": "Production", "department_url": "https://aiagentsformanufacturing.com/departments/production/", "evidence": "Verified", "summary": "AI job-shop scheduling: J/E Bearing on-time delivery up 37%; Yeager from under 80% to over 90% OTD in 2 months; Three Sigma 15% fewer production delays.", "specialized": false, "who_this_is_for": null, "role": "Planning & scheduling", "problem": "In a low-volume, high-mix job shop every order differs. The ERP produces a schedule that meets due dates and material timelines but ignores efficiency: it does not group similar parts to cut setups or send each part to the machine best suited to it. The experienced schedulers know how to do this, but that know-how lives in their heads. Yeager Manufacturing, a high-mix precision manufacturer in Winter Park, Florida, already ran ProShop as its ERP but could not turn that data into a plan that survived the shop floor: a machine can be free while the qualified operator is not, material moves, people call in sick, hot orders arrive, and the resource that looks overloaded is not always the one making jobs late; its on-time delivery was stuck under 80%. Three Sigma Manufacturing, a precision machining company in Kent, Washington serving aerospace, defense and industrial customers, scheduled 47 machine work centers with Excel and the knowledge of senior staff, with a planning horizon of one week.", "how_ai_helps": "The schedulers' rules (which parts to group, which machine suits which part) were captured from the people who held them and turned into a scheduling algorithm, built as a prototype with Rotman MBA students, that groups similar parts to cut setups, assigns jobs to machines and sequences them. A 'geometry layer' adds part dimensions and patterns as an input to match each part to the machine best suited to make it. The source does not say which software or optimiser is used, nor whether the schedule is used live or reviewed in simulation. Commercial AI scheduling tools follow the same idea at a larger scale: Skody, used at Yeager and Three Sigma, reads the jobs, routings and current state from the ERP (ProShop in both cases), computes a finite schedule that takes machines, operators' skills and availability, and material readiness into account, and recomputes it on demand. Planners use it for what-if questions (add a qualified operator, bring material in earlier, someone is absent, take a new order) and check whether customer delivery dates actually move before acting.", "result": "J/E Bearing & Machine Ltd. (precision machining job shop, Tillsonburg, Ontario), built with Rotman MBA students in the Toronto Region Board of Trade's Growth Development Program: the company reports a 37% improvement in on-time delivery after implementing the AI-driven scheduling system (Toronto Region Board of Trade article, 27 May 2026). No baseline, period or setup-hour figure is given, and the article is published by the programme that sells the course, so this is a pilot-level result. Yeager Manufacturing (high-mix precision manufacturer, Florida), Skody AI scheduler on top of ProShop ERP: on-time delivery from under 80% to over 90% in two months ('We were stuck under 80% OTD,' Zachary Yeager, President); after an absence, a new plan is computed and validated in minutes (Skody case studies, a vendor claim). Three Sigma Manufacturing (aerospace and defense machining, Washington), Skody on top of ProShop: 15% reduction in production delays, scheduling time from days to seconds, planning horizon from one week to one year, 47 machine work centers scheduled, and the planner accepts the computed baseline schedule about 99% of days instead of rebuilding it (Skody case study, a vendor claim).", "companies": [ { "name": "J/E Bearing & Machine", "url": "https://aiagentsformanufacturing.com/companies/#j-e-bearing-machine" }, { "name": "Three Sigma Manufacturing", "url": "https://aiagentsformanufacturing.com/companies/#three-sigma-manufacturing" }, { "name": "Yeager Manufacturing", "url": "https://aiagentsformanufacturing.com/companies/#yeager-manufacturing" } ], "tools_deployed": [ { "name": "Skody AI", "url": "https://aiagentsformanufacturing.com/tools/#skody-ai" } ], "tools": [ "Custom algorithm built in-house with Rotman School of Management students through the World Trade Centre Toronto Growth Development Program: used at J/E Bearing & Machine", "Skody AI scheduler (finite scheduling of machines, qualified operators and material, on top of ProShop ERP): used at Yeager Manufacturing and Three Sigma Manufacturing", "Other tools to evaluate: ProShop, Fulcrum, PlanetTogether, Siemens Opcenter APS", "Python with an open-source optimiser (Google OR-Tools) for a do-it-yourself pilot" ], "sources": [ "https://bot.com/News/From-Gut-Feel-to-Algorithm-How-One-Manufacturer-Used-AI-to-Rethink-How-Work-Gets-Done", "https://skody.ai/case-study", "https://skody.ai/case-studies/yeager-manufacturing", "https://skody.ai/case-studies/three-sigma-manufacturing" ], "date_added": "2026-10-14", "last_updated": "2026-10-29" }, { "title": "Automatic feasibility check of customer drawings at RFQ intake", "url": "https://aiagentsformanufacturing.com/use-cases/drawing-feasibility-check-at-rfq-werk24/", "markdown": "https://aiagentsformanufacturing.com/use-cases/drawing-feasibility-check-at-rfq-werk24.md", "department": "Quality", "department_url": "https://aiagentsformanufacturing.com/departments/quality/", "evidence": "Verified", "summary": "AI reads customer drawings at RFQ intake and flags out-of-capability tolerances: 35% fewer NCRs from misread specs at a plastics supplier. Steps and a prompt.", "specialized": false, "who_this_is_for": null, "role": "Quality (feasibility check of customer drawings at RFQ intake)", "problem": "Customer drawings for injection-moulded and machined parts arrive with mixed units, different tolerance and PMI conventions from each customer, and finish notes buried in PDFs. Reading them by hand caps how many RFQs the team can take in, and under time pressure tight tolerances, special finishes or threads are sometimes missed. They then surface after the order is won, as nonconformance reports (NCRs), scrap or a margin lost on a part that was harder to make than quoted.", "how_ai_helps": "AI reads each 2D drawing (components and assemblies, including acceptable scans) and extracts the title-block data and the features that drive feasibility: dimensions with upper and lower tolerances, GD&T, threads and surface roughness, plus outer dimensions and process hints. Each field comes with a confidence score. Policy checks then normalize units and compare every feature with the company's capability rules (tolerance bands per process, finishes it can and cannot do), giving a Go / No-Go result at intake. High-confidence fields are accepted automatically; low-confidence fields and every No-Go go to a quality engineer before the quote goes out.", "result": "Global plastics (injection-moulding) supplier (anonymous), with Werk24 feasibility gates at RFQ intake: 35% fewer NCRs caused by misread specifications (no starting number, period or method is given); 95% extraction accuracy (an average: it does not say how many critical requirements are missed), up to about 99% with custom quality gates on critical features; typically 5 to 20 seconds per page instead of manual reading. Werk24 describes the rollout as API key, sample drawings and policy mapping in days 1 to 2, intake gates live with tuned confidence thresholds in week 1, then extension to inspection planning and QMS reporting from week 2, with no labelled training data needed (Werk24 case study). The same extraction is used for quoting at Laserhub (see the AI quoting case). The figures are published by the vendor.", "companies": [], "tools_deployed": [ { "name": "Werk24", "url": "https://aiagentsformanufacturing.com/tools/#werk24" } ], "tools": [ "Werk24 drawing-reading API (structured output with per-field confidence, REST API, Python SDK and webhooks into ERP or QMS): used at the plastics supplier", "Other tools to evaluate: CADDi (drawing data platform), Paperless Parts (quoting with drawing analysis)", "Any multimodal LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below, on a few past drawings, for a manual first test; check every extracted tolerance" ], "sources": [ "https://werk24.io/case-studies/plastics-company-feasibility-analysis" ], "date_added": "2026-10-14", "last_updated": "2026-10-26" }, { "title": "Operator assistant over SOPs, troubleshooting guides and process data", "url": "https://aiagentsformanufacturing.com/use-cases/operator-assistant-sops-process-data-c3/", "markdown": "https://aiagentsformanufacturing.com/use-cases/operator-assistant-sops-process-data-c3.md", "department": "Production", "department_url": "https://aiagentsformanufacturing.com/departments/production/", "evidence": "Verified", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Production", "problem": "New operators struggle to find the root cause when equipment misbehaves. They must ask the few experienced technicians or read long troubleshooting guides, then compare historian readings with target values by hand. It is slow and error-prone, and unplanned downtime grows.", "how_ai_helps": "A generative AI assistant combines SOPs, troubleshooting guides, textbooks and job aids with live process data from the historian. The operator asks a question in plain language and gets probable causes, the parameters that are out of range and the corrective steps from the procedures, with sources.", "result": "Large manufacturer (anonymous; C3's page labels it 'F500 manufacturing company, 80+ manufacturing facilities' without spelling out F500), C3 Generative AI deployed at one major plant, 20 weeks to a production-ready solution: 5 data sources unified (3 document types and 2 process data types), 1,000+ documents and 5 billion+ rows of structured process data, with a C3-finetuned model tested against a library of operator queries; '90% reduction in time to accurate and relevant insights' (C3 does not define how this was measured or what the baseline was) and 85%+ accuracy for answers retrieved from documents (C3 AI case study, a vendor claim).", "companies": [], "tools_deployed": [ { "name": "C3 Generative AI", "url": "https://aiagentsformanufacturing.com/tools/#c3-generative-ai" } ], "tools": [ "C3 Generative AI: used at the plant", "Others: Siemens Industrial Copilot for operations, AVEVA / Cognite assistants, Microsoft Copilot Studio over documents plus historian data" ], "sources": [ "https://c3.ai/customers/manufacturing-operator-assistance-with-c3-generative-ai" ], "date_added": "2026-10-14", "last_updated": "2026-10-26" }, { "title": "Condition-monitoring alerts ranked by an expert so teams act on the right ones (Boise Cascade, Arbec, Kerry Group)", "url": "https://aiagentsformanufacturing.com/use-cases/prioritized-condition-monitoring-alerts-assetwatch/", "markdown": "https://aiagentsformanufacturing.com/use-cases/prioritized-condition-monitoring-alerts-assetwatch.md", "department": "Maintenance", "department_url": "https://aiagentsformanufacturing.com/departments/maintenance/", "evidence": "Verified", "summary": "Expert-ranked condition-monitoring alerts: Boise Cascade scaled to 17 sites with 130 documented saves; Arbec prevented 15 failures in six months.", "specialized": false, "who_this_is_for": null, "role": "Maintenance (reliability: condition monitoring and alert triage)", "problem": "Monitoring systems that send every anomaly as an alert create alert fatigue: maintenance teams cannot tell which warnings threaten production, so real problems get ignored along with the noise. Multi-site companies add another problem: each plant has its own practices and level of experience, much of the work is reactive, and there are few people trained in condition-based maintenance. At Boise Cascade, maintenance practices varied significantly from site to site and much of the organization relied on reactive maintenance and aging assets. On a line where one machine failing takes the whole line down, as on Arbec's oriented strand board (OSB) lines, the cost of a missed warning is the line.", "how_ai_helps": "Wireless sensors on pumps, motors, gearboxes, fans and compressors send vibration and temperature data continuously. Machine learning screens the readings and flags possible faults; AssetWatch's condition-monitoring engineers (people) then review those findings, rank them and send only prioritised, prescriptive recommendations: what is wrong, how urgent, what to do. The AI does the screening; the ranking is human expert triage. Recommendations flow into the CMMS as work, with enough notice to source parts and repair in a planned stop.", "result": "Boise Cascade (engineered wood products, US): after two sites piloted it independently, the program scaled from 1 to 17 manufacturing sites and from 96 to 1,224 monitored points in 13 months (November 2023 to December 2024); at the 13-month mark, 362 assets monitored, 130 documented saves, an estimated $4M in maintenance, repair and lost-production costs avoided, and a 5x return on investment. Arbec (oriented strand board maker; the page is titled for the Miramichi, NB plant, while its text names plants in Miramichi, NB, Saint-Georges, QC and Amos and does not say whether the results below cover all of them): started in early 2024 with 200 sensors per plant on critical machines, installed in 2 days; first value in about 4 weeks, 15 failures prevented in the first six months, a 12x six-month ROI and a planned increase to 330 sensors. Kerry Group (food and beverage site): about 20 critical assets and 100 sensors on dryers, blowers and washdown-area equipment; three major issues caught in the first two months, about $43,000 in estimated avoided costs. Engineered-wood and plywood producer (anonymous, 15 facilities): six-month proof of concept with 11 faults fixed before failure and $113,000 documented savings, then deployed at all 15 facilities with 130 resolved downtime risks, $2M+ in savings and an ongoing 5x ROI. All are AssetWatch case studies (vendor-published); savings are estimates and combine AI screening with expert review.", "companies": [ { "name": "Arbec", "url": "https://aiagentsformanufacturing.com/companies/#arbec" }, { "name": "Boise Cascade", "url": "https://aiagentsformanufacturing.com/companies/#boise-cascade" }, { "name": "Kerry Group", "url": "https://aiagentsformanufacturing.com/companies/#kerry-group" } ], "tools_deployed": [ { "name": "AssetWatch", "url": "https://aiagentsformanufacturing.com/tools/#assetwatch" } ], "tools": [ "AssetWatch (wireless sensors, AI fault detection, dedicated condition-monitoring engineers, CMMS integration): used at Boise Cascade, Arbec, Kerry Group and the anonymous wood producer", "Others: Augury, Tractian, Waites, SKF, Siemens Senseye", "Any LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below, on a week of alerts and your asset criticality list, to rank them for the maintenance meeting" ], "sources": [ "https://www.assetwatch.com/casestudies/boise-cascade", "https://www.assetwatch.com/casestudies/arbec-miramichi", "https://www.assetwatch.com/casestudies/kerry-group", "https://www.assetwatch.com/casestudies/wood-products" ], "date_added": "2026-10-14", "last_updated": "2026-10-26" }, { "title": "AI finds duplicate and excess MRO spare parts across plants", "url": "https://aiagentsformanufacturing.com/use-cases/mro-spare-parts-inventory-optimization-verusen/", "markdown": "https://aiagentsformanufacturing.com/use-cases/mro-spare-parts-inventory-optimization-verusen.md", "department": "Maintenance", "department_url": "https://aiagentsformanufacturing.com/departments/maintenance/", "evidence": "Verified", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Maintenance & reliability", "problem": "Spare-parts (MRO) stock is decided plant by plant, with non-standard descriptions, several ERP instances and no cross-site view. The same part sits under different numbers in different plants, some sites overstock while others risk stockouts, and money stays tied up in parts nobody uses.", "how_ai_helps": "AI harmonises MRO data from all plants into one view, detects probable duplicate materials despite different descriptions, compares stock with actual usage and risk, and recommends updated stocking policies and transfers between sites. Planners approve the changes.", "result": "Industrial manufacturer (anonymous, 29 plants; the page text says only 'industrial manufacturer', and 'Fortune 500 industrial equipment manufacturer' appears only in its URL), Verusen: $10.5M of inventory savings verified, out of $20.9M identified; 3,000+ potential duplicate materials found; about 2,000 materials found at risk of stockout (Verusen gives no period for the $20.9M and $10.5M; a separate figure of $55K per hour in verified reductions is measured over the first six months, and 'verified' is the vendor's word); 800+ stocking policies updated (Verusen case study, a vendor claim). Duplicates are 'potential' until a planner confirms fit, form and function.", "companies": [], "tools_deployed": [ { "name": "Verusen", "url": "https://aiagentsformanufacturing.com/tools/#verusen" } ], "tools": [ "Verusen: used at the equipment manufacturer", "Others: GenAlpha, Prometheus Group, SAP MRO analytics; Excel or an LLM to group descriptions for a first pass" ], "sources": [ "https://verusen.com/use_case/use-case-fortune-500-industrial-equipment-manufacturer/" ], "date_added": "2026-10-14", "last_updated": "2026-10-26" }, { "title": "Connected-worker platform with AI to speed up operator onboarding", "url": "https://aiagentsformanufacturing.com/use-cases/connected-worker-ai-onboarding-augmentir/", "markdown": "https://aiagentsformanufacturing.com/use-cases/connected-worker-ai-onboarding-augmentir.md", "department": "Office & management", "department_url": "https://aiagentsformanufacturing.com/departments/office/", "evidence": "Verified", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Training & HR", "problem": "At the battery maker, average tenure among operations staff was falling and experience levels were at an all-time low (its own description). New operators take too long to become proficient, training material is hard to create and maintain, qualifications are tracked by hand, and information operators need is in table-top books or other systems instead of at the line.", "how_ai_helps": "A connected-worker platform puts training, digital work instructions, one-point lessons and troubleshooting at the point of work, and tracks skills and qualifications in real time. Augmentir says the platform is built on its 'Smart AI' foundation, and the case-study page lists two AI functions: AI-based workforce intelligence, which identifies workforce improvement opportunities, and Augie, a generative AI assistant that helps workers with troubleshooting, proactive insights and decision-making. The page does not say which of these the battery maker used, or how much of the result comes from AI versus moving from paper to digital.", "result": "Global battery and power-storage manufacturer (anonymous), Augmentir: onboarding time down 40% (no onboarding duration before or after is given), worker productivity up over 17%, time spent looking up instructions in table-top books down 35%, and time spent going to other systems for data down 60% (Augmentir case study, undated, a vendor claim). These results cover the whole connected-worker rollout (digitised procedures, training in the flow of work); the source does not attribute them to AI specifically.", "companies": [], "tools_deployed": [ { "name": "Augmentir", "url": "https://aiagentsformanufacturing.com/tools/#augmentir" } ], "tools": [ "Augmentir: used at the battery maker", "Others: Dozuki, SwipeGuide, Tulip, Poka, Azumuta" ], "sources": [ "https://www.augmentir.ai/case-studies/battery-manufacturer" ], "date_added": "2026-10-14", "last_updated": "2026-10-26" }, { "title": "AI fabric inspection on knitting machines stops the machine at the first recurring defect (Yeşim, Ekoten, A. Sampaio)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-fabric-inspection-knitting-yesim/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-fabric-inspection-knitting-yesim.md", "department": "Quality", "department_url": "https://aiagentsformanufacturing.com/departments/quality/", "evidence": "Verified", "summary": "AI on knitting machines stops at the first recurring defect: Yeşim cut defects nearly 70%; Ekoten avoided 6,008 kg of defective fabric.", "specialized": true, "who_this_is_for": "Textile mills that knit fabric, mainly on circular knitting machines (most of the results come from them; one Portuguese pilot used two open-width machines), and that can mount cameras on the machines.", "role": "Quality (in-process inspection: circular knitting)", "problem": "In circular knitting (machines that knit tube-shaped fabric), fabric is usually inspected only after a roll comes off the machine, sometimes 100% by operators on an inspection table. By then a needle, yarn, elastane (stretch fiber, also sold as spandex or Lycra) or oil defect may have run through a whole roll. The defective fabric still goes on to be dyed and finished, so yarn, water, energy and dyehouse capacity are spent on fabric that will be scrapped or downgraded; recurring defects repeat because nobody sees them in time, and tired operators miss subtle defects or grade inconsistently. In the A. Sampaio pilot, needle defects ran for an average of 27 minutes, elastane and yarn defects 47 minutes and oil marks and dashes 24 minutes before being caught.", "how_ai_helps": "Cameras and lighting mounted inside each knitting machine photograph the fabric as it is knitted, and machine-learning models compare it with clean fabric to detect lines, holes, needle, elastane and yarn defects and oil marks. When a critical defect is detected or recurs, the machine stops so the operator fixes the cause before more fabric is lost (Smartex calls this a 'Golden Stop'). At Yeşim, after a stop a password-protected screen requires an authorized supervisor to investigate and resolve the root cause before production can resume (the source does not say the cause must be recorded). Defect data by machine, article and yarn lot feeds root-cause analysis, and automatic roll grading lets the team inspect only the rolls likely to be poor.", "result": "Yeşim Group (integrated knitting-to-garment maker, Bursa, Turkey): in Lycra jersey production, machines equipped with Smartex show nearly 70% fewer defects, with less waste downstream in dyeing, finishing, cutting and sewing; oil-mark defects were traced to the oil and fixed by changing it (Just Style Excellence Awards profile, 2025). Ekoten (Sun Tekstil Group, Izmir, Turkey), which co-developed the system with Smartex from 2020: in the pilot on eleven machines the error rate fell by about 50%; since November 2021, 376 hours of defective production and 6,008 kg of defective fabric avoided, worth €30,040, with a payback of four to six months (Ekoten's own claim; the Apparel Impact Institute profile gives a general payback of 9–18 months and a project length of 6–18 months for the vendor); 28 knitting operators trained for 14 days (ITMF award submission, 2022). Later figures for Ekoten, with 17 equipped machines (no measurement date is given, so they cannot be added to the figures above): 80% fewer kilograms of defective fabric, on average 20 high-impact continuous defects stopped per machine per month, and over 4.5 tons of raw fabric not produced each year (Apparel Impact Institute solution profile). A. Sampaio (Portugal, EU DigiTVC pilot on two open-width machines, April to October 2022): 72 Golden Stops, about 34 hours of machine production and 744 kg of raw material saved, about €6,696 at €9/kg, and less manual roll inspection (DigiTVC project page). Impetus (Portugal), four-week pilot of automatic roll grading: rolls needing inspection down from 33% to 16% (Smartex blog). Figures are published by the companies, an EU project and the vendor.", "companies": [ { "name": "A. Sampaio", "url": "https://aiagentsformanufacturing.com/companies/#a-sampaio" }, { "name": "Ekoten", "url": "https://aiagentsformanufacturing.com/companies/#ekoten" }, { "name": "Yeşim Group", "url": "https://aiagentsformanufacturing.com/companies/#yesim-group" } ], "tools_deployed": [ { "name": "Smartex", "url": "https://aiagentsformanufacturing.com/tools/#smartex" } ], "tools": [ "Smartex CORE (in-machine cameras and AI defect detection with machine stop), FACT and LOOP (quality data and roll grading): used at Yeşim, Ekoten, A. Sampaio and Impetus", "Other fabric-inspection systems to evaluate: automatic inspection on the knitting machine from the machine maker, and roll inspection systems from Cognex, Keyence, EVS or Shelton", "Any LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below, on your defect records, to find patterns by machine, yarn lot and shift" ], "sources": [ "https://www.just-style.com/featured-company/yesim-group-ai-inspection-technology-innovation/", "https://www.itmf.org/images/dl/awards/2022/Ekoten-Tekstil-and-Smartex-ai.pdf", "https://apparelimpact.org/resources/solutions/smartex-ai-enabled-real-time-quality-control/", "https://digitvc.eu/portuguese-pilot/", "https://www.smartex.ai/post/boosting-final-inspection-with-automatic-production-grading" ], "date_added": "2026-10-14", "last_updated": "2026-10-26" }, { "title": "Ask an AI agent about live production instead of reading yesterday's ERP reports", "url": "https://aiagentsformanufacturing.com/use-cases/ai-agent-live-production-data-sauder/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-agent-live-production-data-sauder.md", "department": "Production", "department_url": "https://aiagentsformanufacturing.com/departments/production/", "evidence": "Verified", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Production", "problem": "Plants often manage from yesterday's reports. At Sauder Woodworking, supervisors had operational data through the ERP but often saw it about 24 hours after the fact, so meetings reviewed what had happened instead of fixing it during the shift.", "how_ai_helps": "A connected-workforce platform collects real-time production data from the lines. Leaders and team leads ask an AI agent in plain language what is going on or what their biggest issues are, and purpose-built agents (line lead, continuous improvement) surface what needs attention during the shift, which also feeds daily and tier meetings.", "result": "What the AI agent does: supervisors and team leads ask Redzone's ChampionAI agents in plain language what is happening on their line and get the biggest issues surfaced during the shift from real-time production data, instead of waiting about 24 hours for ERP reports. Programme results (Redzone connected-workforce programme as a whole, not the agent alone): Sauder Woodworking (ready-to-assemble furniture, largest North American supplier to IKEA) reports average changeover time down 93% (from 19 minutes to 84 seconds) and productivity up 40% in the first year; OEE in the pilot area from about 50% to over 70%; three times as many improvement ideas as on paper. A Nucleus Research ROI study reported 665% ROI with a seven-month payback. The source does not isolate the agent's share (ERP Today, Aug 2026, based on Redzone's announcement and case study).", "companies": [ { "name": "Sauder Woodworking", "url": "https://aiagentsformanufacturing.com/companies/#sauder-woodworking" } ], "tools_deployed": [ { "name": "Redzone (QAD)", "url": "https://aiagentsformanufacturing.com/tools/#redzone-qad" } ], "tools": [ "Redzone Connected Workforce with ChampionAI (QAD): used at Sauder", "Others: Tulip, Fabriq, Evocon, MachineMetrics, Sight Machine; Microsoft Copilot or ChatGPT over a production data export for a manual start" ], "sources": [ "https://erp.today/redzone-agentic-ai-sauder-woodworking/" ], "date_added": "2026-10-14", "last_updated": "2026-10-14" }, { "title": "An AI agent that turns your emails and messages into a live, prioritised to-do list and day plan", "url": "https://aiagentsformanufacturing.com/use-cases/ai-agent-live-todo-list-from-email-and-chat/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-agent-live-todo-list-from-email-and-chat.md", "department": "Office & management", "department_url": "https://aiagentsformanufacturing.com/departments/office/", "evidence": "Method", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Office & management", "problem": "In a plant, the work you are asked to do rarely arrives as a task. It is buried in a supplier email ('can you confirm by Thursday?'), a Teams message from production, a comment in Asana or ClickUp, or the fourth reply of a long thread. Nobody re-reads all of it, so requests slip, the urgent item is found at 4 p.m., and the day is planned around whatever landed last. Microsoft's own field data shows the size of the problem: in one large study the average worker spent 11.7 hours a week reading and answering email.", "how_ai_helps": "An agent with read-only access to your mailbox, chat and work tools wakes up every 15 minutes, reads only what is new since its last run, and pulls out every request, commitment and deadline hidden in the messages. It merges duplicates (the same ask in an email and in Teams), scores each task P1/P2/P3 with clear rules (deadline, sender, impact, blocking others), keeps one live to-do list, and proposes a day plan that it re-shuffles when a new P1 arrives. You confirm, edit or dismiss; it never replies or acts for you.", "result": "Method case: no published figure measures this exact agent end to end. Related published figures: (1) Microsoft researchers, randomised study of about 6,000 workers at 56 firms (\"Early Impacts of M365 Copilot\", April 2025): \"Workers with access to Copilot reduced this time by 7% on average (12 minutes per week)\" of email reading, and \"Treated workers who used Copilot at least once a week reduced their weekly time reading emails by half an hour (18%).\" (2) Microsoft researchers, 7,137 workers (\"Shifting Work Patterns with Generative AI\", 2025): \"workers with Copilot reduced their time using Outlook for email by 1.4 hours per week, a 12% reduction\". (3) ClickUp customer stories (vendor-reported): VanHomes (modular home manufacturer), where \"AI takes customer emails, breaks them into subtasks, and categorizes each issue automatically\" and \"it saves our team at least 10–15% of their time\"; Bell Direct (online broker, not a manufacturer), where an agent triages \"800+ emails\" a day, with \"2 FTEs' work handled by Super Agents\" and a \"20% boost in operational efficiency\". Practitioner note (Frédéric D'Amours): I run this kind of agent for my own work. Every 15 minutes it reads what's new in my mail and messages, adds the tasks it finds to my list with a P1, P2 or P3, and updates my plan for the day. I don't have a measured before/after figure; the gain I notice is that requests no longer get lost in threads and I start each day with a plan rather than an inbox.", "companies": [], "tools_deployed": [ { "name": "ClickUp Brain", "url": "https://aiagentsformanufacturing.com/tools/#clickup-brain" } ], "tools": [ "Microsoft 365 Copilot (Outlook, Teams) + Microsoft Planner / To Do; Copilot Studio to build a scheduled agent", "Gmail + Google Tasks / Google Calendar (Gemini in Workspace, or Apps Script on a time trigger)", "Zapier, Make or n8n: scheduled 15-minute polling, connectors to Outlook, Gmail, Teams, Slack, Asana, ClickUp, and an LLM step", "ClickUp Brain / ClickUp Super Agents (email-to-task, used at VanHomes and Bell Direct)", "Asana AI (task extraction and prioritisation inside Asana)", "Motion and Reclaim.ai: turn the prioritised list into time blocks on your calendar", "Any general LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below to test the logic on a pasted batch first" ], "sources": [ "https://arxiv.org/abs/2504.11443", "https://arxiv.org/abs/2504.11436", "https://clickup.com/customers/vanhomes", "https://clickup.com/customers/bell-direct" ], "date_added": "2026-10-14", "last_updated": "2026-10-14" }, { "title": "AI warranty claim triage: check each claim against policy, flag duplicates and fraud", "url": "https://aiagentsformanufacturing.com/use-cases/ai-warranty-claim-triage-policy-fraud-check/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-warranty-claim-triage-policy-fraud-check.md", "department": "Quality", "department_url": "https://aiagentsformanufacturing.com/departments/quality/", "evidence": "Verified", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Quality", "problem": "Warranty claims arrive from dealers, distributors and end customers with photos, invoices, serial numbers and free-text fault descriptions. Someone has to check each one against the warranty policy (coverage period, proof of purchase, exclusions), spot duplicates and repeat abusers, and decide approve / reject / ask for more. It takes minutes per claim, decisions vary between reviewers, and fraudulent or duplicate claims slip through when the queue is long.", "how_ai_helps": "AI reads each incoming claim (form, attachments, photos, history of the same serial number, customer or dealer), checks it against the written warranty rules, looks for duplicates and abuse patterns across all past claims, and proposes a decision with the reasons and the evidence it used. Simple, clean claims can be pre-approved by rules; anything with a flag goes to a person. A human keeps the final decision, credits and any reply to the customer.", "result": "Apex Tool Group (described on the page as a global tools manufacturer): AI fraud detection was added on top of its existing warranty-claim approval app built on AgilePoint, and reached production in under three months. Reported results: 76% less time and labour per warranty claim, 80% less manual processing time across the queue, 71% of incoming claims AI-assisted, duplicate and abuser identification 10 times faster, and CSR throughput up 400% (AgilePoint page, undated, a vendor claim; the page gives no baseline minutes per claim and no claim volume). Kawasaki Engines USA: after centralising more than 98,000 claims on Tavant's platform, more than 83% of claims are auto-approved, cycle times fell from weeks to hours as volume grew, headcount stayed flat, and customer satisfaction rose from 30% to 83% (Tavant article, 17 Sep 2025, citing Kawasaki's Senior Manager of IT Business Strategy in a webinar; a vendor claim). Note that Tavant says Kawasaki's auto-approvals come from rule-based engines, not from AI.", "companies": [ { "name": "Apex Tool Group", "url": "https://aiagentsformanufacturing.com/companies/#apex-tool-group" }, { "name": "Kawasaki Engines USA", "url": "https://aiagentsformanufacturing.com/companies/#kawasaki-engines-usa" } ], "tools_deployed": [ { "name": "AgilePoint", "url": "https://aiagentsformanufacturing.com/tools/#agilepoint" }, { "name": "Tavant Warranty", "url": "https://aiagentsformanufacturing.com/tools/#tavant-warranty" } ], "tools": [ "AgilePoint (workflow platform + AI Control Tower): used at Apex Tool Group", "Tavant Warranty: used at Kawasaki Engines USA", "Other tools to evaluate: Syncron Warranty, PTC Servigistics / Arbortext warranty modules, SAP warranty claim processing, Salesforce Service Cloud; document-AI (Azure Document Intelligence, ABBYY) for invoices and proof of purchase", "Any LLM (ChatGPT, Claude, Copilot) with the prompt below to test the checks on a sample of past claims" ], "sources": [ "https://www.agilepoint.com/warranty-claim-fraud-detection", "https://tavant.com/aibytes/stop-reacting-start-predicting-with-ai-powered-early-warnings-in-the-aftermarket/" ], "date_added": "2026-10-15", "last_updated": "2026-10-15" }, { "title": "AI supplier contract review and renewal tracking", "url": "https://aiagentsformanufacturing.com/use-cases/ai-supplier-contract-review-renewal-tracking/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-supplier-contract-review-renewal-tracking.md", "department": "Procurement", "department_url": "https://aiagentsformanufacturing.com/departments/procurement/", "evidence": "Verified", "summary": null, "specialized": false, "who_this_is_for": null, "role": "Purchasing", "problem": "Supplier agreements, NDAs, quality agreements and price agreements pile up waiting for review. Each one is read clause by clause against the company's standard positions (liability, payment terms, warranty, termination, price adjustment, Incoterms), which takes hours of legal or procurement time and delays sourcing. Once signed, contracts sit in shared drives as PDFs, so notice periods, auto-renewals and price-review dates are missed.", "how_ai_helps": "AI screens each incoming contract against your playbook (your standard and fallback positions), flags only the clauses that deviate and suggests your standard wording, so the reviewer works on the exceptions. For signed contracts, it extracts the key terms (parties, dates, renewal and notice periods, price-adjustment clauses, liability caps, payment terms, Incoterms) into a register and alerts before deadlines. Legal or procurement still approves every change and signs.", "result": "Flex (one of the world's largest electronics manufacturers, 140,000+ employees): its Global Procurement and Supply Chain Legal team embedded DocJuris AI screening and playbook-aligned reviews in its workflow; DocJuris says each agreement took an average of eight days to review and redline, mentions back-and-forth email chains and Word redlines, and reports five minutes after (DocJuris case study, undated, a vendor claim; the page does not say how the days and minutes are measured or how much of the eight days was working time, so do not read it as a labour saving). Trench Group (the Luminance page does not describe the company): Luminance does the first-pass review and suggests standard alternatives; average contract review time fell from 150 minutes to approximately 30 minutes, Sales and Operations are empowered to handle 80% of contracts, and a tariff review project took less than 2 hours versus 2.5 weeks without Luminance (Luminance customer story, undated, a vendor claim; the page gives no measurement method and does not say that Legal is out of the loop).", "companies": [ { "name": "Flex", "url": "https://aiagentsformanufacturing.com/companies/#flex" } ], "tools_deployed": [ { "name": "DocJuris", "url": "https://aiagentsformanufacturing.com/tools/#docjuris" }, { "name": "Luminance", "url": "https://aiagentsformanufacturing.com/tools/#luminance" } ], "tools": [ "DocJuris: used at Flex", "Luminance: used at Trench Group", "Other tools to evaluate: Ironclad, Icertis, Sirion, Agiloft, Spellbook, SAP Ariba Contracts, Coupa Contract Lifecycle Management; Microsoft Copilot or Google Gemini over a contract folder for a first test", "Any LLM (ChatGPT, Claude, Copilot) with the prompt below for a first pass on a few contracts" ], "sources": [ "https://www.docjuris.com/case-studies/how-flex-reduced-contract-review-from-8-days-to-5-minutes", "https://www.luminance.com/customers/how-luminances-legal-grade-ai-is-transforming-trench-groups-contract-activity/" ], "date_added": "2026-10-15", "last_updated": "2026-10-26" }, { "title": "AI freight planning: spot-rate quotes in minutes, consolidated loads and carrier scorecards", "url": "https://aiagentsformanufacturing.com/use-cases/ai-tms-spot-freight-rfq-load-consolidation-pando/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-tms-spot-freight-rfq-load-consolidation-pando.md", "department": "Logistics / Supply Chain", "department_url": "https://aiagentsformanufacturing.com/departments/logistics/", "evidence": "Verified", "summary": "AI freight planning at a Fortune 500 engine maker: spot RFQs in 20 minutes, 94% on-time delivery, 12% freight savings. Steps, cautions and a prompt to try.", "specialized": false, "who_this_is_for": null, "role": "Logistics & transportation", "problem": "Outbound freight is still run from email and spreadsheets in many plants. Every spot shipment means emailing a few carriers, waiting hours for quotes and picking one with no benchmark. Orders going to nearby customers ship on separate partial trucks because nobody sees them together in time, carrier on-time performance is argued about instead of measured, and proof of delivery comes back on paper days later. At a Fortune 500 engine manufacturer described by its TMS vendor, the ERP handled transactions but not freight optimization, spot freight procurement was entirely manual and took hours per request, and proof of delivery was paper-based across more than 28,000 consignees in 190 countries.", "how_ai_helps": "An AI-assisted transportation management system (TMS) sits next to the ERP. It sends spot RFQs to a wider carrier network and benchmarks the quotes against past and market rates, groups open orders by lane and date to suggest consolidated full-truck loads, tracks shipments and electronic proof of delivery in one place, and scores carriers on their actual on-time performance. A planner still approves the carrier award and the load plan; the AI does the quote chasing, the grouping and the measuring.", "result": "Anonymous Fortune 500 engine manufacturer (products used in construction, mining and commercial fleets; more than 28,000 consignees in 190 countries; road and air freight), using Pando's AI-powered TMS integrated with its existing ERP: spot RFQ turnaround reduced to 20 minutes (from hours per request), 94% on-time delivery achieved, 12% freight cost savings, 18% improvement in consolidated shipments, delivery consolidation giving 5.5% cost savings (the page does not say how these overlap, so do not add them), 96% e-proof-of-delivery adoption, and carrier scorecards with automated SLA monitoring. Source: Pando's own customer story (undated page, read October 2026). The figures are vendor-reported, the company is not named, and the page gives no baseline for on-time delivery or freight cost, so treat them as an indication of what to measure rather than a promise.", "companies": [], "tools_deployed": [ { "name": "Pando", "url": "https://aiagentsformanufacturing.com/tools/#pando" } ], "tools": [ "Pando (AI-powered TMS): used at the Fortune 500 engine manufacturer", "Other TMS and freight platforms to evaluate: Oracle Transportation Management, SAP Transportation Management, Manhattan TMS, Blue Yonder TMS, e2open, Uber Freight TMS; for smaller shippers, freight marketplaces with instant quotes", "Any LLM (ChatGPT, Claude, Copilot) with the prompt below, on an export of your open orders and past shipments, to find consolidation candidates and draft a spot RFQ before buying anything" ], "sources": [ "https://pando.ai/fortune-500-engine-manufacturer-achieves-94-otif-with-ai-powered-tms" ], "date_added": "2026-10-15", "last_updated": "2026-10-26" }, { "title": "AI predicts simulation and test results from your past data, so engineers run fewer prototypes", "url": "https://aiagentsformanufacturing.com/use-cases/ai-surrogate-models-fewer-simulations-physical-tests/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-surrogate-models-fewer-simulations-physical-tests.md", "department": "Engineering", "department_url": "https://aiagentsformanufacturing.com/departments/engineering/", "evidence": "Pilot", "summary": "AI trained on past simulation and test data: Sumitomo Riko predicts in under 5 minutes, Nissan cut physical tests 17%. Steps, cautions and a prompt to try.", "specialized": true, "who_this_is_for": "Engineering teams that already run many similar simulations or physical tests on the same family of parts, and have kept the results (dozens to hundreds of comparable runs). Without that history, there is nothing to train a model on.", "role": "Engineering & R&D", "problem": "Every design change triggers another round of finite-element (computing how a part behaves under load by splitting it into many small pieces) or multiphysics (several physical effects at once, such as heat and stress) simulation, and often another physical prototype test. Each run needs setup by a specialist and can take hours, so engineers explore only a few variants and test labs become the bottleneck. Meanwhile years of past simulation results and test reports sit in folders and are rarely reused. At Sumitomo Riko, a rubber automotive components maker, engineers must run hundreds of multiphysics simulations to understand how parts behave under extreme loads, and pre-processing such as defining geometric parameters takes extensive time and expertise.", "how_ai_helps": "Machine-learning models, called surrogate models (fast stand-ins for the slow simulation), are trained on the simulation results and test data you already have. Once trained, they predict the outcome for a new design in minutes instead of hours, recommend which physical tests are still needed, and flag results that look abnormal. Engineers use the fast predictions to screen many variants, then confirm the shortlisted designs with full simulation and the physical tests that the AI recommends keeping.", "result": "Sumitomo Riko (global maker of rubber automotive components such as vibration isolators and hoses), using Ansys SimAI trained on previously generated simulation data: AI models make performance predictions in under five minutes, over one hour of time saved for each new design, with accuracy comparable to high-fidelity simulation; initial testing showed simulation cycles over 10x faster for the mechanical performance of certain rubber bushes. Uses include anti-vibration design, battery cooling, magnetic field analysis and mixing heat transfer (Ansys press release, 8 Oct 2025). Nissan Technical Centre Europe (Cranfield, UK), using Monolith with more than 90 years of vehicle testing data: on bolt joints in vehicle chassis, the AI recommended the optimal torque range and prioritised which additional tests to run, giving an overall 17% reduction in physical testing compared with the non-AI process; Nissan says applying the approach across its European range could cut testing time by half, which is a projection, not a result (The Manufacturer, 3 Nov 2025). Both results are early or first-application results reported by the vendor or the company.", "companies": [ { "name": "Nissan", "url": "https://aiagentsformanufacturing.com/companies/#nissan" }, { "name": "Sumitomo Riko", "url": "https://aiagentsformanufacturing.com/companies/#sumitomo-riko" } ], "tools_deployed": [ { "name": "Ansys SimAI", "url": "https://aiagentsformanufacturing.com/tools/#ansys-simai" }, { "name": "Monolith", "url": "https://aiagentsformanufacturing.com/tools/#monolith" } ], "tools": [ "Ansys SimAI: used at Sumitomo Riko", "Monolith (Next Test Recommender, Anomaly Detector): used at Nissan Technical Centre Europe", "Other tools to evaluate: Siemens Simcenter (Reduced Order Modeling, HEEDS), Altair physicsAI / romAI, PhysicsX, Neural Concept, and the AI features in your current CAE suite", "Any LLM (ChatGPT, Claude, Copilot) with the prompt below to inventory your past simulation and test data and plan a pilot" ], "sources": [ "https://www.automotiveworld.com/news-releases/sumitomo-riko-leverages-ansys-ai-technology-to-accelerate-simulation-over-10x-for-automotive-component-design-and-manufacturing/", "https://www.themanufacturer.com/articles/nissan-harnesses-power-of-ai-to-speed-up-physical-vehicle-tests/" ], "date_added": "2026-10-15", "last_updated": "2026-10-26" }, { "title": "AI rewrites the production schedule in minutes when a line slows or an order changes", "url": "https://aiagentsformanufacturing.com/use-cases/ai-dynamic-production-rescheduling-changeovers-bottler/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-dynamic-production-rescheduling-changeovers-bottler.md", "department": "Production", "department_url": "https://aiagentsformanufacturing.com/departments/production/", "evidence": "Verified", "summary": "AI rescheduling at a beverage bottler: 75% less non-value-added time, changeover downtime down nearly 80%. Steps, cautions and a prompt to try.", "specialized": false, "who_this_is_for": null, "role": "Production planning & scheduling", "problem": "The schedule is wrong by mid-morning: a line slows, a machine goes down, a rush order arrives or a material is late, and the scheduler pulls plant managers, engineering and operations into a room to rebuild the plan. The sequence matters too, because the wrong product order adds cleaning and changeover time. In Microsoft's customer story, a major beverage bottler was replanning production schedules 10 to 15 times per week through manual meetings, using averages and operator experience, while the experienced engineers who held the scheduling know-how were approaching retirement.", "how_ai_helps": "Live plant data (line speeds per SKU, actual changeover and ramp-up times, machine status) is combined with orders, current schedules and the constraints written in plain language. A small language model turns that description into a mathematical optimization model, and a solver produces the best sequence. When a line slows or goes down, the change triggers a new optimized schedule in minutes, and the team can ask what-if questions in plain English before a maintenance stop or demand shift. Schedulers review and release the plan; they no longer rebuild it by hand.", "result": "Major beverage bottler (not named), using Sight Machine with Microsoft's OptiMind model in Microsoft Foundry: non-value-added production time cut by 75%, production capacity up by more than 5%, changeover-related downtime down by nearly 80%, ramp-up delays down by nearly 60%, clean-in-place (CIP) sanitation downtime down by almost 90%, and hours of manual planning work removed every week, without expanding production infrastructure. Microsoft and Sight Machine also describe the general approach as raising overall plant productivity by 10% or more and cutting scheduling time by 75% (Microsoft customer story, read October 2026). The figures are published by Microsoft and its partner Sight Machine about an unnamed customer.", "companies": [], "tools_deployed": [ { "name": "Sight Machine", "url": "https://aiagentsformanufacturing.com/tools/#sight-machine" } ], "tools": [ "Sight Machine (industrial data platform) with Microsoft OptiMind in Microsoft Foundry: used at the beverage bottler", "Other scheduling tools to evaluate: advanced planning and scheduling (APS) systems such as Siemens Opcenter APS, DELMIA Ortems, PlanetTogether, Asprova, and solvers such as Google OR-Tools or Gurobi for in-house models", "Any LLM (ChatGPT, Claude, Copilot) with the prompt below to write down your scheduling rules and find the changeover data you are missing" ], "sources": [ "https://www.microsoft.com/en/customers/story/26648-sight-machine-microsoft-foundry" ], "date_added": "2026-10-15", "last_updated": "2026-10-15" }, { "title": "AI cash application and deduction management: read complex remittances, track customer deductions (Eagle Family Foods, Clorox, Shearer's Foods)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-cash-application-remittances-deductions-eagle-foods/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-cash-application-remittances-deductions-eagle-foods.md", "department": "Office & management", "department_url": "https://aiagentsformanufacturing.com/departments/office/", "evidence": "Verified", "summary": "AI cash application at Eagle Family Foods, Clorox and Shearer's Foods: 80% of payments auto-matched, cash posting 48 h to under 24 h.", "specialized": false, "who_this_is_for": null, "role": "Finance (accounts receivable)", "problem": "Customer payments arrive with remittance advices in every format: PDFs, emails, portal downloads, spreadsheets with thousands of lines. Someone reworks them in Excel to match invoices before posting, so cash posts late and open receivables look worse than they are. Short payments and deductions (promotions, damages, pricing, penalties) are approved by email with backup scattered on personal drives, so nobody knows which are valid and money that should be recovered is written off. At Eagle Family Foods Group, a food maker with 406 employees, it took up to 2 days to post one day of cash, the largest customer's remittances had 2,000 to 3,000 lines that took more than 8 hours to post, and the team worked holidays to keep up. At Shearer's Foods, a snack food contract manufacturer with 14 plants in North America, matching imprecise payments to open invoices relied on the knowledge of long-time employees; more customers paid through payment consolidators with little payer information, the share of ACH payments carrying remittance data was falling fast, and customers rarely said why they took a deduction. At The Clorox Company, a consumer products maker, cash posting was inaccurate and delayed, remittances came in through emails and web portals, and a high volume of deductions was coded by hand.", "how_ai_helps": "AI reads each remittance, whatever its layout, extracts the invoice numbers, amounts and deduction codes, and proposes the match to open invoices in the ERP for the analyst to confirm. Deductions get a case each: the AI pre-fills the claim lines, the backup documents and approvals are stored with the claim, and the team sees the status and which deductions are invalid and should be disputed or repaid.", "result": "Eagle Family Foods Group (Cleveland, Ohio; sweetened condensed milk, evaporated milk, popcorn, Hamburger Helper; 406 employees across 4 locations; ERP Infor M3), using Esker Cash Application and Deductions Management with Esker Synergy AI: 50% reduction in cash application posting time, from 48 hours to less than 24 hours; prior day's cash now posted by 2:00 p.m.; remittance processing for the largest customer down from more than 8 hours to 5 to 10 minutes (the highlights say 10 minutes, a 98% reduction); 80% reduction in open deductions; roughly $2.3 million reduction in non-trade deductions; four times as many deductions processed while adding one analyst; cash application handled with a move from one to two analysts despite strong sales growth after a 2022 acquisition. Source: Esker customer story (PDF, 2025). The figures are published by the vendor with the company's cash application lead quoted. The Clorox Company (consumer packaged goods, North America), HighRadius Cash Application: 80% automated cash posting, 64% reduction in the time taken to apply payments, 224 man-hours saved per month, 33% reduction in manpower, automated deduction coding, 27% reduction in past dues (HighRadius case study, a vendor claim). Shearer's Foods (largest contract manufacturing and private label supplier in the North American snack industry, 14 plants), HighRadius AI with ACH and payment data received through PNC Bank: auto-match rate improved to 80%, with the other 20% handled as exceptions; DSO lowered by three days; cash application head count lowered by 50%; more than 50% process improvement. The AI parses email remittances and downloads payment data from large retailers' websites, and learns customers' regular deductions over time (AFP Treasury case study, July 2023, based on a 2022 AFP conference session presented by Shearer's VP Corporate Accounting and PNC Bank).", "companies": [ { "name": "Clorox", "url": "https://aiagentsformanufacturing.com/companies/#clorox" }, { "name": "Eagle Family Foods", "url": "https://aiagentsformanufacturing.com/companies/#eagle-family-foods" }, { "name": "Shearer's Foods", "url": "https://aiagentsformanufacturing.com/companies/#shearer-s-foods" } ], "tools_deployed": [ { "name": "Esker", "url": "https://aiagentsformanufacturing.com/tools/#esker" }, { "name": "HighRadius", "url": "https://aiagentsformanufacturing.com/tools/#highradius" } ], "tools": [ "Esker Cash Application, Deductions Management and Synergy AI: used at Eagle Family Foods Group", "HighRadius Cash Application (AI remittance capture, invoice matching and deduction coding): used at The Clorox Company and Shearer's Foods", "Other tools to evaluate: Billtrust, Versapay, SAP Cash Application, Oracle and Microsoft Dynamics AR automation, Quadient AR (formerly YayPay)", "Any LLM (ChatGPT, Claude, Copilot) with the prompt below on one anonymized remittance to see what matching can be automated" ], "sources": [ "https://cloud.esker.com/fm/others/033-esker-accounts-payable-accounts-receivable-customer-story-eagle-foods-en.pdf", "https://www.highradius.com/resources/case-studies/the-clorox-company/", "https://www.financialprofessionals.org/training-resources/resources/articles/Details/overcoming-the-loss-of-receivables-transaction-data" ], "date_added": "2026-10-15", "last_updated": "2026-10-28" }, { "title": "AI-built load planning tool fills trailers better and cuts outbound freight 12% at a trailer maker", "url": "https://aiagentsformanufacturing.com/use-cases/ai-load-planning-trailer-utilization-outbound-freight/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-load-planning-trailer-utilization-outbound-freight.md", "department": "Logistics / Supply Chain", "department_url": "https://aiagentsformanufacturing.com/departments/logistics/", "evidence": "Verified", "summary": "AI-built load planning at a trailer maker: 12% lower freight spend, trailer utilization 65% to 80%, load lock window 21 to 3 days. Steps, cautions and a prompt.", "specialized": false, "who_this_is_for": null, "role": "Logistics & transportation (outbound load planning)", "problem": "In many plants, outbound loads are still planned by hand. The planner knows which products stack together, what fits on which trailer and which orders can ride together, but that knowledge lives in their head and in a handful of Excel workbooks that do not talk to each other. Trailers leave partly empty, loads have to be locked weeks ahead so the planner can keep up, and any late change in orders is hard to absorb. At a mid-market trailer manufacturer owned by a private-equity firm, outbound freight costs were squeezing margins, load planning was entirely manual and relied on the planner's experience and institutional knowledge, and trailer utilization was sub-optimal.", "how_ai_helps": "The consultants found the data and logic needed to optimize loads already existed but were referenced by hand. Instead of replacing the planning platform, they used AI together with the company's own planners to quickly build a purpose-built load planning tool that encodes the experts' stacking logic and replaces the fragmented Excel workbooks with one connected tool. The source does not say which AI was used or whether the tool itself uses machine learning or rules; the planner still owns the final load plan.", "result": "Anonymous private-equity-owned mid-market trailer manufacturer, project by the consulting firm SSA & Company: rapid discovery of the current load planning process, a single-site pilot in week 6, then the tool deployed across all facilities within 8 weeks. Reported results: freight spend reduced by 12% across all plants; load lock window shortened from 21 to 3 days, giving planners back 18 days of flexibility; trailer utilization up from 65% to 80%; and the company delayed a TMS decision, avoiding a platform overhaul and a multi-year implementation. Source: SSA & Company case study (undated page, read October 2026). The figures are published by the consultancy about an unnamed client; no cost of the project is given.", "companies": [], "tools_deployed": [], "tools": [ "Purpose-built load planning tool developed by SSA & Company with AI and the client's planners: used at the trailer manufacturer (tool not named)", "Other tools to evaluate: load-building and cube optimization software (e.g. EasyCargo, Cargo-Planner, PackVol, Magaya Cargo Planner), the load planning module of your TMS or WMS", "Any LLM (ChatGPT, Claude, Copilot) with the prompt below, to write your planners' stacking rules down and specify a spreadsheet or script before buying anything" ], "sources": [ "https://ssaandco.com/case-studies/delivered-12-freight-savings-in-8-weeks-using-ai-at-pe-owned-trailer-manufacturer/" ], "date_added": "2026-10-16", "last_updated": "2026-10-16" }, { "title": "AI orchestrates a plant-attached warehouse: more loads shipped straight from the plant, fewer shuttle moves (P&G Lima)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-plant-warehouse-orchestration-pg-lima/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-plant-warehouse-orchestration-pg-lima.md", "department": "Logistics / Supply Chain", "department_url": "https://aiagentsformanufacturing.com/departments/logistics/", "evidence": "Verified", "summary": "AI warehouse orchestration at P&G Lima: direct plant shipments 57% to 83.4%, bring-backs 16% to 8.7%, planning 8 h to 20 min a day. Steps and a prompt.", "specialized": false, "who_this_is_for": null, "role": "Warehouse & shipping (plant-attached warehouse)", "problem": "A warehouse attached to a plant has to feed raw materials to the lines and ship finished goods at the same time, with limited space, a production schedule that keeps changing and a crew whose size varies from day to day. Supervisors rebuild the plan by hand every day; when the plan is imperfect, finished goods are moved to outside warehouses and later brought back to complete shipments, and shuttle trucks run between buildings all day. At P&G's Lima, Ohio plant, with seven nearby satellite warehouses and more than 250 outbound full-vehicle shipments a day (85% drop-and-hook, 15% live loads), the vendor lists raw-material firefighting to support production, missed shipments, labour coordination, staging and storage limits, travel across the site and siloed data between production, warehouse and logistics systems.", "how_ai_helps": "An AI-driven warehouse orchestration platform sits on top of the existing WMS and ERP and reads the production schedule. It plans labour, inventory moves and dock activity in real time against the constraints (dock doors, labour, equipment), sequences raw-material moves to the lines, decides which loads can ship directly from the plant, and re-plans when production changes. Supervisors work from the plan (the case mentions a 'Load Now' button) instead of building it by hand.", "result": "Procter & Gamble, Lima, Ohio manufacturing plant with seven satellite warehouses, using AutoScheduler.AI on top of its WMS and ERP: direct dock shipments from the plant up from 57% to 83.4% (+46.3%); bring-backs from satellite warehouses down from 16% to 8.7% (-45.6%); workforce planning time cut from 8 hours to 20 minutes per day (96%); 30% higher throughput during a product crunch; 99%+ on-time shipment performance; shuttle moves to outside warehouses cut by nearly 50%; more than $4 million in total savings. Source: AutoScheduler's own case study (PDF, undated). The figures are vendor-published, and the PDF gives no time frame for the savings, no system cost and no breakdown of the $4 million. It also says trucks shipped directly from the plant 'doubled' (a count), while 57% to 83.4% is a share of shipments (+26.4 points, which is the +46.3% the PDF quotes in relative terms); without total volumes and periods the two figures cannot be reconciled, but they are different measures, not necessarily a contradiction.", "companies": [ { "name": "Procter & Gamble", "url": "https://aiagentsformanufacturing.com/companies/#procter-gamble" } ], "tools_deployed": [ { "name": "AutoScheduler.AI", "url": "https://aiagentsformanufacturing.com/tools/#autoscheduler-ai" } ], "tools": [ "AutoScheduler.AI (warehouse orchestration over WMS and ERP): used at P&G Lima", "Other tools to evaluate: the labour management and wave planning modules of your WMS (Manhattan, Blue Yonder, SAP EWM, Körber), dock scheduling software (Opendock, C3 Reservations), yard management systems", "Any LLM (ChatGPT, Claude, Copilot) with the prompt below on tomorrow's schedule and open loads, to draft a shift plan and see where the conflicts are" ], "sources": [ "https://autoscheduler.ai/wp-content/uploads/2025/03/AutoScheduler-PG-Case-Study.pdf" ], "date_added": "2026-10-16", "last_updated": "2026-10-26" }, { "title": "Generative AI trained on past designs proposes new product designs that meet cost and manufacturability targets (Eaton)", "url": "https://aiagentsformanufacturing.com/use-cases/generative-ai-product-design-cost-dfm-eaton/", "markdown": "https://aiagentsformanufacturing.com/use-cases/generative-ai-product-design-cost-dfm-eaton.md", "department": "Engineering", "department_url": "https://aiagentsformanufacturing.com/departments/engineering/", "evidence": "Verified", "summary": "Eaton's generative AI, built on past designs and aPriori cost and DFM data, cut lighting fixture design time 87% (16 weeks to 2). Steps, cautions and a prompt.", "specialized": true, "who_this_is_for": "Engineering teams that design customer-specific, configurable products and have a library of past designs with simulation, cost and manufacturability data.", "role": "Engineering / R&D (product design)", "problem": "Customer-specific products take months to design because each request needs input from several disciplines and many manual iterations before the design meets cost, performance and manufacturing requirements, and customers wait. At Eaton (power management; 92,000 employees), customers regularly ask for custom components such as a unique passenger-car valve stem or a new lighting fixture; a manual design that is cost-effective and meets manufacturing specifications takes months, and a lighting fixture can take up to 16 weeks because it needs thermal, electrical, mechanical, optical and manufacturing engineering.", "how_ai_helps": "Eaton trained its generative AI (AI that creates new designs rather than only analyzing existing ones) on actual historical product design data and simulation results, including cost and design-for-manufacturing (DFM: checking that a design can be built at the target cost with the plant's processes) analysis from aPriori that runs programmatically in the cloud. The AI runs thousands of design iterations in minutes and proposes the top five designs; those go through a high-fidelity (detailed but slower) simulation, and Eaton's digital design and engineering team reviews them against the requirements before any design is released.", "result": "Eaton (Dublin, Ireland; $23.2 billion revenue and 92,000 employees as given in the ASSEMBLY article, undated), generative AI built on historical design data and aPriori (aP Pro) cost and DFM simulation: design time for an automated lighting fixture reduced by 87%, from up to 16 weeks to two weeks per new design request; design time for a high-speed gear reduced by 65%; weight of a liquid-to-air heat exchanger reduced by 80% (no starting weight, performance constraints or validation status are given). Eaton plans to double the output of its new product innovation investments. Sources: aPriori customer case study and an ASSEMBLY magazine article with the same figures, quoting Uyiosa Abusomwan, Eaton's senior global technology manager of digital design and engineering. The figures are published by the software vendor and the company, not independently audited, and are described as early results.", "companies": [ { "name": "Eaton", "url": "https://aiagentsformanufacturing.com/companies/#eaton" } ], "tools_deployed": [ { "name": "aPriori", "url": "https://aiagentsformanufacturing.com/tools/#apriori" } ], "tools": [ "aPriori aP Pro (cost, DFM and sourcing simulation run programmatically): used at Eaton, with Eaton's own generative AI and simulation tools", "Other tools to evaluate: generative design in Autodesk Fusion, PTC Creo or Siemens NX; nTop; Ansys and Simcenter simulation with AI; should-cost tools; CAD configurators for product families", "Any LLM (ChatGPT, Claude, Copilot) with the prompt below, to frame a design-automation pilot from your past variants before buying software" ], "sources": [ "https://www.apriori.com/resources/case-study/eatons-generative-ai-cuts-product-design-time-by-87/", "https://www.assemblymag.com/articles/99091-generative-ai-slashes-design-time" ], "date_added": "2026-10-16", "last_updated": "2026-10-26" }, { "title": "Technicians identify a spare part from a photo or a few words and see its stock and location (Deutsche Bahn, ENGEL)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-spare-part-identification-photo-search-partium/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-spare-part-identification-photo-search-partium.md", "department": "Maintenance", "department_url": "https://aiagentsformanufacturing.com/departments/maintenance/", "evidence": "Verified", "summary": "AI spare-part search by photo or text: Deutsche Bahn reports 46,800 person-days saved in a year across 95 maintenance plants. Steps, cautions and a prompt.", "specialized": false, "who_this_is_for": null, "role": "Maintenance (spare-parts identification)", "problem": "A technician standing at a broken machine often holds a worn part with no readable label and has to find its part number in catalogues, parts lists or a messy material master, or ask the one colleague who knows. On older equipment the documentation is incomplete or missing. The search takes time, the wrong part gets ordered, and OEMs field the same identification calls on their hotline. manroland Goss says its printing systems can have up to 40,000 parts in up to 500 assemblies, with 5,000 to 10,000 consumable parts depending on the press type; at DB Cargo, fleets span decades and documentation is often incomplete.", "how_ai_helps": "The technician photographs the removed part, types a few words or reads the label, QR or barcode, and an AI search (image recognition plus semantic text search over the material master, parts lists and drawings) returns the matching part with its SAP or part number, manufacturer, storage location and stock. When there is no confident match, the request goes to an expert with the photos and what the technician already tried.", "result": "Deutsche Bahn, with Partium: started at DB Fernverkehr in 2019; five business units (Fernverkehr, Cargo, Regio, vehicle maintenance and the DB Netz machine pool) now use one app built with Partium, DB Mat.ID, which shows the SAP number, manufacturer, storage location and stock within seconds; 95 maintenance plants, 12,200 technicians and users, and 46,800 person-days saved in one year (Partium customer page). ENGEL (injection molding machines; 7,000 employees, 750 technicians) put Partium Find in its parts portal for customers; Partium reports fewer part-search hotline calls, more self-served transactions and more requests for quotes, without figures. manroland Goss (printing presses) has offered its customers Partium image, semantic text and parts-list search since February 2023, with a call-an-expert function that takes up to five photos (press release, June 2023; no results given). Deutsche Bahn is a rail operator, not a manufacturer; it is included because its vehicle maintenance plants run the same workflow as plant maintenance. The figures are vendor-published.", "companies": [ { "name": "Deutsche Bahn", "url": "https://aiagentsformanufacturing.com/companies/#deutsche-bahn" }, { "name": "ENGEL", "url": "https://aiagentsformanufacturing.com/companies/#engel" } ], "tools_deployed": [ { "name": "Partium", "url": "https://aiagentsformanufacturing.com/tools/#partium" } ], "tools": [ "Partium Find (image and semantic multi-modal part search): used at Deutsche Bahn (DB Mat.ID), ENGEL and manroland Goss", "Other tools to evaluate: visual part search in your CMMS or EAM if it has one, OEM parts portals with image search, shape-search tools for 3D parts, MRO master-data tools (Sparetech, Verusen)", "Any multimodal LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below, on a photo and an extract of your spare-parts list, to test identification on a few real parts" ], "sources": [ "https://www.partium.io/customers/deutsche-bahn", "https://www.partium.io/customers/engel", "https://www.manrolandgoss.com/en/read-press-release/items/manroland-goss-makes-part-identification-easier-via-partium-the-web-app-for-spare-parts.html" ], "date_added": "2026-10-16", "last_updated": "2026-10-16" }, { "title": "Machine learning recommends the minimum alloy addition for each heat at a steel mill (Gerdau, Fero Labs)", "url": "https://aiagentsformanufacturing.com/use-cases/ml-minimum-alloy-addition-per-heat-gerdau-fero-labs/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ml-minimum-alloy-addition-per-heat-gerdau-fero-labs.md", "department": "Production", "department_url": "https://aiagentsformanufacturing.com/departments/production/", "evidence": "Verified", "summary": "Machine learning recommends the minimum alloy per heat at Gerdau's Petersburg mill: $3/ton lower alloy cost, 15% less quality variation. Steps and a prompt.", "specialized": true, "who_this_is_for": "Steel mills that add alloys heat by heat and have years of recorded process and lab data (Gerdau's model learned from more than 20,000 heats). The method may interest other batch process plants, but the case itself comes from steelmaking.", "role": "Production (process optimization: recipe and additive amounts per batch)", "problem": "In batch processes such as steelmaking, every batch starts from different raw materials, so plants set conservative recipe limits to make sure every batch meets spec. Quality stays high, but the plant overspends on additives. At Gerdau's Petersburg mill (about 200 products and 25 grades, over 4,000 heats a year (a heat is one batch of molten steel), hundreds of process readings), scrap quality, alloy recovery (the share of an added alloy that ends up in the steel), slag (the waste layer floating on the molten steel) and rolling mill temperatures all vary, and the alloy addition range for each grade was set at a conservative level. The capability analyses (statistical checks that a process stays within spec) used to tune those ranges were done only once or twice a year, on a small subset of the data, and assume linear relationships and normally distributed data.", "how_ai_helps": "A white-box machine learning model (one whose reasoning can be inspected, unlike a black box), trained on melt shop, rolling mill and test lab data (over 20,000 heats covering the mill's whole grade portfolio), predicts the final yield and tensile strength while the steel is still in the ladle (the vessel that carries the molten steel). After each ladle sample, it recommends the lowest-cost alloy addition that still meets the specification with a set confidence, within about 30 seconds of the spectrometer reading (a chemical analysis of a steel sample). The ladle operator sees the recommendation on a screen in the pulpit (the operator control cabin) and makes the addition. The quality team owns the model constraints and the safety buffers added to the spec limits. The model shows confidence bands and rules itself out on new, rare or hard-to-make grades, where operators keep the standard grade limits.", "result": "Gerdau Petersburg mill (Virginia, USA), with Fero Labs, live since 2020: thousands of heats cast with ML-optimized chemistry; alloy cost down $3 per ton on average while meeting quality specs; alloy additions down about 8%; more than 500,000 pounds of manganese, carbon, niobium and vanadium conserved; 15% less quality variation (AISTech 2023 paper co-written by Gerdau and Fero Labs; Fero Labs, March 2023). Moving the operator screen to a central monitor in the ladle furnace pulpit led to a 2.5-fold increase in tonnage cast with the recommendations and a threefold improvement in profits within one month (the paper's wording; what 'profits' covers, whether the alloy savings or something wider, is not defined). The mill kept deliberate buffers on the spec limits, which in some cases cut the per-ton savings potential by up to 55%; the paper puts the true potential alloy reduction at twice the 8% achieved. Keeping the model up to date takes less than 10 minutes a week, with retraining about once a month (upkeep of an established model, not the data collection, validation and governance needed to set it up). Gerdau has expanded Fero to five other plants in North America (Fero Labs case study). The figures are published by the vendor and the mill.", "companies": [ { "name": "Gerdau", "url": "https://aiagentsformanufacturing.com/companies/#gerdau" } ], "tools_deployed": [ { "name": "Fero Labs", "url": "https://aiagentsformanufacturing.com/tools/#fero-labs" } ], "tools": [ "Fero Labs (no-code, white-box machine learning for process data): used at Gerdau", "Other tools to evaluate: process analytics and optimization software that reads your historian and lab data (for example Seeq or Braincube), or your MES or historian vendor's analytics module", "For a first offline test: any LLM with a data-analysis or code tool (ChatGPT, Claude, Copilot, Gemini) on an export of past batches, with the prompt below" ], "sources": [ "https://craft-ferolabs.files.svdcdn.com/production/PDF/AISTech2023_Fero_Gerdau_Adopting_Machine_Learning_Based_Workflows_for_Reducing_Production_Risk_and_Cost_2024-02-14-193044_sgvr.pdf?dm=1707939044", "https://www.ferolabs.com/insights/post/how-gerdau-saved-3-ton-of-steel-with-white-box-machine-learning", "https://www.ferolabs.com/case-studies/gerdau" ], "date_added": "2026-10-19", "last_updated": "2026-10-26" }, { "title": "AI proposes material formulations so R&D runs fewer lab experiments (Resonac, formerly Showa Denko)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-formulation-design-fewer-lab-experiments-resonac/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-formulation-design-fewer-lab-experiments-resonac.md", "department": "Engineering", "department_url": "https://aiagentsformanufacturing.com/departments/engineering/", "evidence": "Verified", "summary": "Materials informatics at Resonac: AI proposes formulations so R&D runs a third of the experiments (3 months to 1 for an adhesive). Steps and a prompt.", "specialized": true, "who_this_is_for": "R&D labs that develop materials by blending ingredients (adhesives, films, coatings, compounds) and have a stack of past experiment records, ideally with a data scientist on staff.", "role": "Engineering / R&D (materials and formulation development)", "problem": "Developing a functional material such as an adhesive, film, coating or compound means trying many combinations of raw materials and ratios until several properties meet target at the same time, often properties that pull against each other (strong adhesion but no residue when peeled off, transparency plus adhesion to copper). Skilled developers rely on knowledge and experience and repeat formulate-and-test cycles. Experiment records are scattered, and the qualitative information (spectra, structure images, notes) gets lost when data is reformatted for analysis.", "how_ai_helps": "Materials informatics (using data and AI to speed up the development of materials): AI models trained on the company's own past experiments predict the properties of a formulation and propose candidate formulations that should meet the required properties (Showa Denko Materials cites Bayesian optimization (a method that picks the next experiment to try based on the results so far) for multi-objective (several properties at once) targets). Developers then test only the most promising candidates. At Resonac, results typed into electronic lab notebooks flow automatically into the database and the models (MLOps, the routine upkeep that keeps models retrained and current), so developers always work with up-to-date predictions, and a data scientist who doubts a prediction can click back to the raw experiment data.", "result": "Showa Denko (now Resonac), adhesive for electronic devices that must adhere strongly yet peel off without residue: using the MI app to focus on promising formulations cut the number of experiments to about one-third of the conventional method and experiment time from three months to one month, less than a year after the data pipeline went live (press release, November 2022). Showa Denko Materials (formerly Hitachi Chemical, now part of Resonac), transparent adhesive film that must be transparent and adhere to copper foil: in AI performance tests, its AI engine reached the target with about 30% fewer experiments than skilled developers, and when the target was raised it reached a value the developers had not yet reached (press release, December 2021). The company says almost all its business divisions use the MI app daily in R&D; about 400 electronic lab notebook IDs were registered as of September 2022. Figures are published by the company.", "companies": [ { "name": "Resonac", "url": "https://aiagentsformanufacturing.com/companies/#resonac" } ], "tools_deployed": [], "tools": [ "Resonac's in-house MI app and MI platform (built with QUATRE-i SCIENCE and Hitachi), with the BIOVIA Notebook electronic lab notebook from Dassault Systèmes", "Other tools to evaluate: materials informatics platforms (for example Citrine Informatics or Uncountable), and design-of-experiments software (JMP, Minitab) as a first step", "Any LLM with a data-analysis or code tool (ChatGPT, Claude, Copilot, Gemini) with the prompt below, on an export of your past formulation trials, to pick the next experiments" ], "sources": [ "https://www.resonac.com/news/2022/11/02/2167.html", "https://www.resonac.com/news/2021/12/07/2071.html", "https://www.resonac.com/jp/news/2022/11/02/2164.html" ], "date_added": "2026-10-19", "last_updated": "2026-10-26" }, { "title": "AI reads customer tenders and specification packages clause by clause and builds the deviation list (ROBEL, Bosch Manufacturing Solutions)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-tender-specification-review-deviation-list-robel-bosch/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-tender-specification-review-deviation-list-robel-bosch.md", "department": "Engineering", "department_url": "https://aiagentsformanufacturing.com/departments/engineering/", "evidence": "Verified", "summary": "AI reviews tenders and RFQ specifications clause by clause: ROBEL saves 95 hours per RFQ; Bosch Manufacturing Solutions reports 53% time savings.", "specialized": false, "who_this_is_for": null, "role": "Engineering (application and sales engineering: requirements review at RFQ)", "problem": "Machine builders and engineer-to-order manufacturers receive tenders and RFQ specification packages that run to hundreds of documents and thousands of pages: technical requirements, safety certifications, referenced standards, environmental criteria, compatibility with the customer's equipment. Engineers list the documents in Excel, split them between departments, collect deviations and open points in a central list, and consolidate it before quoting, often within a few weeks. It costs hundreds of hours per project, not every page gets read in depth, the same questions get answered again and again, foreign-language tenders must be translated first, and a missed requirement surfaces after the order is won.", "how_ai_helps": "The team uploads the whole tender or specification package. The AI extracts each requirement, pre-evaluates it against internal engineering standards and the answers given on past projects, highlights what differs from previous projects with the same customer, flags deviations, risks and open questions clause by clause, and assigns each requirement to the right department. Referenced norms are matched against the actual requirements. Engineers review only their part, in parallel, and decide; the AI then drafts sub-supplier RFQs and first drafts of proposal documents such as technical summaries and test plans.", "result": "ROBEL Bahnbaumaschinen (railway construction and maintenance machines, Freilassing, Germany), with Atira: 95 hours of sales and engineering time saved per RFQ; the tender is annotated clause by clause with red flags on day one; 18 specialist departments review in parallel; a structured norm compliance assessment takes hours instead of weeks of manual lookups (Atira customer page). Bosch Manufacturing Solutions (automated production and assembly systems), with BoWatt: RFQ packages of hundreds of documents and thousands of pages are uploaded as folders, requirements are pre-evaluated against internal standards and past projects, and the team reports 53% time savings on large specification projects; after use in the German subsidiary, other locations have started testing it (BoWatt success story, March 2026). The figures are published by the vendors.", "companies": [ { "name": "Bosch", "url": "https://aiagentsformanufacturing.com/companies/#bosch" }, { "name": "ROBEL", "url": "https://aiagentsformanufacturing.com/companies/#robel" } ], "tools_deployed": [ { "name": "Atira", "url": "https://aiagentsformanufacturing.com/tools/#atira" }, { "name": "BoWatt", "url": "https://aiagentsformanufacturing.com/tools/#bowatt" } ], "tools": [ "Atira (sales engineering agents for tenders): used at ROBEL", "BoWatt (AI requirements management for RFQ specifications): used at Bosch Manufacturing Solutions", "Other tools to evaluate: requirements and proposal tools that shred RFPs into compliance matrices, or a document-reading assistant in your existing PLM or requirements tool", "Any LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below, on one specification document and your standard offer, for a first test" ], "sources": [ "https://atira.ai/customers/robel", "https://www.bowatt.com/resources/customer-success-story-bosch-manufacturing" ], "date_added": "2026-10-19", "last_updated": "2026-10-19" }, { "title": "AI answers customers' 'where is my order?' emails from live ERP data (promotional-products maker, medical device maker)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-order-status-email-replies-from-erp/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-order-status-email-replies-from-erp.md", "department": "Office & management", "department_url": "https://aiagentsformanufacturing.com/departments/office/", "evidence": "Verified", "summary": "AI drafts replies to 'where is my order?' emails from live ERP data: 47% of reply-needed emails handled at one manufacturer, 85%+ resolved at another.", "specialized": false, "who_this_is_for": null, "role": "Office (customer service and inside sales: order-status inquiries)", "problem": "A large share of the customer service inbox at a manufacturer is routine: what is the status of my order, when will it ship, what is your returns policy. Each one means reading the email, looking up the order in the ERP and typing a reply, while complex requests and quotes wait. When volume rises, response times slip, satisfaction drops, and sales staff get pulled in to help answer the queue instead of selling.", "how_ai_helps": "The AI reads each incoming email and first classifies it: needs no reply (acknowledgements, notifications, marketing), routine and in scope (order status, shipping dates, policy questions), or out of scope (custom orders, complaints, escalations), which goes to a person. For routine questions, it looks up the order in the ERP and drafts the reply from that data. A quality check scores each draft; depending on confidence, the AI either replies directly, places a draft in the representative's inbox for review, or hands the email to a person without drafting.", "result": "Promotional-products manufacturer and wholesaler (anonymous, Los Angeles; tens of thousands of customer emails a month, seven-person inside sales team), built by HatchWorks AI on OpenAI, n8n, NetSuite and Outlook: 47% of the emails that needed a reply are now drafted automatically and pass the quality gate; about 50% of incoming emails turned out to need no reply at all; up to 70% of responses to routine inquiries are send-ready (HatchWorks case study). Medical device manufacturer (anonymous), Quiq email AI agent on its Oracle environment, run silently alongside human agents before go-live: more than 85% of the emails the agent answers directly are fully resolved with no follow-up from the customer, response-time SLAs were exceeded, and sales staff pulled into support went back to selling (Quiq customer page). Both are anonymous vendor case studies.", "companies": [], "tools_deployed": [ { "name": "Quiq", "url": "https://aiagentsformanufacturing.com/tools/#quiq" } ], "tools": [ "Custom build on OpenAI, n8n, NetSuite and Microsoft Outlook (HatchWorks AI): used at the promotional-products manufacturer", "Quiq email AI agent integrated with Oracle: used at the medical device manufacturer", "Other tools to evaluate: the AI features of your help desk or CRM, Microsoft Copilot in Outlook, or a workflow tool (n8n, Power Automate) with read-only ERP access", "Any LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below, on a few real emails and an ERP export of the matching orders, for a first test" ], "sources": [ "https://hatchworks.com/case-studies/ai-powered-email-automation-frees-sales-team-to-focus-on-high-value-orders/", "https://quiq.com/customers/medical-device-manufacturer/" ], "date_added": "2026-10-19", "last_updated": "2026-10-19" }, { "title": "A walking robot runs thermal and acoustic inspection rounds and turns what it finds into work orders (AB InBev, BMW, Nestlé Purina)", "url": "https://aiagentsformanufacturing.com/use-cases/robot-thermal-acoustic-inspection-rounds-ab-inbev-bmw/", "markdown": "https://aiagentsformanufacturing.com/use-cases/robot-thermal-acoustic-inspection-rounds-ab-inbev-bmw.md", "department": "Maintenance", "department_url": "https://aiagentsformanufacturing.com/departments/maintenance/", "evidence": "Verified", "summary": "A walking robot runs thermal and acoustic rounds: AB InBev Leuven found nearly 150 anomalies in six months and cut average repair time to 13 days.", "specialized": false, "who_this_is_for": null, "role": "Maintenance (reliability and preventive maintenance: inspection rounds)", "problem": "Inspection rounds are the first preventive task to slip. Walking every line to take thermal readings on motors, gearboxes and electrical panels, and listening for compressed-air or gas leaks, is repetitive work and there are rarely enough people to do it on schedule. Leaks are hard to hear over running machines, so some teams shut lines down to find them or bring in contractors with specialist equipment. A motor running hot or a gearbox low on oil goes unnoticed until it stops a packaging line, and problems found late wait months for a repair. At BMW Hams Hall, manual inspection could take a couple of hours per shift for some of the most experienced specialists.", "how_ai_helps": "A four-legged robot (Boston Dynamics Spot) walks pre-programmed routes on its own while production runs, and stops at fixed inspection points. A thermal camera captures heat on motors, gearboxes and electrical parts; an acoustic imager (Fluke SV600) picks out the sound frequencies of compressed air or gas escaping under pressure, even through background noise; a 360° camera and LIDAR record the area. When the robot finds an anomaly it sends an alert to the maintenance and production managers and enters it in the system, so work orders are generated in advance from its rounds. At BMW the data feeds the plant's digital twin, where employees view each inspection and submit a maintenance request. Technicians spend their time on repairs instead of on the rounds.", "result": "AB InBev brewery in Leuven, Belgium (ten packaging lines, over 50,000 containers an hour): after a test run in 2022, Spot now performs 1,800 individual inspections a week; in its first six months it found nearly 150 anomalies, more than double what the team expected, and average repair time fell from a few months to 13 days. Finds include a faulty transport motor that would have stopped a packaging line for at least 6 hours, gearboxes low on oil, faulty motor fans and a very small ammonia leak. Most anomalies are compressed-air and gas leaks; the team estimates an average air leak at about $550 in wasted product and leaks of ammonia and other expensive gases at up to $15,000. BMW Group Plant Hams Hall (engine plant near Birmingham, England, robot acquired in 2023): six autonomous walks a day, routes set by order of importance; in the first weeks of acoustic inspections it found air leaks in unexpected places, and its readings fill inspection gaps the digital twin had revealed. Nestlé Purina: thermal inspections on non-critical equipment and acoustic compressed-air leak detection in thirty-plus missions on packaging lines, to give time back to technicians. All three are Boston Dynamics case studies (vendor-published).", "companies": [ { "name": "AB InBev", "url": "https://aiagentsformanufacturing.com/companies/#ab-inbev" }, { "name": "BMW", "url": "https://aiagentsformanufacturing.com/companies/#bmw" }, { "name": "Nestlé", "url": "https://aiagentsformanufacturing.com/companies/#nestle" } ], "tools_deployed": [ { "name": "Boston Dynamics Spot", "url": "https://aiagentsformanufacturing.com/tools/#boston-dynamics-spot" } ], "tools": [ "Boston Dynamics Spot with a thermal camera and a Fluke SV600 acoustic imager: used at AB InBev Leuven, BMW Hams Hall and Nestlé Purina", "Other options to evaluate: fixed thermal cameras and continuous acoustic or ultrasound sensors on critical assets, a handheld acoustic imager on a manual route, wheeled or rail-mounted inspection robots", "CMMS to receive the work orders: SAP PM, IBM Maximo, Fiix, MaintainX, UpKeep", "Any LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below, on a week of inspection readings, to rank the findings and draft work orders" ], "sources": [ "https://bostondynamics.com/case-studies/energy-savings-predictive-maintenance-at-ab-inbevs-largest-european-brewery/", "https://bostondynamics.com/case-studies/spot-inspires-innovation-at-bmw-group-plant/", "https://bostondynamics.com/case-studies/digital-transformation-at-nestle-purina/" ], "date_added": "2026-10-20", "last_updated": "2026-10-20" }, { "title": "AI reads technicians' free-text work orders and assigns the failure codes, so the CMMS shows what really fails (Janssen Pharmaceutica, oil and gas operator)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-failure-codes-from-maintenance-work-order-text/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-failure-codes-from-maintenance-work-order-text.md", "department": "Maintenance", "department_url": "https://aiagentsformanufacturing.com/departments/maintenance/", "evidence": "Pilot", "summary": "AI assigns failure codes from technicians' notes: at Janssen's Geel site a model got all three CMMS codes right on 83% of notifications.", "specialized": false, "who_this_is_for": null, "role": "Maintenance (reliability engineering: CMMS data and failure analysis)", "problem": "Every failure notification in the CMMS asks for codes: what was defective, the cause, the action taken. Technicians pick from drop-down lists in a hurry and write the real story in short, abbreviated free text. Codes end up wrong, vague or missing, so bad-actor lists, failure-mode analysis, preventive maintenance plans and spares decisions rest on data nobody trusts. Recoding years of history by hand against a failure-mode framework is a months-long job that rarely gets done.", "how_ai_helps": "A model reads the free text of each work order or notification (telegraphic, full of abbreviations, sometimes in the local language) and proposes the codes: defect, cause and action, or component, failure mechanism and cause against a failure-mode taxonomy. Each proposal carries a confidence level; low-confidence records, records that fit no code and records that are not failures (planned work, testing) go to a reliability engineer instead of being forced into a code. Comparing the proposals with the codes already entered also surfaces likely mislabels, ambiguous codes and gaps in the code list.", "result": "Janssen Pharmaceutica (Johnson & Johnson) production site in Geel, Belgium, with Ghent University: tens of thousands of Dutch-language failure notifications from 2010 to 2020, each carrying three codes (one of 16 defect, 17 cause and 15 action codes). The best model predicted the codes with 0.95 average accuracy per code and got all three right on 83% of notifications. A manual review of the disagreements found that only about 4.2% of notifications were truly mislabelled by the model; about 7.4% had at least one ambiguous code and about 5.7% had text too poor for anyone to code, which points to codes and writing rules to fix. The authors also note room for improvement in how the cause codes are defined (peer-reviewed paper in Computers in Industry). Global oil and gas operator (anonymous), proof of concept by Shivaan Asset Management with custom AI agents on SAP data: 3,126 historical ball-valve work orders classified in one batch run against a failure-mode taxonomy; 99.2% classified (the 24 others had nothing written against them), 73.7% came back actionable with 76.4% average confidence, and 40.7% described more than one failure (consultant's project page).", "companies": [ { "name": "Janssen Pharmaceutica", "url": "https://aiagentsformanufacturing.com/companies/#janssen-pharmaceutica" } ], "tools_deployed": [], "tools": [ "Custom text-classification models built by Ghent University on the site's CMMS data: Janssen Pharmaceutica, Geel", "Custom AI agents with SAP and Nexaan APM (Shivaan Asset Management): oil and gas proof of concept", "Other tools to evaluate: AI features of your CMMS or asset performance management tool (SAP PM, IBM Maximo, Fiix, MaintainX), reliability consultancies that offer work-order text analytics", "Any LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below, on an export of 100 to 200 work orders and your code list, for a first test" ], "sources": [ "https://doi.org/10.1016/j.compind.2022.103830", "https://biblio.ugent.be/publication/01GPB14QZFDBS0BJN3MT21V4HM", "https://shivaanam.com.au/projects/ai-work-order-failure-mode-classification-oil-gas/" ], "date_added": "2026-10-20", "last_updated": "2026-10-20" }, { "title": "AI reads supplier delivery notes and prepares the goods receipt in the ERP (Siemens, RAUMEDIC)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-delivery-note-goods-receipt-erp-siemens-raumedic/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-delivery-note-goods-receipt-erp-siemens-raumedic.md", "department": "Logistics / Supply Chain", "department_url": "https://aiagentsformanufacturing.com/departments/logistics/", "evidence": "Verified", "summary": "AI reads supplier delivery notes and prepares the goods receipt: Siemens reports 90%+ touchless processing and 98% accuracy; RAUMEDIC over 99%.", "specialized": false, "who_this_is_for": null, "role": "Logistics (receiving and goods receipt: inbound deliveries)", "problem": "Every delivery that reaches the dock comes with a delivery note in the supplier's own layout: printed, smudged, scanned or sent as an attachment. Someone keys the PO number, items, quantities and sometimes batch numbers and expiry dates into the ERP and checks them against the purchase order. With hundreds or thousands of suppliers it is slow, repetitive and error-prone; ERP stock lags physical stock, differences surface weeks later at invoice matching, and in food or regulated production a wrong batch or expiry date can mean scrapped material or a recall.", "how_ai_helps": "The AI reads each delivery note whatever the layout, extracts the supplier, PO number, items, quantities, batch numbers and dates, and checks them against the open purchase order, order confirmation and receiving requirements in the ERP. Clean notes are posted, or prepared for posting, automatically; differences such as quantities over or under the order, unknown items or missing batch data go to a person with the reason and the document attached. Receiving staff handle exceptions instead of keying every line.", "result": "Siemens (materials from more than 1,000 vendors; delivery notes previously entered by hand into SAP), with DeepOpinion, now Otera, an LLM-based document platform: over 90% touchless processing within 2 weeks, 98% accuracy, more than €5M in annual ROI, and delivery-note processing cut from days to minutes (CaseStudies.com summary of the vendor case; Otera success story). RAUMEDIC, at its logistics centers, with MHP on UiPath RPA plus Document Understanding: the AI model, first trained on delivery notes from three suppliers chosen for their high document volume, reads delivery-note data with over 99% accuracy; the remaining 1% is reworked by a person, the proof of concept took a few weeks and the next phase rolls it out to more suppliers (MHP case study). Food producer with over 10,000 employees (anonymous), Nordoon AI agents: delivery notes checked against purchase orders and order confirmations, with batch numbers, production and expiry dates fed into the ERP so MRP and scheduling can respect expiry; no figures published (Nordoon customer story). All vendor or integrator-published.", "companies": [ { "name": "RAUMEDIC", "url": "https://aiagentsformanufacturing.com/companies/#raumedic" }, { "name": "Siemens", "url": "https://aiagentsformanufacturing.com/companies/#siemens" } ], "tools_deployed": [ { "name": "Nordoon", "url": "https://aiagentsformanufacturing.com/tools/#nordoon" }, { "name": "Otera (formerly DeepOpinion)", "url": "https://aiagentsformanufacturing.com/tools/#otera-formerly-deepopinion" }, { "name": "UiPath Document Understanding", "url": "https://aiagentsformanufacturing.com/tools/#uipath-document-understanding" } ], "tools": [ "DeepOpinion / Otera (LLM-based document agents): used at Siemens", "UiPath Document Understanding with RPA, implemented by MHP: used at RAUMEDIC", "Nordoon AI agents: used at the food producer", "Other tools to evaluate: document AI in or around your ERP (SAP Document AI, Microsoft Dynamics 365), ABBYY, Rossum, Hyperscience", "Any LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below, on photos of a few delivery notes and the matching open POs, for a first test" ], "sources": [ "https://www.casestudies.com/company/deepopinion/case-study/siemens-jumped-to-90-touchless-processing-in-2-weeks", "https://www.otera.ai/success-stories/siemens-autonomous-delivery-processing", "https://www.mhp.com/en/insights/our-experience/raumedics-sustainable-automated-incoming-goods-process", "https://www.nordoon.ai/customer-stories/ai-agents-simplify-delivery-note-processing-regulated-manufacturing" ], "date_added": "2026-10-20", "last_updated": "2026-10-20" }, { "title": "AI predicts when shipments will really arrive, so planners act before a late delivery stops a line (Faurecia, Magna)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-predictive-eta-shipment-tracking-line-stops-faurecia-magna/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-predictive-eta-shipment-tracking-line-stops-faurecia-magna.md", "department": "Logistics / Supply Chain", "department_url": "https://aiagentsformanufacturing.com/departments/logistics/", "evidence": "Verified", "summary": "Machine-learning ETAs warn planners before late shipments stop a line: Faurecia (now Forvia) cut urgent transport costs by €0.5M and Shippeo projects €1.9M a year in total.", "specialized": false, "who_this_is_for": null, "role": "Logistics (transportation and supply chain: shipment tracking and expediting)", "problem": "For a plant that runs just in time or just in sequence, a late inbound shipment means a line stop, and a late delivery to the customer means penalties. Visibility usually ends at the carrier's booking confirmation: the planner learns about the delay when the parts do not show up, then calls carriers, books expensive urgent transport and reshuffles the schedule. Unplanned waiting times and extra distance come back later as extra carrier charges. Faurecia Seating described all three costs: extra carrier costs, very costly line stops when supply was disrupted, and high penalties for late deliveries.", "how_ai_helps": "A visibility platform connects to carriers' tracking data (GPS, telematics, carrier systems) and to the orders in the ERP or transport management system. A machine-learning model recalculates each shipment's estimated time of arrival (ETA) from its position and the constraints of the route, and alerts the right people automatically when a shipment will miss its slot or the time the material is needed. The planner decides early: expedite, re-sequence, build another product on the line, or warn the customer. The same data shows dwell times and carrier performance for carrier reviews and milk-run design.", "result": "Faurecia Seating (74 production sites in 23 countries, customers including Porsche, BMW and Mercedes), Shippeo real-time transportation visibility with machine-learning ETAs, deployed in all its European plants by its supply chain operations team in Poland: 180 carriers connected through more than 40 API connections, 17,000 orders tracked a month, over 40 daily users (Industry Europe, 2022); 15% less workload for the supply chain team and €0.5 million less in urgent, unplanned transport costs; Shippeo labels the €1.9 million in combined annual savings as projected, not achieved (Shippeo customer story, Nov 2020, which names the company Forvia, the group Faurecia became). Magna Steyr, Graz (contract vehicle manufacturing, about 1,700 supplier pick-up points): real-time tracking of highly critical transports since 2016, with an automatic notification when a transport fails to arrive on schedule, and trailer-level tracking for long-distance just-in-sequence deliveries; when parts for one model are delayed, a line building several models can temporarily build more of the others (Magna article, 2024). Magna with project44 on long-distance just-in-sequence deliveries across Europe: 4-week rollout, shipments doubled with no added headcount, 40% more on-time deliveries and 40% cost savings, with no cost base, period or amounts given (project44 one-page case study, 2019). Figures are published by the vendors or the companies.", "companies": [ { "name": "Faurecia", "url": "https://aiagentsformanufacturing.com/companies/#faurecia" }, { "name": "Magna", "url": "https://aiagentsformanufacturing.com/companies/#magna" } ], "tools_deployed": [ { "name": "project44", "url": "https://aiagentsformanufacturing.com/tools/#project44" }, { "name": "Shippeo", "url": "https://aiagentsformanufacturing.com/tools/#shippeo" } ], "tools": [ "Shippeo (real-time transportation visibility with machine-learning ETAs): used at Faurecia", "project44 (visibility platform): used at Magna for long-distance just-in-sequence deliveries", "Other tools to evaluate: FourKites, Transporeon (Trimble) visibility, Descartes MacroPoint, the tracking module of your TMS, and your forwarder's visibility service for ocean and air shipments", "Any LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below, on an export of open shipments with their latest ETAs and the production plan, to see which delays matter" ], "sources": [ "https://industryeurope.com/sectors/transportation/faurecia-limits-risk-gains-competitive-advantage-with-rttv/", "https://www.shippeo.com/resources/explore/customer-stories/forvia", "https://www.magna.com/stories/inside-automotive/2024/going-digital-in-the-name-of-resilience-and-sustainability---multi-oem-logistics-of-contract-manufacturers-in-the-automotive-industry", "https://cdn.featuredcustomers.com/CustomerCaseStudy.document/p44_EU_Case_Study_-_Case_Study_-_Magna.pdf" ], "date_added": "2026-10-20", "last_updated": "2026-10-26" }, { "title": "An AI agent chases a late PO in Zoho Inventory and updates it after approval (Northline Components)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-agent-late-po-follow-up-zoho-inventory/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-agent-late-po-follow-up-zoho-inventory.md", "department": "Procurement", "department_url": "https://aiagentsformanufacturing.com/departments/procurement/", "evidence": "In-house demo", "summary": "In-house demo by Frédéric D'Amours with Grok Bot: an AI agent finds a late PO in Zoho Inventory, drafts the supplier email, reads the reply and updates the PO after approval.", "specialized": false, "who_this_is_for": null, "role": "Procurement (buyer or purchasing coordinator: purchase order follow-up and expediting)", "problem": "When a purchase order passes its delivery date without being received, someone has to notice it, write to the supplier, wait for the answer, then update the ERP with the new date and any price change, and confirm back with a revised PO. The steps are simple but easy to forget, and each change has to be checked against the original PO before anything is written. This page documents a demo Frédéric D'Amours built himself to test that loop end to end on a small ERP.", "how_ai_helps": "An AI agent built on Grok Bot, the agent platform, and set up as a purchasing assistant, works with Zoho Inventory through its REST API, using a few Python scripts. On request, a read-only script lists purchase orders that are issued, open or partially received, not fully received, and past their delivery date. The agent drafts an email to the supplier asking for a firm new date and tracking. When the reply arrives, it extracts the new date and any price change, checks the 'from' price against the PO, and asks the buyer to approve: date only, date and price, or nothing. Only after a yes does it write to Zoho, then reads the PO back to confirm. Finally it drafts the confirmation to the supplier with the revised PO PDF attached and asks again before anything goes out. Writes run in dry-run by default, every change is logged with its original value so it can be restored, and every record the agent creates is tagged.", "result": "In-house demo developed and tested by Frédéric D'Amours with Grok Bot, not a customer case. Northline Components Inc. is a fictional company on a free Zoho Inventory organisation (Canadian data centre); all items, suppliers, POs and customers are test data marked NLDEMO, and Frédéric D'Amours wrote the supplier's replies himself, playing the supplier. Tested on 1 purchase order (NLDEMO-PO-OPEN-001), made late on purpose by moving its delivery date to Oct 18, 2026. The agent found it as 1 day late with 200 units outstanding (USD 548.00), drafted the supplier email, read the reply (new ship date Nov 3, 2026; NL-CAB-0006 from USD 4.59 to 4.84 per unit for a copper surcharge, +USD 25.00), and after approval wrote Nov 3 into the PO's Zoho field delivery_date (the field the late check reads; the script also added a note line 'ETA updated' and logged the original value, Oct 18) and raised the NL-CAB-0006 line rate to 4.84, for a total of USD 573.00, then prepared the reply with the revised PO PDF. 33 Zoho Inventory API calls were sent in total over two days of building and testing (28 reads, 5 writes; the free plan allows 1,000 a day). Processing time per PO and manual time saved were not measured: the timestamps from the test session include retakes and filming. During testing, the agent caught a wrong reference price in one supplier reply (6.20 instead of the PO's 4.59) and refused to apply it.", "companies": [], "tools_deployed": [], "tools": [ "Zoho Inventory (free plan) through its REST API v1, with OAuth: used in the demo", "Python scripts: a shared Zoho client plus scripts to detect late POs (read-only), draft the supplier email, parse the reply, apply the new date and price, and reset the demo data", "Grok Bot, the agent platform: the agent runs there and keeps emails as Gmail drafts that a person sends; it never sends from Zoho", "Other tools to evaluate: the same loop works with any ERP or inventory system that has an API (purchase order read and update) and a mailbox the agent can read", "Any LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below, on an export of open POs and the supplier's reply" ], "sources": [ "https://www.zoho.com/inventory/api/v1/" ], "date_added": "2026-10-20", "last_updated": "2026-10-26" }, { "title": "AI reads each molding cycle and tells the setter which machine setting to adjust, so scrap and start-up losses fall (Beko, Zahoransky)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-injection-molding-setpoint-recommendations-beko-zahoransky/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-injection-molding-setpoint-recommendations-beko-zahoransky.md", "department": "Production", "department_url": "https://aiagentsformanufacturing.com/departments/production/", "evidence": "Verified", "summary": "AI reads each molding cycle and advises which setting to change: Beko cut scrap 52% on one part; Zahoransky cut start-up scrap 10–17%.", "specialized": true, "who_this_is_for": "Plastic injection molders in series production that can record data from each cycle, have a validated process window (the settings range known to give good parts) and have setters or process engineers to act on recommendations.", "role": "Production (injection molding: process setting and shift operation)", "problem": "An injection molding recipe that was right at sampling drifts in series production: recycled material varies from batch to batch, humidity and temperature change, and the same machine, mold and material do not give the same part every day. Short shots (incomplete parts), flash (excess plastic at the edges), sink marks (small dents) and weld lines (visible seams where plastic flows meet) appear, and the fix depends on how fast and how well the setter on shift reacts. Experienced setters are hard to find and slow to train. At a global medtech manufacturer quoted in the Zahoransky article, changing material batches caused significant scrap and up to 30% swings in OEE (overall equipment effectiveness, a standard measure of productive time) between shifts and batches; the Beko researchers note that scrap rates vary with operators' competence and their ability to intervene quickly.", "how_ai_helps": "Sensors capture every cycle: cavity pressure curves (pressure measured inside the mold during each shot) at Beko, and at Zahoransky the machine and peripheral data (hot runner, the heated channels that feed melted plastic into the mold; dryer, temperature control units), the properties of the material batch in use, ambient conditions and quality test results. A model learns which pattern belongs to good parts and, when a cycle drifts out of that window, recommends which setting to change and in which direction (injection pressure, speed or time, holding pressure or time, shot volume at Beko; up to 60 setting values at Zahoransky, always inside the validated limits). The setter or process engineer confirms the change; at Zahoransky the vendor says the tool can also write settings back to the machine within the validated windows.", "result": "Beko Corporate (home appliances) and its suppliers, peer-reviewed study with Koç University: cavity pressure sensors and a neural network that suggests increases or decreases for six machine parameters, used in mass production for 6 months on two parts. On a dishwasher bottom cover molded from 100% recycled 30% glass-filled polypropylene on a 600-ton machine: cycle time down 18%, scrap rate down 800 PPM, and 500 minutes of assembly-line downtime caused by this part eliminated. On a refrigerator inner panel (HIPS, 1,000-ton machine): visual defects eliminated after 6 months, scrap down 52% and start-up scrap down 49%. Model accuracy was 97% and 93% on the two parts (a classification accuracy, not proof that each recommended change gives a good part) (International Journal of Advanced Manufacturing Technology, 2025, open access). Zahoransky (machine builder, Freiburg) with plus10's Hopper on the ramp-up of a 16-cavity line for staked-needle syringes in COC/COP: according to customer feedback, ramp-up time 20% shorter than on similar earlier projects and scrap in the early production phase down 10 to 17%, at an annual output in the millions (Kunststoffe International, 2025, article co-written by plus10 and Zahoransky).", "companies": [ { "name": "Beko", "url": "https://aiagentsformanufacturing.com/companies/#beko" }, { "name": "Zahoransky", "url": "https://aiagentsformanufacturing.com/companies/#zahoransky" } ], "tools_deployed": [ { "name": "plus10", "url": "https://aiagentsformanufacturing.com/tools/#plus10" } ], "tools": [ "Cavity pressure sensors (Kistler) with a custom neural network built by Beko and Koç University: used at Beko", "plus10 Hopper (live-learning setting recommendations) and Shannon (operator assistance during stops): used at Zahoransky", "Other tools to evaluate: your machine maker's assistance systems (for example Arburg, Engel, KraussMaffei), cavity pressure monitoring (Kistler ComoNeo, RJG eDART), and AI process tools such as Gauss ML Optimyzer", "Any LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below, on a shift's cycle data and defect log, to prepare a troubleshooting sheet for the setter" ], "sources": [ "https://link.springer.com/article/10.1007/s00170-025-15611-x", "https://en.kunststoffe.de/a/specialistarticle/smart-machine-ramp-ups-stable-series-pro-7129063" ], "date_added": "2026-10-21", "last_updated": "2026-10-26" }, { "title": "Automation and AI check supplier certificates of analysis against your specifications before material is released (Georgia-Pacific, C.H. Guenther, two anonymous manufacturers)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-supplier-certificate-of-analysis-check-georgia-pacific/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-supplier-certificate-of-analysis-check-georgia-pacific.md", "department": "Quality", "department_url": "https://aiagentsformanufacturing.com/departments/quality/", "evidence": "Verified", "summary": "Automation and AI check supplier certificates of analysis against your specs: Georgia-Pacific cut manual CoA inspections 95% (automation); an anonymous Fortune 500 maker runs 80%+ touchless (vendor figures).", "specialized": false, "who_this_is_for": null, "role": "Quality (supplier quality and incoming inspection: certificates of analysis)", "problem": "Raw materials arrive with a certificate of analysis (CoA) from the supplier: a PDF or scan, in the supplier's own layout, test names, units and sometimes language. Before the lot can be used, someone reads each result, compares it with the material specification, and often types the values into SAP QM or a LIMS. At volume this becomes one of the largest manual jobs in quality: lots wait in quarantine, production schedules absorb the delay, and errors creep in at re-keying. C.H. Guenther (food manufacturer) had one technician compare every CoA with the specification sheet and a second person repeat the check because customers required double verification.", "how_ai_helps": "The AI reads each incoming CoA in the formats and languages it has been set up and tested for (Otera reports more than 20 languages), extracts the material, batch, tests, values and units, maps each supplier test name to your own inspection characteristic, and checks every result against the specification and supplier rules. In-spec, complete lots are recorded or proposed for release; out-of-spec, missing or ambiguous results go to a quality specialist with the value, the limit and the source page already shown. The usage decision stays with QA under the rules it approves.", "result": "Georgia-Pacific (pulp and paper, over 30,000 employees, 180+ locations), Specright specification management: CoAs from paper suppliers scored automatically on arrival and failing inbound materials flagged for immediate review; manual CoA inspections cut by 95%, from hundreds a day to a handful, now expanding beyond the paper division (Specright case study). Specright describes this as automation of CoA scoring against specifications; the case study does not say that AI reads or compares the certificates. Fortune 500 pharma manufacturer (anonymous, over 580,000 CoAs a year from thousands of suppliers in more than 20 languages), Otera AI agents: over 80% of CoAs now flow touchless into SAP and LIMS; validation workload from the equivalent of 70 full-time employees to 9 (Otera calls this 87% lower cost; it is freed capacity, not a measured drop in total spending, and software and integration costs are not given); cycle time from 2 to 5 days to hours; exception handling from 15 minutes to under 3; error rate 3.5 times lower than manual (Otera success story). Regulated manufacturer (anonymous, about 30,000 SAP QM inspection lots a year), Artificio: 91% of lots recorded touchless and clean-lot processing from about 12 minutes to 40 seconds (Artificio case study, customer-provided metrics). C.H. Guenther with TraceGains integrated to SAP: CoAs compared with specifications automatically and out-of-spec lots put on hold in SAP, so a shipment can be rejected before it leaves the supplier; no figures published, and the testimonial does not describe any AI in this set-up, only electronic comparison with the specifications and blocking rules. All figures are vendor-published.", "companies": [ { "name": "Georgia-Pacific", "url": "https://aiagentsformanufacturing.com/companies/#georgia-pacific" } ], "tools_deployed": [ { "name": "Artificio", "url": "https://aiagentsformanufacturing.com/tools/#artificio" }, { "name": "Otera (formerly DeepOpinion)", "url": "https://aiagentsformanufacturing.com/tools/#otera-formerly-deepopinion" }, { "name": "Specright", "url": "https://aiagentsformanufacturing.com/tools/#specright" }, { "name": "TraceGains", "url": "https://aiagentsformanufacturing.com/tools/#tracegains" } ], "tools": [ "Specright (specification management with CoA scoring): used at Georgia-Pacific", "Otera, formerly DeepOpinion (AI agents for document validation): used at the Fortune 500 pharma manufacturer", "Artificio (CoA to SAP QM results recording): used at the regulated manufacturer", "TraceGains (supplier compliance, integrated with SAP): used at C.H. Guenther", "Other tools to evaluate: the supplier portal or quality module of your ERP (SAP QM, QAD, Plex), your LIMS, and document AI platforms such as ABBYY or Rossum", "Any LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below, on one CoA and the matching specification, for a first test" ], "sources": [ "https://cdn.featuredcustomers.com/CustomerCaseStudy.document/specright_georgia-pacific_894736.pdf", "https://www.otera.ai/success-stories/fortune-500-manufacturer-autonomous-coa-validation", "https://artificio.ai/case-studies/sap-result-recording", "https://tracegains.com/testimonials/c-h-guenther-case-study/" ], "date_added": "2026-10-21", "last_updated": "2026-10-26" }, { "title": "AI runs a first-pass check of drawings and models before the design review, so engineers discuss real issues (INFICON, McDermott)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-first-pass-drawing-design-review-inficon-mcdermott/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-first-pass-drawing-design-review-inficon-mcdermott.md", "department": "Engineering", "department_url": "https://aiagentsformanufacturing.com/departments/engineering/", "evidence": "Verified", "summary": "AI checks drawings before the design review: INFICON cut review closure from 62 to 7 days; McDermott checks 100 P&IDs in about 3 minutes.", "specialized": false, "who_this_is_for": null, "role": "Engineering (design engineering and drafting: drawing checking and design review)", "problem": "Drawing checks are repetitive and detail-heavy: missing dimensions, conflicting tolerances, wrong hole callouts, title blocks, material notes, and the same mistakes the team has flagged before. They are easy to miss when engineers are at capacity, and a mistake found late costs a new loop through drawings, suppliers and parts. At INFICON, reviewers often saw a design for the first time in the meeting, so a one-hour review turned into several hours and the format depended on who led it; its engineering manager estimates that a significant late-stage problem can nearly double an 8 to 12 week development cycle. At McDermott, checking piping and instrumentation diagrams (P&IDs), some up to 100 pages, is done several times per project by experienced engineers and can take up to tens of hours per project.", "how_ai_helps": "Each new drawing or drawing update goes through an AI checker before a person reviews it. At INFICON, CoLab's AutoReview runs a first pass on 2D drawings and 3D parts against engineering standards and the team's own checklists, and puts its markups on the design; engineers then review those findings and add their own feedback asynchronously before the meeting. At McDermott, the in-house ReCo tool uses image recognition to find common mistakes in P&IDs (and later in piping isometrics, including third-party PDFs); the engineer decides whether each flag is a real error and how to fix it. The AI covers the baseline checks; engineers keep the judgment calls.", "result": "INFICON (measurement and control instruments, Bad Ragaz, Switzerland; teams in Europe, the US and Asia), CoLab with AutoReview since March 2026: design-review closure time from 62 days to 7 (89% shorter); design feedback captured from about 50 to more than 400 pieces a month; one full development program reached first-time-right with no additional engineering changes, which the team estimates removed four to six weeks from that cycle. These results come from the whole move to CoLab's shared review workflow, of which the AI first pass is one part (CoLab case study, vendor-published). McDermott (engineering and construction for the energy industry, 54+ countries), ReCo tool built on the VIKTOR platform: checks that can take up to tens of hours per project are done in seconds, at 100 P&IDs in about 3 minutes; the tool also gives junior engineers a reference manual of common P&ID mistakes (VIKTOR customer interview with McDermott's AI projects manager, 2023).", "companies": [ { "name": "INFICON", "url": "https://aiagentsformanufacturing.com/companies/#inficon" }, { "name": "McDermott", "url": "https://aiagentsformanufacturing.com/companies/#mcdermott" } ], "tools_deployed": [ { "name": "CoLab AutoReview", "url": "https://aiagentsformanufacturing.com/tools/#colab-autoreview" } ], "tools": [ "CoLab with AutoReview (AI first pass on drawings and 3D models against standards and checklists): used at INFICON", "McDermott's in-house ReCo tool (deep-learning image recognition), built on VIKTOR: used for P&IDs and piping isometrics", "Other tools to evaluate: the checking functions of your CAD or PLM (for example SOLIDWORKS Design Checker, NX Check-Mate), drawing-reading AI such as Werk24 or CADDi, and DFM checkers from your CAD vendor", "Any LLM that reads images (ChatGPT, Claude, Copilot, Gemini) with the prompt below, on one exported drawing and your checklist, for a first test" ], "sources": [ "https://www.colabsoftware.com/case-studies/inficon", "https://www.colabsoftware.com/product/autoreview", "https://www.viktor.ai/customer-cases/46/mcdermott-artificial-intelligence-AI-piping-instrumentation-diagram-image-recognition" ], "date_added": "2026-10-21", "last_updated": "2026-10-21" }, { "title": "AI collections: prioritised worklists and automatic reminders get customers to pay on time (EZG Manufacturing, Yaskawa, Lodge Cast Iron)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-collections-past-due-invoices-ezg-yaskawa-lodge/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-collections-past-due-invoices-ezg-yaskawa-lodge.md", "department": "Office & management", "department_url": "https://aiagentsformanufacturing.com/departments/office/", "evidence": "Verified", "summary": "AI collections for manufacturers: EZG automated 95% of outreach and cut DSO 5 days; Yaskawa cut DSO 5.5 days with AI-prioritised worklists.", "specialized": false, "who_this_is_for": null, "role": "Finance (accounts receivable: credit and collections)", "problem": "In many plants, collections is a part-time job: the same one or two people also handle cash application, billing and other finance work. Reminders go out when someone has time, the worklist is a spreadsheet or the aging report, and the biggest or most at-risk accounts are not always called first. Past dues grow, days sales outstanding (DSO) creeps up, and the company finances its customers. Lodge Cast Iron ran collections and dunning in spreadsheets with faxes and Outlook; Yaskawa America's past-due receivables were rising because the collections worklist was prioritised by hand; EZG Manufacturing's two-person AR team had a clear strategy but could not execute it consistently next to its other duties.", "how_ai_helps": "The collections software connects to the ERP's open invoices and payment history. It ranks accounts by amount, age and the customer's payment behaviour so the team works the right ones first, and it runs the routine outreach on its own: due-date reminders, follow-up emails, automated phone calls at EZG, with each reply and promise to pay logged. Disputes, short payments and sensitive accounts go to a person. The AR team decides on holds, payment plans and escalation.", "result": "EZG Manufacturing (industrial manufacturer with several companies under one umbrella, two-person AR team), Stuut AI collections: live within weeks, phased from due-date reminders to weekly emails and automated phone calls; 95% of collections outreach now automated across 1,597 touchpoints, $11.67 million collected automatically (43% of total AR collected), DSO 5 days lower and about 20 hours a week returned to the team; extended to sister company Malta Dynamics (Stuut case study). Yaskawa America (drives, motion control and robotics; 1,000+ employees), HighRadius with AI-based worklist prioritisation and automated correspondence: DSO down 5.5 days, A/R team productivity up 60% and a collection effectiveness index of about 90% (HighRadius case study). Lodge Cast Iron (cookware maker, 550+ employees), HighRadius prioritised automated collections workflows plus AI short-payment identification: the headline says 'past dues down 13%' and 'DSO efficiency up 5 days', but the customer's own quote in the PDF gives the share of aging invoices going from 22% to 9% in a year (13 percentage points) and the gap between best-possible and actual DSO going from 11 to 6 days (not a 5-day cut in DSO itself); 84% collections efficiency; and 96% net recovery rate on invalid deduction disputes, not on all past-due invoices (HighRadius case study PDF). All figures are vendor-published.", "companies": [ { "name": "EZG Manufacturing", "url": "https://aiagentsformanufacturing.com/companies/#ezg-manufacturing" }, { "name": "Lodge Cast Iron", "url": "https://aiagentsformanufacturing.com/companies/#lodge-cast-iron" }, { "name": "Yaskawa", "url": "https://aiagentsformanufacturing.com/companies/#yaskawa" } ], "tools_deployed": [ { "name": "HighRadius", "url": "https://aiagentsformanufacturing.com/tools/#highradius" }, { "name": "Stuut", "url": "https://aiagentsformanufacturing.com/tools/#stuut" } ], "tools": [ "Stuut (AI collections outreach by email and phone): used at EZG Manufacturing", "HighRadius (AI-prioritised collections worklists, automated correspondence, deductions): used at Yaskawa America and Lodge Cast Iron", "Other tools to evaluate: Esker, Billtrust, Versapay, Upflow, Chaser, Tesorio, and the collections module of your ERP (SAP, Microsoft Dynamics 365, NetSuite)", "Any LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below, on your aging report, to build this week's call list and draft reminders" ], "sources": [ "https://www.stuut.ai/case-studies/ezg-manufacturing", "https://www.highradius.com/en-gb/resources/case-studies/yaskawa/", "https://cdn.featuredcustomers.com/CustomerCaseStudy.document/highradius_lodge-cast-iron_893087.pdf" ], "date_added": "2026-10-21", "last_updated": "2026-10-26" }, { "title": "AI agents negotiate with long-tail suppliers that buyers never had time to call (Honeywell, and separately an anonymous German consumer and industrial goods maker)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-agent-tail-spend-supplier-negotiation-honeywell-pactum/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-agent-tail-spend-supplier-negotiation-honeywell-pactum.md", "department": "Procurement", "department_url": "https://aiagentsformanufacturing.com/departments/procurement/", "evidence": "Verified", "summary": "AI agents negotiate with long-tail suppliers: Honeywell ran 2,500+ negotiations on $500M+ of spend; a German maker saved $1.2M+.", "specialized": false, "who_this_is_for": null, "role": "Procurement (indirect and tail spend: price, payment terms and rebate negotiations)", "problem": "Buyers negotiate the big contracts with the top suppliers. The long tail, hundreds or thousands of smaller suppliers, indirect items and non-strategic parts, is bought at whatever price and payment terms the supplier last quoted, because each negotiation takes time that nobody has. Honeywell describes it as hundreds of supplier contracts that were too time-consuming to negotiate manually yet too numerous to ignore: standardizing payment terms across a sprawling supplier base was slow, inconsistent and resource-heavy, so savings were left on the table.", "how_ai_helps": "Procurement sets the mandate: which suppliers and categories are in scope, the targets (price, payment terms, rebates, item discounts) and the limits the agent may trade within. The AI agent then contacts each supplier through a chat interface, negotiates within that mandate, and records the agreed terms; at the German manufacturer it is integrated with SAP Ariba for requisition and indirect cost negotiations. Strategic suppliers, anything outside the mandate and every final contract stay with the buyers.", "result": "Honeywell (Fortune 100 industrial company), Pactum AI agents, starting with an indirect spend pilot: over $500M of spend managed across categories, more than 2,500 autonomous negotiations concluded, and 80% of suppliers contacted actively engaged; in one negotiation alone, $65,000 in savings with zero buyer involvement (a single example: no typical or median saving per negotiation is given). Global consumer and industrial goods manufacturer headquartered in Germany (anonymous), a Pactum customer for nearly three years in more than 20 countries: over $131M of spend processed through more than 11,000 completed autonomous negotiations, delivering over $1.2M in savings, starting with requisition and indirect cost negotiations integrated with SAP Ariba, then post-sourcing negotiations, tactical sourcing and price-list renegotiations. A leading automotive group (anonymous) used the same approach on non-strategic OEM parts in direct procurement and reports a 15.3% value improvement; the summary does not define 'value' (price, discount or payment terms), the comparison base or the period (CaseStudies.com summary). All figures are published by the vendor.", "companies": [ { "name": "Honeywell", "url": "https://aiagentsformanufacturing.com/companies/#honeywell" } ], "tools_deployed": [ { "name": "Pactum AI", "url": "https://aiagentsformanufacturing.com/tools/#pactum-ai" } ], "tools": [ "Pactum AI (autonomous supplier negotiation agents, SAP Ariba integration): used at Honeywell, the German manufacturer and the automotive group", "Other tools to evaluate: tail-spend sourcing tools such as Fairmarkit or Keelvar, and the negotiation or sourcing modules of your procurement suite (SAP Ariba, Coupa, Jaggaer)", "Any LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below, on your spend data, to pick the suppliers worth negotiating and draft a first request" ], "sources": [ "https://pactum.com/testimonial/honeywell", "https://pactum.com/testimonial/1.2m-in-savings-for-a-global-manufacturer", "https://www.casestudies.com/company/pactum-ai/case-study/leading-automotive-group-customer-case-study" ], "date_added": "2026-10-22", "last_updated": "2026-10-26" }, { "title": "Inventory drones count the plant warehouse on a regular schedule, so inventory is right without stopping production (LAPP USA, GKN Automotive, Hyundai Glovis)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-drone-inventory-cycle-counts-plant-warehouse-lapp-gkn/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-drone-inventory-cycle-counts-plant-warehouse-lapp-gkn.md", "department": "Logistics / Supply Chain", "department_url": "https://aiagentsformanufacturing.com/departments/logistics/", "evidence": "Verified", "summary": "Drones count the plant warehouse on a schedule: LAPP USA went from 2 to 26 full counts a year; GKN divided stock-check time by three.", "specialized": false, "who_this_is_for": null, "role": "Logistics (plant warehouse: cycle counts and inventory accuracy)", "problem": "High-bay racks are counted by people on lifts, so full counts happen rarely, often on weekends or by stopping work. Between counts, misplaced pallets and reels, wrong labels and stock that is in the system but not on the shelf delay orders and production. LAPP USA ran only two full warehouse passes a year, with five associates (about 12.5% of the warehouse team) doing weekend cycle counts and still missing targets. GKN Automotive's inventory specialist checked locations from an aerial lift, and twice a year a full inventory needed 5 aerial lifts and 10 operators; the company wanted regular stock takes without stopping production for a whole day.", "how_ai_helps": "Indoor drones fly the aisles on a schedule (LAPP's Corvus drones fly nightly missions after people leave, with no human pilot), and read pallet and rack labels with cameras and computer vision, without GPS or changes to the racks. The set-ups differ: EYESEE, for example, locates the drone with two mobile ultra-wideband (UWB) tags placed on the ground during each flight, with no fixed infrastructure. Each scan is compared with the warehouse management system (WMS) and the drone keeps a photo of every location, so the team gets a daily discrepancy list (missing, misplaced, mislabelled) and corrects it from a desk instead of walking the racks. People still decide each correction in the WMS.", "result": "LAPP USA (industrial cable manufacturing and distribution, 134,000 sq ft site in Brownsburg, Indiana), Corvus One drones flying nightly missions from 6 PM: 26 full-warehouse counts a year (about every two weeks) instead of two; inventory team from five associates to two (60% less labor), the other three moved to picking and manufacturing; weekend cycle-count overtime eliminated; mislabelled, misplaced or missing reels flagged for next-day resolution, and label-placement practices standardized after the scans showed inconsistencies (vendor case study). GKN Automotive (driveline systems), EYESEE drones: since 2022 a stock take every month in all its warehouses without stopping production, time spent on stock checks divided by three, and less work at height (vendor page with GKN's supply chain manager quoted). Bayer CropScience (Villefranche-Limas site, France), EYESEE: cycle inventory all year since 2021, 100% pallet reading rate, an average of 400 pallets per hour and 10 seconds per location, and no work-related accidents during inventories (vendor page). Hyundai Glovis at the Hyundai Motor Group Metaplant America consolidation center (Ellabell, Georgia): two autonomous drones inspect the inventory of semi-knockdown car parts, with inspection cycles cut from 300 minutes to 30 (Aju Press, 2025).", "companies": [ { "name": "Bayer", "url": "https://aiagentsformanufacturing.com/companies/#bayer" }, { "name": "GKN Automotive", "url": "https://aiagentsformanufacturing.com/companies/#gkn-automotive" }, { "name": "Hyundai Glovis", "url": "https://aiagentsformanufacturing.com/companies/#hyundai-glovis" }, { "name": "LAPP USA", "url": "https://aiagentsformanufacturing.com/companies/#lapp-usa" } ], "tools_deployed": [ { "name": "Corvus Robotics", "url": "https://aiagentsformanufacturing.com/tools/#corvus-robotics" }, { "name": "EYESEE (Hardis Group)", "url": "https://aiagentsformanufacturing.com/tools/#eyesee-hardis-group" } ], "tools": [ "Corvus Robotics Corvus One (autonomous inventory drones and AIMS platform): used at LAPP USA", "EYESEE drone inventory (invented by Hardis Group, marketed since March 2022 by Darwin Drones): used at GKN Automotive and Bayer", "Hyundai Glovis in-house drone system linked to the site's consolidation-center system: used at HMGMA", "Other tools to evaluate: Gather AI, Verity, and fixed or forklift-mounted camera systems", "Any LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below, on a discrepancy export, to sort and explain what to fix first" ], "sources": [ "https://www.corvus-robotics.com/case-study-lapp", "https://www.eyesee-drone.com/client/gkn-automotive/", "https://www.eyesee-drone.com/client/bayer/", "https://www.ajupress.com/view/20250707152254306", "https://www.eyesee-drone.com/the-eyesee-inventory-drone-plots-a-new-course/" ], "date_added": "2026-10-22", "last_updated": "2026-10-26" }, { "title": "AI watches conveyor belts and stops them quickly when a foreign object hits, to limit the damage (Cemex, an Indian steel plant, JFE Steel)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-conveyor-belt-rip-foreign-object-detection-cemex-jfe/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-conveyor-belt-rip-foreign-object-detection-cemex-jfe.md", "department": "Maintenance", "department_url": "https://aiagentsformanufacturing.com/departments/maintenance/", "evidence": "Verified", "summary": "AI stops a conveyor within metres of a foreign-object hit: Cemex estimates £700–800K avoided in one incident; a steel plant cut belt downtime 38%.", "specialized": true, "who_this_is_for": "Sites with long, heavy-duty conveyor belts that carry bulk material (quarries, cement plants, steel mills), where a torn belt stops the whole line.", "role": "Maintenance (bulk material handling: conveyor belt condition and protection)", "problem": "Conveyor belts carry raw materials around the clock in steel plants, cement plants and quarries. A piece of scrap, a fallen tool or a steel plate landing on the belt can start a rip that runs along the belt in seconds, and a torn belt means hours to weeks of downtime, repair costs and idle downstream equipment. Inspection is usually periodic walkthroughs: by the time someone sees the object, the damage is done. At an integrated steel plant in eastern India, unplanned downtime events ran at 3 to 4 a month, a single belt damage event cost over $86,000 in lost production and repairs in the month before the project, and incidents were logged on paper.", "how_ai_helps": "Two approaches are in use. Camera-based: a fixed industrial camera films the belt and a computer vision model (software that interprets camera images) trained on the site's images detects foreign objects or belt damage; at the Indian plant a confirmed detection stops the conveyor through the PLC (programmable logic controller, the industrial computer that runs the conveyor) and alerts staff by WhatsApp and SMS, and JFE Steel also uses the images to spot belt edge misalignment. Drive-based: the Flexco system at Cemex learns the normal trend of the motor current and drive data and triggers a shutdown alarm on an abnormal deviation, with no camera or sensor in the belt. Every event is logged with time and, for cameras, a snapshot, so maintenance can see patterns.", "result": "Cemex, Dove Holes Quarry (UK, about five million tonnes of limestone aggregate a year), Flexco Elevate Belt Rip Prevention on its most critical production conveyor: when a steel plate fell from equipment above the primary crusher onto the running belt, the system detected the impact, raised a shutdown alarm, notified maintenance and the belt stopped within 12 meters of the incident, which kept the rip from spreading (the plate still damaged the belt: the system limited the damage after the impact, it did not prevent the impact); Cemex estimated that losses of £700,000 to £800,000 were avoided in this single incident (article based on the vendor's case study). Integrated steel plant in eastern India (anonymous), camera-based detection with automatic PLC stop by KGT Solutions, live since February 2026: over 3+ months, 57+ foreign objects detected automatically (the page body says 57+, while its header now says 900+ with no new period, so we use the 57+ with the 3-month window), zero unplanned downtime events and no belt damage cost since February 2026, 38% less conveyor downtime and 18% higher productivity (none of these is defined: what counts as downtime, or productivity of what, is not stated), payback within 90 days (vendor case study, which says the results were verified by the client). JFE Steel (Japan) installed AI image-recognition monitoring of raw-material conveyors, up to two kilometers of belt per yard, at its East Japan Works (Chiba) and West Japan Works (Kurashiki), with plans to roll it out to all its domestic yards; no figures published (company news release, 2024).", "companies": [ { "name": "Cemex", "url": "https://aiagentsformanufacturing.com/companies/#cemex" }, { "name": "JFE Steel", "url": "https://aiagentsformanufacturing.com/companies/#jfe-steel" } ], "tools_deployed": [ { "name": "Flexco Elevate Belt Rip Prevention", "url": "https://aiagentsformanufacturing.com/tools/#flexco-elevate-belt-rip-prevention" }, { "name": "KGT Solutions", "url": "https://aiagentsformanufacturing.com/tools/#kgt-solutions" } ], "tools": [ "Flexco Elevate Belt Rip Prevention (drive-data AI, local processing, no belt sensors): used at Cemex", "KGT Solutions custom computer vision with PLC (Modbus) stop and alerts: used at the Indian steel plant", "JFE Steel's in-house AI image recognition with a video management system: used at JFE", "Other tools to evaluate: belt rip detection loops and belt scanning systems from conveyor suppliers, and general vision platforms", "Any LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below, on your conveyor incident log, to find patterns and draft a protection plan" ], "sources": [ "https://www.worldports.org/ai-powered-conveyor-belt-monitoring-prevents-high-maintenance-costs-at-cemex/", "https://kgtsolutions.com/resources/case-studies/ai-conveyor-belt-monitoring", "https://www.jfe-steel.co.jp/en/release/2024/01/240125.html" ], "date_added": "2026-10-22", "last_updated": "2026-10-26" }, { "title": "AI pre-processes medical device complaints, from intake to reportability, so specialists focus on investigations (Medtronic, a patient-monitoring device maker)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-medical-device-complaint-handling-reportability-medtronic/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-medical-device-complaint-handling-reportability-medtronic.md", "department": "Quality", "department_url": "https://aiagentsformanufacturing.com/departments/quality/", "evidence": "Verified", "summary": "AI in medical device complaint handling: Medtronic halved touch time (158,000 to 79,000 h/yr) and cut lead time from 26 to 16 days.", "specialized": true, "who_this_is_for": "Medical device makers that receive patient or user complaints and must decide, under regional regulations and strict deadlines, whether each one is reportable to the health authorities. Quality, complaint-handling and regulatory affairs teams are the main users.", "role": "Quality (post-market: complaint handling, complaint coding and adverse-event reportability)", "problem": "Every complaint about a medical device must be recorded, assessed, coded, checked for whether it is a reportable adverse event (a patient-safety incident that must be reported to the health authorities) within strict deadlines, investigated when needed and closed, with a complete audit trail. Most of this is manual reading and typing. At Medtronic the complaint process was almost completely manual from intake to closure, and only 50% of incoming complaints contained complete information at intake, which triggers time-consuming 'good faith effort' follow-ups. A global patient-monitoring device maker using an accurate but slow manual decision tree could not keep up when an acquisition and a regulatory audit added a backlog of complaints and service records, and hiring and training more evaluators was too slow. A European MedTech maker had more than 3,000 reopened complaint files waiting.", "how_ai_helps": "Intake forms ask product-specific questions instead of free text, so validated rules can make the simple, deterministic decisions (the same input always gives the same answer). AI then helps with the judgment-heavy reading: at Medtronic an LLM (large language model, the kind of AI behind chatbots) drafts the complaint summary from the intake data and follow-ups, and recommends whether a good-faith effort to collect missing data is needed; at the patient-monitoring maker a validated model mirrors the existing decision tree to separate potential complaints from other records, assess patient impact and evaluate reportability by severity and regional regulation, and sends records it cannot process to people. Specialists review every AI output and make the reporting decisions.", "result": "Medtronic, pilot with the FDA and the Medical Device Innovation Consortium (MDIC) started in September 2023, using rule-based automation and generative AI: complete complaint information at intake from 50% to 75%; human touch time for the automated tasks (codes, reporting decisions, event narrative) from 158,000 to 79,000 hours a year, a 50% reduction; process lead time for the steps involved from 26 to 16 days (39% better); accuracy of AI complaint summaries against Medtronic's procedural standard from 38% to 77%; the LLM generated false information 1% of the time and up to 23% of summaries needed edits by a specialist, with summary writing time down 40% on average; every LLM output is reviewed by a trained specialist (MDIC case study, 2026). Global patient-monitoring device maker (anonymous), validated AI model by USDM integrated with TrackWise: backlog of complaints and service records eliminated, 'no missed reporting' goal for adverse events upheld, and manual processing resources reduced by 75% (vendor case study). Europe-based MedTech maker (anonymous), Tata Elxsi automation of complaint coding combined with process changes: complaint backlog down 70% and productivity up 40% (vendor case study).", "companies": [ { "name": "Medtronic", "url": "https://aiagentsformanufacturing.com/companies/#medtronic" } ], "tools_deployed": [ { "name": "Tata Elxsi", "url": "https://aiagentsformanufacturing.com/tools/#tata-elxsi" }, { "name": "USDM", "url": "https://aiagentsformanufacturing.com/tools/#usdm" } ], "tools": [ "Medtronic in-house rule-based automation and an LLM for complaint summaries, built with MDIC and FDA input", "USDM validated AI complaint evaluation model integrated with TrackWise: used at the patient-monitoring maker", "Tata Elxsi complaint coding automation and process redesign: used at the European MedTech maker", "Other tools to evaluate: complaint-handling modules with AI in QMS suites (for example Veeva, MasterControl, Sparta TrackWise Digital)", "Any approved private LLM with the prompt below, on de-identified complaint text, to draft a structured summary and a checklist for the specialist" ], "sources": [ "https://mdic.org/resources/case-study-advanced-complaint-handling-automation-at-medtronic/", "https://mdic.org/wp-content/uploads/2026/02/MDIC_AMCH_Medtronic_Complaint_Handling_Case_Study-vf.pdf", "https://www.usdm.com/resources/case-studies/complaint-processing-enhanced-by-ai-in-medical-device-manufacturing", "https://www.tataelxsi.com/insights/case-study/transforming-complaint-management-for-enhanced-medical-device-performance" ], "date_added": "2026-10-22", "last_updated": "2026-10-26" }, { "title": "AI adjusts the cement kiln in real time to burn less fuel and keep clinker quality steady (Heidelberg Materials, SECIL Sibline, Holcim)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-cement-kiln-process-control-fuel-clinker-heidelberg-materials/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-cement-kiln-process-control-fuel-clinker-heidelberg-materials.md", "department": "Production", "department_url": "https://aiagentsformanufacturing.com/departments/production/", "evidence": "Verified", "summary": "AI kiln control at Heidelberg Materials Mokra: fuel cost index -4.1%, heat consumption -2.2%, C3S variance -33%. Steps and a prompt.", "specialized": true, "who_this_is_for": "Cement plants that already run their kiln with advanced process control (APC), a process-data historian (the database that stores sensor readings over time) and regular laboratory results. It is a specialist case: most manufacturers do not have this setup.", "role": "Production (continuous process control: kiln and pyroprocess optimization)", "problem": "A cement kiln is a hard process to run steadily. Raw meal chemistry varies, and plants burn more and more alternative fuels such as refuse-derived fuel (fuel made from sorted waste), whose calorific value (energy content) changes from one load to the next. Operators and advanced process control (APC) keep correcting fuel, air and feed, but each shift reacts after the fact: the kiln is run with safety margins that cost fuel, clinker (the intermediate product that is ground into cement) quality swings, and process engineers spend their time on manual tuning and shift-to-shift troubleshooting. Heidelberg Materials' Mokra plant (Czechia) already ran a thermal substitution rate (the share of kiln fuel energy that comes from alternative fuels) of up to 86%, mainly with refuse-derived fuels, on top of ABB Ability Expert Optimizer, and wanted to extend that APC with AI.", "how_ai_helps": "Machine-learning models trained on the plant's live process, laboratory and chemical data act as soft sensors (software estimates of values that are slow or hard to measure directly): they continuously predict key values (for example clinker quality indicators such as C3S, the main strength-giving mineral in clinker, or free lime, unreacted lime left in it) and recommend or set the next control moves for fuel mix, feed, kiln speed and air. At Mokra the AI sits on top of the existing ABB APC and gives it dynamic targets in closed loop (its moves are applied directly, without waiting for an operator); at Sibline the optimizer runs the kiln in continuous autopilot; at Holcim's Siggenthal plant an AI digital twin (a model that mirrors the kiln's live behavior) balances fluctuating alternative-fuel calorific values and stabilizes the flame in autopilot. Operators keep the constraints and can take back control at any time. Note that the set-ups differ: Mokra's AI sits on top of an existing APC, while the Gigaton system announced in May 2026 (the new name of Carbon Re) also controls the actuators directly.", "result": "Heidelberg Materials, Mokra plant (Czechia), Carbon Re AI (now named Gigaton) integrated with ABB Ability Expert Optimizer: in the first month of continuous operation, fuel cost index down 4.1%, fuel-derived CO2 down 4.5 kg per tonne of clinker (2%), and C3S variance down 33%; in a one-month on, one-month off test controlling for fuel mix and raw material quality, specific heat consumption down 2.2%; in a Carbon Re-sponsored CemNet article (2 June 2025) the plant director says the live connection and dashboard were delivered in 8 weeks (that is the connection and dashboard, not a full roll-out or the time to the results above) (University of Cambridge Enterprise news and the sponsored article). Sibline Cement (SECIL Group, Lebanon), ES Processing Kiln Process Optimizer: kiln throughput up 2%, thermal energy consumption down about 1.8%, specific electrical energy down 2.8%, free lime standard deviation down 48.5% and C3S variation down 5% (the vendor page gives no period or comparison base), with the kiln in continuous autonomous operation (vendor news, featured in World Cement, October 2025). Holcim (Schweiz), Siggenthal plant, Aix Process KILN AIxperT: deployed to balance alternative-fuel calorific values, predict clinker quality in real time and stabilize the flame in autopilot at high substitution rates; no figures published, and no measured result or duration of operation in the source (CemNet news, 6 March 2025).", "companies": [ { "name": "Heidelberg Materials", "url": "https://aiagentsformanufacturing.com/companies/#heidelberg-materials" }, { "name": "Holcim", "url": "https://aiagentsformanufacturing.com/companies/#holcim" }, { "name": "SECIL", "url": "https://aiagentsformanufacturing.com/companies/#secil" } ], "tools_deployed": [ { "name": "Aix Process KILN AIxperT", "url": "https://aiagentsformanufacturing.com/tools/#aix-process-kiln-aixpert" }, { "name": "Carbon Re (now Gigaton)", "url": "https://aiagentsformanufacturing.com/tools/#carbon-re-now-gigaton" }, { "name": "ES Processing Kiln Process Optimizer", "url": "https://aiagentsformanufacturing.com/tools/#es-processing-kiln-process-optimizer" } ], "tools": [ "Carbon Re, now Gigaton (AI process control on top of ABB Ability Expert Optimizer or FLSmidth PXP): used at Heidelberg Materials Mokra", "ES Processing Kiln Process Optimizer (AI and APC): used at Sibline Cement, SECIL Group", "Aix Process KILN AIxperT (AI digital twin of the kiln): used at Holcim Siggenthal", "Other tools to evaluate: the AI or APC modules of your control-system vendor (ABB, FLSmidth, Siemens), and process analytics tools that read your historian and lab data", "For a first offline study: any LLM with a data-analysis or code tool (ChatGPT, Claude, Copilot, Gemini) on an export of past shifts, with the prompt below" ], "sources": [ "https://www.enterprise.cam.ac.uk/news/heidelberg-materials-adopts-carbon-res-ai-and-abb-ability-for-improved-kiln-operations/", "https://www.cemnet.com/Partner/story/179280/decarbonise-profitably-ai-process-control-that-cuts-costs-and-carbon.html", "https://www.es-processing.com/secil-group-sibline-cement-ai-apc-kiln-process-optimizer/", "https://www.cemnet.com/News/story/178754/holcim-optimises-cement-production-with-kiln-aixpert.html", "https://gigaton.co/article/introducing-Gigaton" ], "date_added": "2026-10-23", "last_updated": "2026-10-26" }, { "title": "AI traces each engineering change to the parts, purchase orders, routings and work instructions it affects, before it reaches the floor (custom-assembly manufacturer, Cytovale)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-engineering-change-impact-analysis-bom-routings-cytovale/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-engineering-change-impact-analysis-bom-routings-cytovale.md", "department": "Engineering", "department_url": "https://aiagentsformanufacturing.com/departments/engineering/", "evidence": "Pilot", "summary": "AI engineering change impact analysis: 40% fewer production disruptions from late design changes; Cytovale pilot 70% faster impact assessments.", "specialized": false, "who_this_is_for": null, "role": "Engineering (engineering change management: ECR/ECO impact analysis)", "problem": "An engineering change rarely stays in engineering. A last-minute part change can delay a purchase order, a moved sub-assembly can invalidate the work instructions, and if nobody connects the change to the open orders, the shop floor discovers it after the job has started, with rework and scrap. Finding every affected item means digging through the BOM, where-used lists, routings, open purchase and production orders, requirements and test documents, often in different systems. A mid-sized manufacturer of custom, complex assemblies was changing designs days before builds and had no quick way to see the downstream impact. At Cytovale, a medical diagnostics device maker, working out which requirements, risks and tests a proposed change affects took up to weeks, depending on the change, and required several subject-matter experts.", "how_ai_helps": "The AI connects the change to the product data around it. At the custom-assembly manufacturer, a dependency graph built from change logs, BOMs, procurement lead times and vendor data, routings, work instructions and production orders is combined with text analysis of the design notes and business rules: every new change is scanned automatically, and buyers, planners and production get an alert within minutes for the high-risk ones. At Cytovale, an AI agent reads hardware descriptions, software specifications, requirements and risk documents linked in one traceability tool, and drafts the impact assessment: affected requirements, a proposed test strategy and the rationale, which engineers then review. Engineers and the change board still approve every change.", "result": "Mid-sized manufacturer of custom, complex assemblies (anonymous), system built by the integrator Captivix and connected to the ERP and PLM: production disruptions from late design changes down 40%, procurement delays linked to overlooked changes down 33%, and several rework or scrap events prevented entirely; the model also flagged past incidents that had slipped through when tested on history (integrator case study). Cytovale (AI-enabled Class II diagnostic system, requirements in Polarion), pilot of Ketryx AI agents for change impact assessments: 70% less time spent and 90% accuracy compared with an engineer-drafted plan, with human review of every AI output (vendor case study of a pilot). Related automation without a stated AI component: Galvanize Therapeutics built an app on Arena PLM's API that extracts every item number in a technical file and checks its status and revision before release, and reports that reviewing a 100-page document went from 13 hours to 67 seconds (PTC case study).", "companies": [ { "name": "Cytovale", "url": "https://aiagentsformanufacturing.com/companies/#cytovale" }, { "name": "Galvanize Therapeutics", "url": "https://aiagentsformanufacturing.com/companies/#galvanize-therapeutics" } ], "tools_deployed": [ { "name": "Arena PLM", "url": "https://aiagentsformanufacturing.com/tools/#arena-plm" }, { "name": "Captivix", "url": "https://aiagentsformanufacturing.com/tools/#captivix" } ], "tools": [ "Captivix custom change-impact model (dependency graph, NLP on design notes, business rules) connected to ERP and PLM: used at the custom-assembly manufacturer", "Ketryx (traceability and AI agents for regulated product development, with Polarion): piloted at Cytovale", "Arena PLM API with an in-house item-check app: used at Galvanize Therapeutics", "Other tools to evaluate: AI assistants in your PLM, for example Siemens Teamcenter Copilot for BOM (impact analysis and part replacement with human confirmation), and the change modules of PTC Windchill, Aras or Arena", "Any LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below, on one change request and exported where-used and open-order lists, for a first test" ], "sources": [ "https://captivix.com/industries/manufacturing/use-cases/engineering-change-impact", "https://www.ketryx.com/case-studies/how-cytovale-transformed-change-impact-assessment-ketryx-ai", "https://www.ptc.com/en/case-studies/galvanize-harnesses-cloud-plm-and-ai", "https://blogs.sw.siemens.com/teamcenter/ai-bom-agent-plm-engineering-change/" ], "date_added": "2026-10-23", "last_updated": "2026-10-23" }, { "title": "AI screens and schedules hourly plant applicants around the clock, so recruiters stop living on the phone (Twin City Fan, General Motors, Nestlé)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-hourly-hiring-screening-interview-scheduling-twin-city-fan-gm/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-hourly-hiring-screening-interview-scheduling-twin-city-fan-gm.md", "department": "Office & management", "department_url": "https://aiagentsformanufacturing.com/departments/office/", "evidence": "Verified", "summary": "AI screening and interview scheduling for plant jobs: Twin City Fan cut time to hire from 60 to 20 days; GM from 5-7 days to 29 min to schedule.", "specialized": false, "who_this_is_for": null, "role": "Office & management (HR and recruiting: screening and interview scheduling for plant roles)", "problem": "Plants need a steady flow of operators, assemblers, welders and machine operators, and many applicants apply from a phone. Recruiters spend their days calling people back, asking the same screening questions and juggling supervisors' calendars, and good candidates take another job while they wait. Twin City Fan, a fan maker with about 1,000 employees and plants in Elkton, Aberdeen and Mitchell, South Dakota, received about 200 applications a month for hourly roles and phone-screened 90% of them: about 250 hours of calls a month for one full-time and one part-time recruiter, on top of $20,000 a month in job advertising, and 60% of hourly hires left within a year. At General Motors, which receives 1 to 2 million applications a year, more than 55 contracted recruiting coordinators were scheduling interviews and it still took 5 to 7 days to schedule one.", "how_ai_helps": "A conversational AI assistant talks to every applicant by phone, text, chat or WhatsApp, at any hour. At Twin City Fan an AI recruiter runs the first phone screen with the same full set of questions for every applicant (shift availability, physical requirements, certifications, distance and pay expectations), and the recruiting team reads the results instead of making the first calls. At GM the assistant books interviews directly in the hiring managers' calendars, integrated with Workday; the 2022 Josh Bersin report describes scheduling for candidates who had already been screened, so GM is a scheduling example, not a screening one. At Nestlé the Paradox page says the assistant automates screening and interview scheduling, integrated with SAP SuccessFactors. Recruiters and hiring managers still interview, decide and make every offer.", "result": "Twin City Fan (fan manufacturer, six facilities), Blucor AI recruiter, started in October 2024 with two openings and extended to the whole hourly funnel: recruiter time on first phone calls from 200 to 250 hours a month to zero (the time to read the AI's results and handle exceptions is not counted); time to hire from 60 to 20 days; screening cost from about $150,000 a year (1.5 recruiters) to $60,000 a year (the subscription); 978 screenings (platform data as of 23 July 2026) across 68 openings and 17 role types. The company also reports company-wide turnover from about 50% to 38% and the number of people hired each year to hold headcount from 427 (2023) to 253 (2025), which the vendor links to more consistent screening (vendor case study, with the before figures, turnover and hiring reported by the company's HR director). General Motors, Paradox conversational assistant 'EV-e' with Workday: time to schedule interviews from 5 to 7 days to 29 minutes, nearly 50,000 interviews scheduled automatically a year, and more than $2 million saved in the first year in direct recruiting costs (2022 report; these gains are about scheduling, not screening), while recruiting coordinators moved to more strategic work (Josh Bersin Company case study and vendor case study). Nestlé, Paradox assistant 'Olivia' with SAP SuccessFactors: interviews scheduled up 600% year over year and about 8,000 hours of recruiting work automated a year, with no starting number, scope or period given for the 600% (vendor case study).", "companies": [ { "name": "General Motors", "url": "https://aiagentsformanufacturing.com/companies/#general-motors" }, { "name": "Nestlé", "url": "https://aiagentsformanufacturing.com/companies/#nestle" }, { "name": "Twin City Fan", "url": "https://aiagentsformanufacturing.com/companies/#twin-city-fan" } ], "tools_deployed": [ { "name": "Blucor", "url": "https://aiagentsformanufacturing.com/tools/#blucor" }, { "name": "Paradox", "url": "https://aiagentsformanufacturing.com/tools/#paradox" } ], "tools": [ "Blucor (AI phone screening recruiter for hourly roles): used at Twin City Fan", "Paradox conversational recruiting assistant (screening and interview scheduling, with Workday or SAP SuccessFactors): used at General Motors and Nestlé, and by Medtronic for high-volume manufacturing roles", "Other tools to evaluate: the scheduling and chat-to-apply features of your applicant tracking system, and manufacturing hiring platforms such as FactoryFix", "Any LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below, to write a consistent screening script and scorecard from your job descriptions, with no candidate data" ], "sources": [ "https://www.blucor.ai/case-studies/twin-city-fan", "https://info.paradox.ai/hubfs/Content%20Marketing%20Assets/GM%20+%20Bersin%20-%20Case%20Study%202023.pdf", "https://www.paradox.ai/case-studies/general-motors", "https://www.paradox.ai/case-studies/nestle" ], "date_added": "2026-10-23", "last_updated": "2026-10-26" }, { "title": "AI checks every spot weld from the welding controller's data, instead of sampling one car a day with ultrasound (Audi, Volkswagen)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-spot-weld-quality-every-weld-from-controller-data-audi/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-spot-weld-quality-every-weld-from-controller-data-audi.md", "department": "Quality", "department_url": "https://aiagentsformanufacturing.com/departments/quality/", "evidence": "Verified", "summary": "Audi's AI checks about 1.5 million spot welds on 300 cars each shift from welding-controller data, instead of ultrasound sampling. Steps and a prompt.", "specialized": true, "who_this_is_for": "Automotive body shops that use instrumented resistance spot welding, where controller data from each weld can be matched with past inspection results (ultrasound or teardown).", "role": "Quality (in-process quality control: resistance spot welding in the body shop)", "problem": "A car body holds thousands of resistance spot welds (metal sheets joined by pressing electrodes together and passing a strong current), and the usual way to check them is destructive or manual sampling. At Audi's Neckarsulm plant, which assembles up to about 1,000 vehicles a day with about 5,000 welds each (5,300 in the body of an Audi A6), more than 5 million welds are made every day. The standard method was to pull one car off the line each day and have 18 engineers test its welds with ultrasound probes, recording the results on clipboards, which left no in-line check of the welds on the other 999 cars. 'Our big goal for this solution is to make it possible for us to inspect 100% of our welds with a very high degree of accuracy,' said Mathias Mayer of Audi.", "how_ai_helps": "A machine-learning model reads the data that the welding-gun controllers already produce for every weld: the electric voltage and current curves, the weld configuration, the metal types and the condition of the electrodes, about 200 variables per weld, each weld lasting 400 to 1,000 milliseconds. The model was trained by comparing its predictions with Audi's real ultrasound inspection results. It flags anomalies in real time, and a lineside screen tells the inspector which welds to check, so people test the suspicious welds instead of a random sample. The proof of concept ran with Intel's Industrial Edge Insights software at the edge (on a computer beside the line rather than in a remote data center); the production version runs in the Volkswagen Industrial Cloud on AWS. To make the method audit- and certification-proof, Audi developed it with the German Association for Quality (DGQ) and the Fraunhofer IAO and IPA institutes.", "result": "Audi, Neckarsulm plant (Germany), in-house WPS-Analytics project: the AI analyses around 1.5 million spot welds on 300 vehicles each shift, and staff now focus on possible anomalies instead of random ultrasound checks of about 5,000 welds per sampled vehicle. After a pilot that had run for two years by early 2022, Audi began the roll-out in June 2023, with the infrastructure being installed at Audi Brussels, Volkswagen Emden and Audi Ingolstadt and the model retrained for each plant's weld settings; Audi is also working on using the weld data for predictive maintenance (Audi press release, June 2023; Automotive Manufacturing Solutions, January 2022). Volkswagen's Emden plant runs its own spot-welding analytics on welding-robot data, which analyses but does not control the process: changes are displayed and then checked by people (Automotive Manufacturing Solutions). Neither company has published a defect-rate or cost figure for the welds.", "companies": [ { "name": "Audi", "url": "https://aiagentsformanufacturing.com/companies/#audi" }, { "name": "Volkswagen", "url": "https://aiagentsformanufacturing.com/companies/#volkswagen" } ], "tools_deployed": [ { "name": "Intel Industrial Edge Insights", "url": "https://aiagentsformanufacturing.com/tools/#intel-industrial-edge-insights" } ], "tools": [ "Audi WPS-Analytics (in-house machine learning on welding-controller data): proof of concept with Intel Industrial Edge Insights, production on the Volkswagen Industrial Cloud (AWS, with Siemens for machine integration); used at Audi Neckarsulm; Audi announced in June 2023 a roll-out to Audi Brussels, Volkswagen Emden and Audi Ingolstadt (we have not checked the current status)", "Other tools to evaluate: the weld-monitoring and data-export options of your welding-controller supplier, and anomaly-detection platforms that read controller or PLC data", "For a first offline study: any LLM with a data-analysis or code tool (ChatGPT, Claude, Copilot, Gemini) on an export of weld data matched with past ultrasound or teardown results, with the prompt below" ], "sources": [ "https://www.audi.com/en/press-releases/audi-begins-roll-out-of-artificial-intelligence-for-quality-control-of-spot-welds-15443", "https://www.intel.com/content/www/us/en/customer-spotlight/stories/audi-automated-factory.html", "https://www.automotivemanufacturingsolutions.com/oems/the-application-of-ai/534291" ], "date_added": "2026-10-26", "last_updated": "2026-10-26" }, { "title": "AI video analytics times every cycle on manual assembly lines, so engineers find and fix the real bottleneck (HELLA, a Tier 1 automotive supplier)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-video-analytics-manual-assembly-cycle-times-hella/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-video-analytics-manual-assembly-cycle-times-hella.md", "department": "Production", "department_url": "https://aiagentsformanufacturing.com/departments/production/", "evidence": "Verified", "summary": "AI video analytics on manual assembly: HELLA recaptured 10% lost efficiency in 3-4 months; payback under 6 months at Dhankot. Steps and a prompt.", "specialized": false, "who_this_is_for": null, "role": "Production (manual assembly lines: cycle time, line balancing and standardized work)", "problem": "Much of assembly is still done by people, and the tools to understand manual work are time and motion studies: hard to scale, based on a few observed cycles, and biased by the observer. Engineers often work on the station they believe is the bottleneck. At a Tier 1 automotive supplier's plant in Guadalajara, Mexico, shift leaders each monitored two to four lines and, however diligent, missed deviations from standardized work. At HELLA, a German Tier 1 supplier of lighting and electronics, one operation required operators to turn their body 90 degrees to grab a bolt from a bin with the left hand while working with the right, and doing it the wrong way caused as much as a 10% productivity slowdown from one operator to another.", "how_ai_helps": "A camera above each station streams video; action-recognition neural networks turn the video into data: the sequence and duration of each step at each station, for every cycle. The system compares this with the standard work, shows bottlenecks on dashboards and can alert supervisors, with the video clip, when a step is missed or takes too long. Engineers then watch the clips behind the numbers to understand why, and change the station, the material feed or the training. At HELLA a small appliance on site pre-processes the video and the analysis runs in the cloud. People decide every change; the data is meant for process improvement.", "result": "HELLA, sensor assembly line, Dhankot plant (India), Drishti: the cycle-time data revealed slowdowns at stations that were not the focus of improvement, the video showed that the station layout forced a twisting reach that tired operators, and the team reconfigured the station and improved line balancing; the investment paid for itself in less than six months. HELLA's head of operational excellence says the company had already reduced cycle times at its Guanajuato, Mexico plant in 2020 (ASSEMBLY magazine, December 2022). On the HELLA line with the bolt-reach problem, the team recaptured all of the 10% lost efficiency in three to four months (Embedded Computing Design, June 2022, from Drishti). A Tier 1 automotive supplier in Guadalajara, Mexico, not named: by acting on standardized-work deviations, it lowered the defect rate on those lines by 30% in 12 weeks and saved an average of $10,000 per defect; a kaizen using more than 1,000 automatically timed cycles found ways to improve efficiency by 11% and reduce scrap by 15%, and took 50% less time than a traditional event (article by Drishti's founder in SME Media, May 2021).", "companies": [ { "name": "HELLA", "url": "https://aiagentsformanufacturing.com/companies/#hella" } ], "tools_deployed": [ { "name": "Drishti", "url": "https://aiagentsformanufacturing.com/tools/#drishti" } ], "tools": [ "Drishti (action-recognition video analytics for manual assembly, with TRACE video and FLOW analytics): used at HELLA Dhankot and the Tier 1 supplier in Guadalajara", "Other tools to evaluate: other video analytics tools for manual assembly lines such as Retrocausal, Invisible AI and PowerArena, and the cycle-time features of your MES", "For a first offline study: any LLM with a data-analysis or code tool (ChatGPT, Claude, Copilot, Gemini) on an export of station cycle times, with the prompt below" ], "sources": [ "https://www.assemblymag.com/articles/97518-video-analytics-help-auto-parts-assembler-improve-cycle-time", "https://embeddedcomputing.com/technology/ai-machine-learning/ai-logic-devices-worload-acceleration/success-story-on-hellas-automotive-assembly-lines-ai-is-helping-humans-too", "https://www.sme.org/technologies/articles/2021/may/using-ai-to-bring-out-the-best-in-people-in-mexico/" ], "date_added": "2026-10-26", "last_updated": "2026-10-26" }, { "title": "AI watches news and events around the suppliers and sub-suppliers you have mapped, so buyers act before a disruption stops the plant (Renault Group, Kärcher, Woodward, Nissan)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-supplier-risk-monitoring-disruption-alerts-renault-karcher/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-supplier-risk-monitoring-disruption-alerts-renault-karcher.md", "department": "Procurement", "department_url": "https://aiagentsformanufacturing.com/departments/procurement/", "evidence": "Verified", "summary": "AI supplier risk monitoring: Renault cut major-crisis impact 30-40%; Woodward cut weather response from 60+ hours to 2. Steps and a prompt.", "specialized": false, "who_this_is_for": null, "role": "Procurement (supplier risk management: disruption monitoring for tier-1 and sub-tier suppliers)", "problem": "A fire at a sub-supplier, a flood, a strike, a cyberattack or an insolvency often reaches the plant as a missing delivery, days after it happened, because nobody was watching thousands of supplier sites, let alone the suppliers behind them. Renault Group manages a supply base of 40,000 tier-1 suppliers for 25 production sites running just-in-time. Kärcher's supplier risk management was fragmented, mostly manual and focused locally rather than across the whole tier-N supply chain, while it also had to comply with Germany's Supply Chain Due Diligence Act (LkSG). At Woodward, an aerospace and industrial manufacturer, responding to an extreme weather disruption took more than 60 hours.", "how_ai_helps": "A supplier risk platform reads news, public data and other sources in many languages, groups the signals into events, and matches each event to the geolocated sites of your suppliers and, where mapped, their own suppliers and materials. It then sends an alert to the right team with the suppliers and parts that may be affected; the newer versions also estimate the business impact against your own parts, stock and revenue. Buyers and risk managers confirm the exposure with the supplier and decide the response: safety stock, alternative source, allocation or customer communication.", "result": "Renault Group, Prewave in a control tower shared by engineering, purchasing, supply chain, manufacturing and public affairs: more than 50 major crises managed in 12 months, 30 to 40% lower business impact from major crises, 70% of incidents avoided with no operational impact on customers and no additional costs, and 12,000 vehicles a day safeguarded across 25 just-in-time factories (vendor case study). Kärcher, Prewave: in Prewave's case study (undated PDF), supplier risk effort is 40 times lower when comparing automation with the earlier manual approach (AI-based monitoring against questionnaires), and Kärcher hears about risk events on average 3 days earlier; the 40x is a comparison of two ways of working, not a measured cut in total risk-management workload, and the 3 days is earlier information, not a measured faster reaction by buyers. The project covers more than 83,000 mapped tier-N suppliers and more than 300 commodities, with tier-N mapping focused on critical products, product groups and suppliers (vendor case study PDF). Woodward, Resilinc: $950,000 in cost savings and $245,000 in estimated cost avoidance within six months, and response time to extreme weather disruptions down from more than 60 hours to 2 hours (vendor case study). Nissan Americas, Everstream Analytics: real-time alerts on tier-1 and sub-tier risks, compliance and cyber threats helped avoid over seven figures in potential disruption costs in the first six months (vendor page for a Gartner symposium session with Nissan).", "companies": [ { "name": "Kärcher", "url": "https://aiagentsformanufacturing.com/companies/#karcher" }, { "name": "Nissan", "url": "https://aiagentsformanufacturing.com/companies/#nissan" }, { "name": "Renault Group", "url": "https://aiagentsformanufacturing.com/companies/#renault-group" }, { "name": "Woodward", "url": "https://aiagentsformanufacturing.com/companies/#woodward" } ], "tools_deployed": [ { "name": "Everstream Analytics", "url": "https://aiagentsformanufacturing.com/tools/#everstream-analytics" }, { "name": "Prewave", "url": "https://aiagentsformanufacturing.com/tools/#prewave" }, { "name": "Resilinc", "url": "https://aiagentsformanufacturing.com/tools/#resilinc" } ], "tools": [ "Prewave (AI supplier risk monitoring, tier-N mapping and impact estimates): used at Renault Group and Kärcher", "Resilinc (AI supply chain risk monitoring and multi-tier mapping): used at Woodward", "Everstream Analytics (supplier risk alerts, including cyber and compliance): used at Nissan Americas", "Other tools to evaluate: Interos, and the supplier risk modules of your procurement suite", "Any LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below, to triage one news item against your supplier list, with no confidential data in a public chatbot" ], "sources": [ "https://www.prewave.com/resources/case-studies", "https://www.prewave.com/cases/how-renault-group-reduced-major-crisis-impact-by-up-to-40-with-prewave", "https://cdn.featuredcustomers.com/CustomerCaseStudy.document/prewave_karcher2_591074.pdf", "https://resilinc.ai/learning-center/case-studies/woodward-saves-950k-in-6-months-with-resilincs-supply-chain-risk-management-software/", "https://www.everstream.ai/events/nissan-journey-resilience/" ], "date_added": "2026-10-26", "last_updated": "2026-10-26" }, { "title": "Cameras at the dock door read every pallet and stop the wrong one before it is loaded (Yazaki, Danfoss, a flooring manufacturer)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-dock-door-pallet-verification-shipping-errors-yazaki-danfoss/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-dock-door-pallet-verification-shipping-errors-yazaki-danfoss.md", "department": "Logistics / Supply Chain", "department_url": "https://aiagentsformanufacturing.com/departments/logistics/", "evidence": "Verified", "summary": "Dock-door cameras verify every pallet: Danfoss went from about 600 loading errors a year to none; Yazaki >99.5% accuracy on 170,001 pallets.", "specialized": false, "who_this_is_for": null, "role": "Logistics (shipping and receiving: pallet verification at the loading dock)", "problem": "At the loading dock, pallets are checked with handheld scanners and paperwork, under time pressure, and many boxes look alike. A pallet on the wrong truck means a return trip, a late delivery and sometimes air freight, and when a customer reports a shortage or damage, there is no proof of what was shipped. Danfoss estimated that about 600 pallets or cartons a year were loaded on the wrong vehicle at its main warehouse in Rødekro, Denmark; in one case a pallet destined for Italy went to Iceland. At Yazaki, a wiring-harness supplier to car makers, inbound and outbound pallets were scanned by hand, which was slow, error-prone and delayed inventory data in a just-in-time supply chain.", "how_ai_helps": "Cameras mounted at each dock door capture every pallet as the forklift passes. Computer vision finds and reads the labels (part numbers, supplier codes, quantities, serial numbers, including AIAG standard labels), can count visible cases and spot damage, and compares the result with the order or shipment in the ERP or WMS. A wrong, unexpected or early pallet triggers a stop signal or an exception to the shift supervisor, and an image of every pallet is stored as proof of what was loaded. People handle the exceptions; the system does the routine scanning.", "result": "Yazaki (three North American sites: a manufacturing plant, a cross-dock and a distribution warehouse), Kargo computer vision towers integrated with SAP: over 99% data accuracy in the first two weeks and over 99.5% on more than 170,001 pallets scanned; 4,436 exceptions flagged to operations teams; inventory recognised at each site on average 8 hours faster than with manual processes; images used to answer customers' receiving discrepancy reports (vendor customer spotlight, January 2024). Danfoss, main warehouse (1.5 million order lines and 80,000 deliveries a year, 14 loading bays), Zetes machine-vision loading verification integrated with the WMS: no loading errors at all in the 10 months after go-live, against about 600 a year before, payback expected in 12 months instead of the planned 2 years, and operators able to load more trucks at the same time (vendor case study with Danfoss's warehouse and distribution consultant). A flooring manufacturer, not named, Accella AI Dock Check with deep learning on two high-resolution cameras: counts cases, reads labels and flags missing, extra or wrong cases on the pallet image in about 6 to 8 seconds per pallet, with inference on site and integration with the PLC, WMS and ERP; no error figures published (camera maker's case study).", "companies": [ { "name": "Danfoss", "url": "https://aiagentsformanufacturing.com/companies/#danfoss" }, { "name": "Yazaki", "url": "https://aiagentsformanufacturing.com/companies/#yazaki" } ], "tools_deployed": [ { "name": "Accella AI Dock Check", "url": "https://aiagentsformanufacturing.com/tools/#accella-ai-dock-check" }, { "name": "Kargo", "url": "https://aiagentsformanufacturing.com/tools/#kargo" }, { "name": "Zetes ZetesMedea", "url": "https://aiagentsformanufacturing.com/tools/#zetes-zetesmedea" } ], "tools": [ "Kargo (computer vision towers at the dock door, integrated with SAP or the WMS): used at Yazaki", "Zetes ZetesMedea loading verification with ImageID machine vision: used at Danfoss", "Accella AI Dock Check (deep learning on LUCID cameras, with Ignition for operator screens): used at the flooring manufacturer", "Other tools to evaluate: the dock and loading verification options of your WMS or label-printing supplier", "Any LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below, to analyse your past shipping complaints and write the requirements for a pilot" ], "sources": [ "https://www.kargo.ai/blog/customer-spotlight-yazaki", "https://www.zetes.com/en/warehouse-solutions/direct-store-delivery-software/from-600-to-zero-shipping-errors-and-roi-in-one", "https://thinklucid.com/case-studies/ai-vision-system-verifies-palletized-shipments/" ], "date_added": "2026-10-26", "last_updated": "2026-10-26" }, { "title": "A camera or a phone photo reads the analog gauges, so inspection rounds are no longer written by hand (Nachi-Fujikoshi, Kobe Steel group, a plastics maker)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-analog-gauge-reading-camera-inspection-rounds-nachi-kobe-steel/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-analog-gauge-reading-camera-inspection-rounds-nachi-kobe-steel.md", "department": "Maintenance", "department_url": "https://aiagentsformanufacturing.com/departments/maintenance/", "evidence": "Verified", "summary": "AI reads analog gauges from a camera or phone photo: Nachi-Fujikoshi checks remote gauges from the office; a plastics maker cut gauge rounds 70%.", "specialized": false, "who_this_is_for": null, "role": "Maintenance (daily equipment checks and inspection rounds on analog gauges and meters)", "problem": "Most plants still have hundreds of analog gauges and small meters (pressure, temperature, current, filter differential pressure) with no data output. Someone walks the plant, reads each needle by eye, writes the value on a paper form and types it into a spreadsheet or system later. It is slow, tiring and error-prone, checks get skipped, and nobody can verify afterwards that the value written was the value shown. At Nachi-Fujikoshi's robot plant in Japan, daily start-of-shift checks covered many kinds of equipment (cranes, compressors, torque wrenches); a check could take 10 minutes and was repeated several times a day, some gauges were a 10-minute round trip from the office and others sat high up where the check itself carried a fall risk. At a plastics manufacturer, operators read analog and digital gauges every hour, wrote them on paper and then keyed them into the IT system; tired operators made reading and copying mistakes and lost forms left gaps in the records for audits.", "how_ai_helps": "An image-recognition model reads the needle or the digits from a photo of the gauge and turns it into a number with a time stamp and the photo attached. There are two ways to take the photo. Fixed: a small battery-powered camera is stuck in front of a gauge and takes a few pictures a day (LiLz Gauge says three shots a day give about three years of battery life), the reading is done in the cloud and shows up on a screen in the office. Mobile: the technician keeps doing the round but takes a photo with a phone or tablet instead of writing the value, and the app reads it and uploads it to the database. Thresholds can then trigger an alert. The existing gauges are not replaced or rewired.", "result": "Nachi-Fujikoshi (machinery maker, consolidated sales of 258 billion yen), robot division plant, LiLz Gauge: cameras were installed first on the gauges that were a 10-minute round trip from the office and on checks at height; they now watch duct temperature (to prevent overheating), refrigerant pressure gauges and current readings, which staff check remotely from the office, detecting equipment problems earlier; the plan is to extend it to all equipment in the robot division; no time or cost figure published (LiLz customer interview, November 2023). Kobe Steel group, hakaru.ai (GMO GlobalSign) connected by API to a Claris FileMaker app, first deployed at Kobelco Research Institute, the group's materials and parts testing company, to record test and inspection readings from analog, digital and panel meters: about 20 users; each value is now stored with its photo, so a third person can check a reading that only the person measuring could confirm before; no more handwritten status notes or misread handwriting; old instruments kept in service. The meter-reading AI was reading 81% of values correctly at one point and was then retrained on the most-used meters, which the user says improved accuracy and efficiency, without a new figure; angles, glare and small decimal points made some instruments hard to read (vendor interview, information as of February 2022). A plastics and synthetic materials manufacturer (not named), Solomon META-aivi on phones and tablets: routine gauge inspections 70% faster than the previous manual process, readings uploaded automatically to its IT system and fewer recording errors (Solomon case study).", "companies": [ { "name": "Kobe Steel", "url": "https://aiagentsformanufacturing.com/companies/#kobe-steel" }, { "name": "Nachi-Fujikoshi", "url": "https://aiagentsformanufacturing.com/companies/#nachi-fujikoshi" } ], "tools_deployed": [ { "name": "hakaru.ai (GMO GlobalSign)", "url": "https://aiagentsformanufacturing.com/tools/#hakaru-ai-gmo-globalsign" }, { "name": "LiLz Gauge", "url": "https://aiagentsformanufacturing.com/tools/#lilz-gauge" }, { "name": "Solomon META-aivi", "url": "https://aiagentsformanufacturing.com/tools/#solomon-meta-aivi" } ], "tools": [ "LiLz Gauge (battery-powered IoT cameras with cloud machine-learning meter reading): used at Nachi-Fujikoshi", "hakaru.ai by GMO GlobalSign (meter-reading AI from photos, used through its API in a Claris FileMaker app): used in the Kobe Steel group", "Solomon META-aivi (photo-based gauge reading on phones and tablets, analog and digital): used at the plastics manufacturer", "Other tools to evaluate: meter-reading apps on industrial camera platforms (for example Kyocera's Meter Reader on Actcast), the gauge-reading options of your CMMS or inspection-rounds app, and a smart transmitter where a gauge is critical enough to justify one", "For a first test: any LLM that reads images (ChatGPT, Claude, Copilot, Gemini) on a few gauge photos, with the prompt below" ], "sources": [ "https://lilz.jp/interviews/8vgn500dic0o", "https://www.hakaru.ai/case/case_kobelco.html", "https://www.solomon-3d.com/case-studies/meta-aivi/automated-gauge-reading-ai-vision-ocr/" ], "date_added": "2026-10-27", "last_updated": "2026-10-27" }, { "title": "Engineers ask questions of their standards, drawings and past project files and get an answer with the source page (Flender, NETZSCH, Victaulic)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-engineering-knowledge-assistant-standards-drawings-past-projects-flender-netzsch/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-engineering-knowledge-assistant-standards-drawings-past-projects-flender-netzsch.md", "department": "Engineering", "department_url": "https://aiagentsformanufacturing.com/departments/engineering/", "evidence": "Verified", "summary": "An AI assistant answers engineers from their own standards and drawings: Flender cut interpretation time from 30 to 5 minutes; NETZSCH saved 5,600 hours.", "specialized": false, "who_this_is_for": null, "role": "Engineering (design, application and project engineering: finding answers in standards, specifications, drawings and past projects)", "problem": "Engineering knowledge is spread over thousands of documents: internal design standards, customer specifications, drawings, test reports, manuals, approvals and past project files, in shared drives, PLM and email. Keyword search finds files, not answers, so engineers spend a long time reading, ask the senior colleague who knows, or redo work that was already done. At Flender, a maker of industrial gearboxes and drives that builds to order, knowledge sat in fragmented repositories and in many formats (documents, drawings, spreadsheets), and interpreting a design standard or drawing took about 30 minutes. At Victaulic, a pipe-joining and fire-protection manufacturer with 5,500 employees and 50 facilities, customer care and application engineering teams searched by hand through manuals, submittals, white papers and approvals to answer technical questions, which slowed answers and made them inconsistent. At NETZSCH Pumpen & Systeme, project, product and development staff needed fast access to project-specific details buried in extensive documentation.", "how_ai_helps": "The company's own documents are indexed into a private assistant. An engineer asks a question in plain language (for example: what flow rate and differential pressure does this customer require, which noise standard applies, which approvals must we pass on to the supplier), and the assistant finds the relevant passages and writes a short answer with links to the source pages, so the engineer can check them. This is called retrieval-augmented generation (RAG): the language model answers from retrieved company documents rather than from general knowledge. At Flender it runs on premises and also reads drawings; at Victaulic it runs on AWS with Claude Sonnet; at NETZSCH it is the MAIA platform. The engineer still reads the source and decides.", "result": "Flender China, sovereign agentic AI engineering assistant built by Atos and deployed on premises: time to interpret design standard documents and engineering drawings cut by 83%, from about 30 minutes to 5 minutes; about 2,000 engineering hours saved a year; knowledge discovery and document retrieval 30 to 50% faster; Atos also reports a sixfold increase in engineering design processing capacity (Atos case study, September 2026). NETZSCH Pumpen & Systeme (positive displacement pumps, about 600 employees in Germany), MAIA: more than 5,600 working hours saved since introduction, 235,500 pages of technical documentation searchable, about 1,400 queries a month; used to review customer specifications, write test procedures from standards and specifications, and pick out the specifications to pass on to suppliers; its head of project management says the team identifies critical details earlier in projects (MAIA case study). Victaulic, conversational assistant built by Nagarro on AWS with Amazon Bedrock and Claude Sonnet: searches dropped from minutes to seconds, 600 to 700 engineering hours saved a year, and response accuracy above 90% (Nagarro success story).", "companies": [ { "name": "Flender", "url": "https://aiagentsformanufacturing.com/companies/#flender" }, { "name": "NETZSCH Pumpen & Systeme", "url": "https://aiagentsformanufacturing.com/companies/#netzsch-pumpen-systeme" }, { "name": "Victaulic", "url": "https://aiagentsformanufacturing.com/companies/#victaulic" } ], "tools_deployed": [ { "name": "Atos sovereign agentic AI", "url": "https://aiagentsformanufacturing.com/tools/#atos-sovereign-agentic-ai" }, { "name": "MAIA", "url": "https://aiagentsformanufacturing.com/tools/#maia" }, { "name": "Nagarro NIA", "url": "https://aiagentsformanufacturing.com/tools/#nagarro-nia" } ], "tools": [ "Atos sovereign agentic AI engineering assistant (RAG and multimodal AI, on premises): used at Flender China", "MAIA (AI knowledge platform for technical documents): used at NETZSCH Pumpen & Systeme", "Nagarro NIA agentic AI platform on AWS (Amazon Bedrock with Claude Sonnet, Amazon Textract, Amazon OpenSearch): used at Victaulic", "Other tools to evaluate: Microsoft 365 Copilot or Copilot Studio over SharePoint, Google Gemini for Workspace, the AI search of your PLM, and enterprise search tools with source citations", "For a first test: any LLM with file upload (ChatGPT, Claude, Copilot, Gemini) on two or three non-confidential documents, with the prompt below" ], "sources": [ "https://www.atos.ai/en/news-insights/case-studies/flender-china-accelerating-engineering-excellence-with-agentic-ai", "https://www.getmaia.ai/en/case-studies/netzsch", "https://www.nagarro.com/en/success-stories/victaulic-transforms-product-support-genai" ], "date_added": "2026-10-27", "last_updated": "2026-10-27" }, { "title": "AI drafts answers to customer questionnaires and RFPs from a library of approved answers and proof documents (Haworth, Ingredion, Cuisine Solutions)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-customer-questionnaires-rfp-sustainability-answers-haworth-ingredion/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-customer-questionnaires-rfp-sustainability-answers-haworth-ingredion.md", "department": "Office & management", "department_url": "https://aiagentsformanufacturing.com/departments/office/", "evidence": "Verified", "summary": "AI drafts customer questionnaire and RFP answers from approved content: Haworth dealers save 35-50% per project; Cuisine Solutions saves 10-15 hours a survey.", "specialized": false, "who_this_is_for": null, "role": "Office (sales support, sustainability and quality: answering customer questionnaires, RFPs and information requests)", "problem": "Customers keep sending questionnaires: sustainability surveys (EcoVadis, CDP, their own supplier forms), RFPs, supplier qualification forms, and detailed technical questions. Many questions repeat from one customer to the next, but the answers and proof documents are scattered, so each questionnaire means emailing the same experts, digging up certificates and copying last year's answers, which may be outdated. At Haworth, an office furniture manufacturer that sells through more than 600 independent dealers, dealers used outdated files and emailed corporate experts for answers, those experts were flooded with repeated questions, and proposals were delayed. At Cuisine Solutions, a sous-vide food producer, several sustainability questionnaires a year had to be answered without a central database or proper collaboration between departments. At Ingredion, data for EcoVadis, which rates both the group and individual plants, was collected largely by hand, with no central place for proof documents.", "how_ai_helps": "The company builds one library of approved answers, data and proof documents (policies, certificates, energy and waste figures, product data sheets). When a questionnaire or RFP arrives, the AI matches each question to the library and drafts an answer with the supporting document, and assigns the questions it cannot answer to the right expert. A conversational assistant lets sales people or dealers ask a question and get an answer with its source. Each new answer that people approve goes back into the library, so the next questionnaire is faster. People review and approve every answer before it is sent.", "result": "Haworth (office furniture and workspace design manufacturer), Responsive: time to first draft of a proposal down from days to hours; dealer partners save 35 to 50% of their time on each project; 16 early-adopter dealers live, with Haworth expecting that number to double by the end of 2026; repetitive questions to corporate experts reduced (Responsive customer story). Ingredion (ingredient manufacturer, more than 12,000 employees), Sunhat: 50% time saved managing sustainability ratings and reporting, more than 125 users and more than 200 projects, used for customer questionnaires, EcoVadis, CDP and SEDEX site questionnaires; Sunhat also lists Ingredion's EcoVadis medals (3 silver, 1 gold, 3 bronze) but does not say how much the tool contributed (Sunhat customer story). Cuisine Solutions (sous-vide food producer, Sterling, Virginia), Sunhat: 10 to 15 hours saved per survey, with 545 sustainability indicators in a central database, starting with the CDP questionnaire (Sunhat customer story).", "companies": [ { "name": "Cuisine Solutions", "url": "https://aiagentsformanufacturing.com/companies/#cuisine-solutions" }, { "name": "Haworth", "url": "https://aiagentsformanufacturing.com/companies/#haworth" }, { "name": "Ingredion", "url": "https://aiagentsformanufacturing.com/companies/#ingredion" } ], "tools_deployed": [ { "name": "Responsive", "url": "https://aiagentsformanufacturing.com/tools/#responsive" }, { "name": "Sunhat", "url": "https://aiagentsformanufacturing.com/tools/#sunhat" } ], "tools": [ "Responsive (AI response management with a content library and the Ask assistant): used at Haworth", "Sunhat (AI platform for sustainability questionnaires, ratings and proof documents): used at Ingredion and Cuisine Solutions", "Other tools to evaluate: Loopio and other RFP response tools, and the questionnaire features of sustainability platforms", "For a small company, a first version: a shared folder of approved answers and proof documents plus any LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below, keeping confidential data in an approved company tool" ], "sources": [ "https://www.responsive.io/customer-stories/haworth", "https://www.getsunhat.com/customer-stories/ingredion-datamanagement", "https://www.getsunhat.com/customer-stories/cuisine-solutions-sustainability-data-efficiency-increase" ], "date_added": "2026-10-27", "last_updated": "2026-10-27" }, { "title": "AI sorts every purchase line into spend categories so buyers see what they really spend with each supplier (Pentair, Barry-Wehmiller)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-spend-classification-spend-analysis-pentair-barry-wehmiller/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-spend-classification-spend-analysis-pentair-barry-wehmiller.md", "department": "Procurement", "department_url": "https://aiagentsformanufacturing.com/departments/procurement/", "evidence": "Verified", "summary": "AI spend classification: Pentair cut spend analysis from weeks to minutes with 90%+ accuracy; Barry-Wehmiller raised classification coverage 40%.", "specialized": false, "who_this_is_for": null, "role": "Procurement (spend analysis: knowing what the company buys, from whom, at what terms)", "problem": "Ask a manufacturer how much it spends on fasteners, packaging or maintenance services, and with which suppliers, and it often takes weeks to answer. Purchase orders and invoices sit in several ERPs, the same supplier appears under five names, item descriptions are free text ('BOLT M8 ZN', 'misc hardware'), and many lines have no category at all. Without a clean view, buyers negotiate site by site and miss volume discounts, better payment terms and duplicate suppliers. At Pentair, a water and fluid management manufacturer with more than 19,000 employees, category managers spent weeks on spend analysis to reconcile gaps between business units. At Barry-Wehmiller, an industrial machinery manufacturer that grew through more than 140 acquisitions, more than a billion dollars of spend had no standard way to categorize purchases across business units, so procurement negotiated in silos.", "how_ai_helps": "Software pulls purchase orders, invoices and supplier data from every ERP into one place. Machine learning then cleans supplier names (grouping 'ACME Inc', 'Acme Corp.' and a subsidiary under one parent) and assigns each purchase line to a category of a common taxonomy, learning from lines already classified and from the corrections people make. Experts review the classification, users can suggest corrections, and approved corrections improve the next run. Buyers then see spend by category, supplier, site and payment terms, and find where to consolidate suppliers or renegotiate.", "result": "Pentair (water and fluid management, Minnesota, more than 19,000 employees), Sievo Spend Analytics: spend analysis cycle time for category managers cut from weeks to minutes (quote from Pentair's VP Global Supply Management); more than 90% accuracy in spend classification; supplier consolidation improved by more than 10% in the year; $15M working capital improvement through payment-term negotiations; deployed globally in two months, with under two hours of training (Sievo case study). Barry-Wehmiller (industrial manufacturing, St. Louis, more than 12,000 employees), Sievo Spend Analytics: 40% relative increase in classification coverage; savings enhanced by 20%; in one case, the new classification showed spend with a key supplier was twice what had been reported, which led to stronger terms and a global agreement (Sievo customer story). Sievo describes its classification as AI that learns from user feedback (approved reclassifications improve the next decisions), backed by its own data managers who review the data, and with an explanation of how each line was classified (Sievo product one-pager).", "companies": [ { "name": "Barry-Wehmiller", "url": "https://aiagentsformanufacturing.com/companies/#barry-wehmiller" }, { "name": "Pentair", "url": "https://aiagentsformanufacturing.com/companies/#pentair" } ], "tools_deployed": [ { "name": "Sievo Spend Analytics", "url": "https://aiagentsformanufacturing.com/tools/#sievo-spend-analytics" } ], "tools": [ "Sievo Spend Analytics (machine-learning spend classification and supplier grouping): used at Pentair and Barry-Wehmiller", "Other tools to evaluate: Coupa, SAP Ariba Spend Analysis, Suplari, SpendHQ, Simfoni, and the spend analytics of your ERP or BI tool", "For a small company, a first version: one export of last year's invoice or PO lines in a spreadsheet plus any LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below, in an approved company tool" ], "sources": [ "https://sievo.com/hubfs/Ebooks%20and%20PDFs/Case%20Studies/Pentair%20Spend%20Analysis%20Case%20Study.pdf?hsLang=en", "https://hub.sievo.com/barry-wehmiller-with-sievo", "https://sievo.com/hubfs/Sievo%20One-pagers/Sievo%20Data%20Cleansing%20%26%20Categorization%20One-pager.pdf?hsLang=en" ], "date_added": "2026-10-28", "last_updated": "2026-10-28" }, { "title": "AI watches CNC spindle load and stops the machine before a worn or broken tool makes scrap (BC Machining, an automotive supplier)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-cnc-tool-wear-breakage-detection-scrap-bc-machining/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-cnc-tool-wear-breakage-detection-scrap-bc-machining.md", "department": "Production", "department_url": "https://aiagentsformanufacturing.com/departments/production/", "evidence": "Verified", "summary": "AI tool monitoring on CNC machines: BC Machining predicts tool failure with 99% accuracy and saves $72,000+ per machine a year by stopping before scrap.", "specialized": false, "who_this_is_for": null, "role": "Production (CNC machining: tool life, scrap and unattended running)", "problem": "On a CNC machine, a drill, tap or end mill wears out or breaks, and the machine keeps cutting. Nobody notices until a bad part is measured or an operator hears it, so a batch of parts is already scrap, the machine may be damaged, and the shift loses time sorting good parts from bad. Shops respond by changing tools early (wasting tool life) or by keeping someone watching the machines, which rules out running unattended. At BC Machining, a maker of aluminium machined components, broken tools were so common that the shop often lost a third of the parts from each shift, operators spent an hour per shift sorting parts to find the scrap, and breakage was only verified by operators looking. At a global automotive supplier machining steering, driveline and suspension parts, operators often realized a tool was broken only after defects had been produced.", "how_ai_helps": "Software reads high-frequency data straight from the machine control (spindle load, torque, spindle speed), with no extra sensors. AI learns what a normal cut looks like for each tool within a few production cycles, then flags a shift in the signal that shows wear or an impending break. It alerts the operator or supervisor and, if set up to, triggers a feed hold so the machine stops before it makes scrap; the operator changes the tool and restarts.", "result": "BC Machining (aluminium machined components for healthcare, defense, transportation and power tools; CNC Swiss lathes), MachineMetrics Tool Monitoring: patterns predicted tool failure with 99% accuracy, and the system stops the Star machines with a feed hold before the tool breaks; the shift-by-shift loss of parts to tool breakage and the hour of sorting per shift were eliminated; 'lights out' production with confidence; savings of more than $72,000 per machine per year (MachineMetrics case study interview with the owner, shop floor manager and manufacturing engineer). Global automotive supplier (not named; high-volume precision machining of steering, driveline and suspension components), MachineMetrics Tool Anomaly Detection, March 2025: AI models learned normal tool behaviour within a few production cycles; alerts go to the operator or supervisor or trigger a feed hold; the company estimates each machine saves thousands of dollars per month in scrap, downtime and maintenance costs (MachineMetrics case study, interview with the principal manufacturing engineer).", "companies": [ { "name": "BC Machining", "url": "https://aiagentsformanufacturing.com/companies/#bc-machining" } ], "tools_deployed": [ { "name": "MachineMetrics", "url": "https://aiagentsformanufacturing.com/tools/#machinemetrics" } ], "tools": [ "MachineMetrics Tool Monitoring and Tool Anomaly Detection (AI on spindle load and other machine-control data, feed hold): used at BC Machining and the automotive supplier", "Other tools to evaluate: the tool monitoring options of your machine or control builder (for example adaptive control and tool breakage detection on Fanuc, Siemens, Heidenhain controls), and add-on systems such as ARTIS (Marposs), Caron Engineering TMAC, Montronix", "To prepare before buying: any LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below, on your scrap and tool-change log, to find which machines and tools to pilot first" ], "sources": [ "https://www.machinemetrics.com/bc-machining", "https://www.machinemetrics.com/blog/tool-anomaly-detection-case-study" ], "date_added": "2026-10-28", "last_updated": "2026-10-28" }, { "title": "AI balloons the part drawing and fills the first article inspection report, instead of retyping every dimension (AdvancedPCB, Lake County Tool Works, B&B Airparts)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-drawing-ballooning-first-article-inspection-report-advancedpcb/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-drawing-ballooning-first-article-inspection-report-advancedpcb.md", "department": "Quality", "department_url": "https://aiagentsformanufacturing.com/departments/quality/", "evidence": "Verified", "summary": "AI ballooning for first article inspection: AdvancedPCB raised FAIR throughput 156% at one site; B&B Airparts cut FAIR time in half. Steps and a prompt.", "specialized": false, "who_this_is_for": null, "role": "Quality (inspection planning: ballooned drawings, first article inspection and PPAP reports)", "problem": "Before a new part ships, many customers want a first article inspection report (FAIR; in aerospace, the AS9102 form) or a PPAP dimensional report. Someone must put a numbered balloon on every dimension, tolerance and note of the drawing (sometimes hundreds), retype each one into a spreadsheet, then type the measured values next to them. It takes hours per drawing, a missed characteristic means a rejected report, and a backlog of reports delays shipments and invoicing. At B&B Airparts, an aerospace machine shop, the quality team had upwards of 350 FAIs a month, and a FAIR could take 2 to 3 hours or 1 to 2 days on some parts. At AdvancedPCB, a circuit board maker, AS9102 reporting varied by site, needed frequent fixes in Excel, and FAIR backlogs directly blocked revenue collection.", "how_ai_helps": "The quality engineer uploads the drawing (PDF or scan). The software finds the dimensions, tolerances, GD&T symbols and notes, places the numbered balloons, and fills the characteristic list of the report with nominal values and limits. Newer tools use AI models trained on manufacturing drawings, so they also read scans and older prints; others use OCR (text recognition) and rules. CMM or gauge results can be imported next to each characteristic. A person reviews every balloon against the drawing, focusing on the ones the tool flags as uncertain, before the report goes to the customer.", "result": "AdvancedPCB (printed circuit board fabrication and assembly for aerospace and defense), GroundControl (AI-powered quality software): pilot in Chandler, Arizona, in August 2024 produced 61 FAIRs in the first month; operators and CMM programmers productive within 30 minutes of training; FAIR throughput up 47% at one facility and 156% at another; more than 200 FAIRs a month across facilities; ROI in under one month (GroundControl case study). Lake County Tool Works North (GroundControl): its CEO is quoted saying the software cut FAI times by 50 to 70%; Mass Precision more than doubled its FAIR output and has had zero rejections since switching (GroundControl buyer's guide, August 2026). B&B Airparts (aerospace machine shop), InspectionXpert ballooning software with automatic capture of tolerances: FAIR time cut in half, and ballooned drawings now used to plan inspection before production (InspectionXpert customer story; it describes automated ballooning and does not say AI).", "companies": [ { "name": "AdvancedPCB", "url": "https://aiagentsformanufacturing.com/companies/#advancedpcb" }, { "name": "B&B Airparts", "url": "https://aiagentsformanufacturing.com/companies/#b-b-airparts" }, { "name": "Lake County Tool Works", "url": "https://aiagentsformanufacturing.com/companies/#lake-county-tool-works" }, { "name": "Mass Precision", "url": "https://aiagentsformanufacturing.com/companies/#mass-precision" } ], "tools_deployed": [ { "name": "GroundControl", "url": "https://aiagentsformanufacturing.com/tools/#groundcontrol" }, { "name": "InspectionXpert (Ideagen Quality Control)", "url": "https://aiagentsformanufacturing.com/tools/#inspectionxpert-ideagen-quality-control" } ], "tools": [ "GroundControl (AI extraction of drawing characteristics, AS9102 FAIR): used at AdvancedPCB, Lake County Tool Works North and Mass Precision", "InspectionXpert (now Ideagen Quality Control; drawing ballooning and inspection reports): used at B&B Airparts", "Other tools to evaluate: 1Factory, Net-Inspect, DISCUS, High QA, and the ballooning features of your CMM or quality software", "For a first test: any LLM that reads images or PDFs (ChatGPT, Claude, Copilot, Gemini) with the prompt below, on one non-confidential drawing, to see what a first pass catches and misses" ], "sources": [ "https://gndctl.com/case-studies/advanced-pcb-as9102-software-case-study", "https://gndctl.com/resources/automated-ballooning-software-fai-buyers-guide-2026/", "https://www.inspectionxpert.com/customers/bbairparts" ], "date_added": "2026-10-28", "last_updated": "2026-10-28" }, { "title": "AI moves fast-selling items closer to the pickers and shortens pick routes in the warehouse (Coram, Gildemeister)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-warehouse-slotting-pick-path-optimization-coram-gildemeister/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-warehouse-slotting-pick-path-optimization-coram-gildemeister.md", "department": "Logistics / Supply Chain", "department_url": "https://aiagentsformanufacturing.com/departments/logistics/", "evidence": "Verified", "summary": "AI warehouse slotting and pick paths: Coram cut travel 25% and picks 15% faster; Gildemeister cut picker travel 15-20% and temporary labour 67%.", "specialized": false, "who_this_is_for": null, "role": "Logistics (warehouse: where each item is stored and the route pickers walk)", "problem": "In a finished-goods or spare-parts warehouse, items are often stored where there was room when they arrived, and pick routes follow a fixed order. Demand changes every week, so fast-selling items end up at the back or on top shelves, and pickers walk far for each order line. In peaks, the company hires temporary workers or pays overtime to keep up. Coram, a Dutch manufacturer of bathroom fixtures and accessories exporting to more than 50 countries, faced rising labour demands as order volumes swung with promotions and large projects. Automotores Gildemeister, a vehicle distributor and dealer group in Chile, Peru and Costa Rica, manages more than 53,000 spare-parts SKUs; its old tool reclassified inventory only once a month, so fast-moving parts could stay outside the best locations for weeks, and each peak season it hired three extra workers just to keep up with picking.", "how_ai_helps": "Two related uses of AI on warehouse transaction history. Pick path optimization: instead of a fixed route, the system computes a shorter route through the warehouse for each pick list, without changing the layout. Dynamic slotting: a machine-learning model reads 12 months of picks, finds the small share of items that make up most of the picks, and recommends moving them to the 'golden zone' (the locations closest to the pickers and at waist height); it reruns daily as demand changes. The recommendations feed the warehouse management system (WMS), so workers keep following their usual put-away and pick instructions.", "result": "Coram (bathroom fixtures and accessories manufacturer, Netherlands), Infor WMS with Infor Velocity Suite AI pick path optimization: 15% faster picking, 25% shorter travel distance, less reliance on temporary labour during demand spikes, with no change to warehouse layout or workflows (Infor customer story). Automotores Gildemeister (automotive distributor; spare-parts distribution centers in Peru and Chile), Infor AI/ML dynamic slotting engine connected to Infor WMS, deployed in about three months: the model found that 6 to 10% of SKUs make up 80% of all picks; golden-zone pick rates rose from about 20 to 25% to 90% in Peru and 70% in Chile; picker travel time down 15 to 20%; temporary labour down 67% and overtime hours down an estimated 25 to 30% in peaks; 'Before, we hired three extra workers every peak season just to keep up with picking. Now we handle more volume with one' (aftermarket logistics manager). Infor also reports distribution center productivity up 70% (4.69 to 7.96 lines per hour), but for the broader digitalization program, not slotting alone (Infor customer story).", "companies": [ { "name": "Coram", "url": "https://aiagentsformanufacturing.com/companies/#coram" } ], "tools_deployed": [ { "name": "Infor Velocity Suite (Infor WMS)", "url": "https://aiagentsformanufacturing.com/tools/#infor-velocity-suite-infor-wms" } ], "tools": [ "Infor WMS with Infor Velocity Suite (AI pick path optimization): used at Coram", "Infor AI/ML (Infor Velocity Suite) and Infor Data Fabric, dynamic slotting engine connected to Infor WMS: used at Gildemeister", "Other tools to evaluate: the slotting module of your WMS (for example Mecalux Easy WMS Slotting, Manhattan, Blue Yonder, SAP EWM), Lucas Systems dynamic slotting, Optricity", "For a first analysis without new software: a pick history export plus any LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below, and a spreadsheet for the counts" ], "sources": [ "https://www.infor.com/en-gb/customer-stories/improving-labor-efficiency-ai-driven-warehouse-pick-path-optimization", "https://dam.infor.com/api/public/content/0ec52d499e074ee6bfbebc175f2b2d77?v=91b86a28" ], "date_added": "2026-10-28", "last_updated": "2026-10-28" }, { "title": "AI reads motor current from the electrical cabinet to warn of failing conveyors, fans and pumps weeks or months ahead (ArcelorMittal)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-motor-current-signature-monitoring-from-the-electrical-cabinet-arcelormittal/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-motor-current-signature-monitoring-from-the-electrical-cabinet-arcelormittal.md", "department": "Maintenance", "department_url": "https://aiagentsformanufacturing.com/departments/maintenance/", "evidence": "Verified", "summary": "AI motor current monitoring from the electrical cabinet: ArcelorMittal caught 27 conveyor failures up to 7 months ahead; a panel plant had no fan failures.", "specialized": false, "who_this_is_for": null, "role": "Maintenance & reliability (condition monitoring of motor-driven equipment)", "problem": "Many motor-driven machines are hard to monitor with handheld routes or vibration sensors: they are hot, dirty, moving, behind guards or simply too many to walk to. Faults in bearings, belts, couplings or rollers then show up as breakdowns or quality problems. At ArcelorMittal's hot strip mill in Ghent, Belgium, conveyors move plates of hot steel through the rolling process, and sensors mounted on the equipment cannot handle the temperatures. 'It's virtually impossible to use manual or vibration-based techniques to assess their condition,' says Andy Roegis, industrial digitalization manager for northern Europe at ArcelorMittal.", "how_ai_helps": "Instead of putting sensors on the machine, small high-frequency sensors are clipped onto the power cables inside the motor control cabinet (the electrical room where the motor starters and drives sit). They measure current and voltage 24/7. AI-based analytics first learn the normal electrical 'signature' of each motor and the machine it drives (about four weeks of automatic learning at ArcelorMittal), then watch for changes: a rising peak at the frequency of a belt, a bearing or the shaft rotation points to a specific developing fault. The system sends an alert with the suspected cause, maintenance inspects and repairs at a planned stop, and the data then confirms that the signature is back to normal.", "result": "ArcelorMittal (steel, hot strip mill in Ghent, Belgium), Samotics SAM4: during a 12-month pilot on the motors driving the hot-strip conveyors, 12 developing faults were detected, all confirmed by the mill's maintenance engineers, some up to 4 months in advance ('That enabled us to schedule maintenance at the optimal time,' Roegis). ArcelorMittal then rolled the system out to 56 critical runout table rollers, where it detected 27 more failures, some up to 7 months in advance, which the mill estimates saved 31 hours of downtime (Samotics case study). Wood-based panel manufacturer (not named), SAM4 on eight exhaust fans of one production line: 6 developing faults detected over 15 months (fixed by replacing a clutch, impeller or bearings, cleaning an impeller, or welding holes in the fan housing), no unexpected fan failure in that period, up to 12 hours of unplanned downtime prevented at a cost of €8,000 per hour, or up to €96,000 (Samotics case study). Food and beverage manufacturer (not named): a slow rise at a cooling pump's rotation frequency led to an alert; three weeks later the trend spiked and the inspection found a coupling missing one of its bolts, fixed at the next planned stop: 'all we had to do was put in a new bolt' (Samotics case study).", "companies": [ { "name": "ArcelorMittal", "url": "https://aiagentsformanufacturing.com/companies/#arcelormittal" } ], "tools_deployed": [ { "name": "Samotics SAM4", "url": "https://aiagentsformanufacturing.com/tools/#samotics-sam4" } ], "tools": [ "Samotics SAM4 (sensors in the motor control cabinet, AI analysis of current and voltage): used at ArcelorMittal Ghent and at the wood-panel and food and beverage plants", "Other tools to evaluate: other electrical signature analysis (ESA or MCSA) systems and portable motor testers; vibration-based monitoring where sensors can be mounted on the machine (see the related vibration case)", "For the first step without new software: your asset list and CMMS failure history, plus any LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below to choose the pilot assets" ], "sources": [ "https://samotics.com/proof/case-studies/how-arcelormittal-prevented-31-hours-of-downtime-by-detecting-27-failures-ahead-of-time", "https://samotics.com/proof/case-studies/prevent-exhaust-fan-failure-in-wood-manufacturing-case-study", "https://samotics.com/proof/case-studies/condition-monitoring-cooling-pump-misalignment-unbalance" ], "date_added": "2026-10-29", "last_updated": "2026-10-29" }, { "title": "AI drafts, checks and translates technical documentation and service manuals (Kistler, Bollinger Motors, Hitachi IES)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-technical-documentation-manuals-drafting-translation-kistler-bollinger/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-technical-documentation-manuals-drafting-translation-kistler-bollinger.md", "department": "Engineering", "department_url": "https://aiagentsformanufacturing.com/departments/engineering/", "evidence": "Pilot", "summary": "AI technical documentation: Kistler cut estimated cost per document from €4,000 to €40 in 14 languages; Hitachi IES cut translation time by over 60%.", "specialized": false, "who_this_is_for": null, "role": "Engineering (technical documentation: manuals, datasheets, service and product information)", "problem": "Every product needs manuals, datasheets, service procedures and wiring information, often in many languages, and they must change each time the product changes. Writing them takes engineers' time, translation agencies are slow and costly, and customers and colleagues still struggle to find the answer in long documents. Kistler, a Swiss maker of measurement technology (sensors and systems), had to translate and distribute technical documentation in 14 languages; the manual process was expensive and slow, and customers and internal teams searched through long files for technical information. Bollinger Motors, a new maker of medium-duty electric trucks in North America, assembled systems from many suppliers that did not supply detailed documentation, so writing service manuals and wiring diagrams meant reverse-engineering similar parts and authoring from scratch. Hitachi Industrial Equipment Systems (air compressors, transformers, motors, Japan) went global from 2021 and needed documents in Japanese and English; free translation tools mistranslated technical terms, and drafting a 10-slide bilingual presentation took two to three days.", "how_ai_helps": "Three uses of the same idea: AI produces the first draft or the translation, and people check it. At Kistler, an AI system called KIA generates technical content with several models (ChatGPT, Gemini and Claude), a second AI agent reviews it for clarity and consistency before publishing, and validated content is translated automatically into 14 languages; a virtual assistant built on 8 GB of technical documentation answers questions. For Bollinger Motors, the documentation services firm MSX International added a generative AI assistant called 'Pete' (built on OpenAI) to its content management system, pubFoundry: it gathers information on a topic, drafts the technical content in the XML structure the system needs, and the MSX team refines and validates it. At Hitachi IES, DeepL Translator is used company-wide to translate whole files, including PowerPoint, with better handling of technical terms than the free tools it replaced.", "result": "Kistler (measurement technology), KIA built by the AI agency margarit.co: estimated cost per document down from €4,000 to €40, production time from days to hours, content in 14 languages, '95% recorded accuracy' (not defined); the assistant handles most frequent questions (margarit.co project page and case study, an agency claim). Bollinger Motors (electric trucks), MSX International pubFoundry with 'Pete': less time and effort to create service content despite missing supplier documentation; the AI draft 'isn't entirely ready for immediate use' but gives the team a solid base; no figure published (MSX case study). Hitachi Industrial Equipment Systems, DeepL Translator: translation time reduced by more than 60%; a presentation that took 2 to 3 days to prepare in both languages now takes one day; the number of users doubled in six months (DeepL customer story).", "companies": [ { "name": "Bollinger Motors", "url": "https://aiagentsformanufacturing.com/companies/#bollinger-motors" }, { "name": "Hitachi Industrial Equipment Systems", "url": "https://aiagentsformanufacturing.com/companies/#hitachi-industrial-equipment-systems" }, { "name": "Kistler", "url": "https://aiagentsformanufacturing.com/companies/#kistler" } ], "tools_deployed": [ { "name": "DeepL Translator", "url": "https://aiagentsformanufacturing.com/tools/#deepl-translator" }, { "name": "MSX International pubFoundry", "url": "https://aiagentsformanufacturing.com/tools/#msx-international-pubfoundry" } ], "tools": [ "KIA (Kistler Artificial Intelligence), built by margarit.co on ChatGPT, Gemini and Claude with a review agent and automatic translation: used at Kistler", "MSX International pubFoundry content management system with the 'Pete' generative AI assistant (OpenAI): used for Bollinger Motors", "DeepL Translator: used at Hitachi Industrial Equipment Systems", "Other tools to evaluate: the AI features of your documentation or content management system, DeepL or other machine translation with your own glossary, translation memory tools such as RWS Trados or memoQ", "Any LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below to structure one procedure and check terminology before choosing a tool" ], "sources": [ "https://margarit.co/en/projects/kistler", "https://margarit.co/blog/caso-de-estudio-kia-kistler-artificial-intelligence-transfor", "https://www.msxi.com/case-studies/advancing-ev-technical-information/", "https://www.deepl.com/en/customer-stories/hitachi-ies" ], "date_added": "2026-10-29", "last_updated": "2026-10-29" }, { "title": "AI sorts the shared customer inbox and routes each email (order, quote, after-sales, invoice) to the right person (Safran Landing Systems, Dossche Mills)", "url": "https://aiagentsformanufacturing.com/use-cases/ai-shared-inbox-email-triage-routing-safran-landing-systems-dossche-mills/", "markdown": "https://aiagentsformanufacturing.com/use-cases/ai-shared-inbox-email-triage-routing-safran-landing-systems-dossche-mills.md", "department": "Office & management", "department_url": "https://aiagentsformanufacturing.com/departments/office/", "evidence": "Verified", "summary": "AI shared-inbox triage: Safran Landing Systems cut manual email processing 70%; a 45-person maker cut daily sorting from 3 h 15 to 53 minutes.", "specialized": false, "who_this_is_for": null, "role": "Customer service and sales administration (shared mailboxes such as orders@ or sales@)", "problem": "Orders, quote requests, delivery questions, complaints, invoices and spam all arrive in the same shared mailbox. Someone has to read each message, decide what it is and forward it, often first thing in the morning, so a message that arrives in the afternoon waits until the next day and urgent orders get lost. Safran Landing Systems, a maker of landing gear and braking systems for aircraft, receives 15,000 to 20,000 emails a month across five countries: requests for quotes, purchase orders, delivery requests and technical questions. Dossche Mills, a Belgian family-owned flour miller founded in 1875, received orders and requests in several multilingual shared inboxes across Europe, triaged them by hand across Outlook, SAP and SharePoint, and saw the workload grow as experienced employees left without being replaced. In a 45-person French components maker, the contact@ and orders@ addresses received 200 to 300 emails a day; a sales assistant spent on average 3 h 15 every morning sorting them, and in the quarter before the project 9 orders were handled late or lost, after which two customers went to competitors.", "how_ai_helps": "The AI reads each incoming email in full (subject, body, sender, attachments), puts it in a category (order, quote request, after-sales, accounting, supplier, other), pulls out the key details (customer, product reference, quantity, urgency) and sends it to the right team or person. Urgent orders are flagged at once instead of waiting for the morning sort. When the AI is not confident, or the message mixes several requests, it goes to a person to decide. In the French SME, product references are checked against the real catalogue and order data goes into the ERP only as a draft; at Dossche Mills the system also prepares draft replies and order entries in SAP; at Safran Landing Systems it works in English, French and Chinese for five sites.", "result": "Safran Landing Systems (aircraft landing and braking systems), Tungsten TotalAgility delivered by the integrator DS Informatique: 70% reduction in manual email processing, more than 20,000 emails a month handled across five sites, emails classified automatically in English, French and Chinese; described as Safran's first successful AI pilot for sales administration (Tungsten case study, a vendor claim). Dossche Mills (flour and bakery ingredients, Belgium), Tekst: 'high accuracy' on email classification, order volume maintained with a reduced workforce, order processing time per case 'meaningfully' cut; no figures published (Tekst customer story). French components maker (45 people, not named), AI agent built by Vantacrew, measured six months after going live at the end of 2025: daily sorting time from 3 h 15 to 53 minutes (-73%), first-reply time from 26 hours to 4 h 30, lost orders from 9 to 1 per quarter, 89% of emails classified without human help, 2.3% routing errors, ERP re-keying down 64%; cost €9,400 setup plus €410 a month, payback 4.2 months by the vendor's calculation; the assistant moved to sales follow-up and nobody was laid off (Vantacrew case study, a vendor claim).", "companies": [ { "name": "Dossche Mills", "url": "https://aiagentsformanufacturing.com/companies/#dossche-mills" }, { "name": "Safran Landing Systems", "url": "https://aiagentsformanufacturing.com/companies/#safran-landing-systems" } ], "tools_deployed": [ { "name": "Tekst", "url": "https://aiagentsformanufacturing.com/tools/#tekst" }, { "name": "Tungsten TotalAgility", "url": "https://aiagentsformanufacturing.com/tools/#tungsten-totalagility" }, { "name": "Vantacrew", "url": "https://aiagentsformanufacturing.com/tools/#vantacrew" } ], "tools": [ "Tungsten TotalAgility (AI email classification and routing), delivered by DS Informatique: used at Safran Landing Systems", "Tekst (email classification, routing, draft replies and SAP order automation): used at Dossche Mills", "Custom AI agent built by Vantacrew on Claude models, connected to the shared mailbox and the ERP: used at the French components maker", "Other tools to evaluate: AI classification in Microsoft Power Automate (AI Builder) or in your help desk or CRM, shared-inbox tools such as Emailgistics", "To test the idea without new software: 30 to 50 anonymized emails and any LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below" ], "sources": [ "https://www.tungstenautomation.com/learn/case-studies/2025/safran-automates-email-processing-with-ai-totalagility", "https://www.tekst.com/customer-stories/dossche-mills", "https://vantacrew.ai/fr/blog/pme-industrielle-tri-emails-commandes-ia-cas-client" ], "date_added": "2026-10-29", "last_updated": "2026-10-29" } ] }