# AI reads supplier delivery notes and prepares the goods receipt in the ERP (Siemens, RAUMEDIC)

> AI reads supplier delivery notes and prepares the goods receipt: Siemens reports 90%+ touchless processing and 98% accuracy; RAUMEDIC over 99%.

- Canonical URL: https://aiagentsformanufacturing.com/use-cases/ai-delivery-note-goods-receipt-erp-siemens-raumedic/
- Department: Logistics / Supply Chain (https://aiagentsformanufacturing.com/departments/logistics/)
- Evidence level: Verified. Real deployment at a named or precisely described company, with a measured result from a primary source.
- Added: 2026-10-20
- Last updated: 2026-10-20
- Publisher: AI Agents for Manufacturing (https://aiagentsformanufacturing.com)

## Role

Logistics (receiving and goods receipt: inbound deliveries). Typical users: receiving clerk, warehouse lead, inbound logistics coordinator, inventory controller, materials manager

## The 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.

## Step by step

1. Count one month of receipts: delivery notes per day, number of suppliers, minutes to key each note, and how many differences are found later at invoice matching.
2. Start with the three to ten suppliers that send the most delivery notes, as RAUMEDIC did, and collect at least 50 sample notes from each.
3. Write down what must match the PO: item, quantity and tolerance, unit of measure, batch, production and expiry dates, certificates.
4. Run in draft mode: the AI extracts and matches, and the receiving clerk confirms every goods receipt; measure accuracy field by field.
5. Allow automatic posting only for clean matches above a confidence threshold; everything else goes to a person with the reason.
6. Keep the physical check: if the counted quantity or condition differs from the note, the clerk's count wins. Track touchless rate, posting errors and time from dock to stock in the ERP.

## 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

## 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.

As reported by the source; most figures are vendor or self-reported and not independently verified.

## Human review: cautions

- The physical count and condition check stay with a person; the AI posts what the paperwork says, not what is on the pallet.
- Start in draft mode and allow automatic posting only for suppliers and fields that have been accurate over weeks.
- For food, pharma and other regulated materials, batch and expiry data must be verified against the label before release to production.
- Figures are vendor or integrator-published; the Siemens figures come from a summary of the vendor's case.

## Try it yourself (prompt)

```text
You are an experienced receiving clerk at a manufacturer.

I will paste: (1) the text or photo of one or more supplier delivery notes; (2) the matching open purchase orders from our ERP [PO number, line, item number, supplier item number, description, quantity ordered, quantity already received, unit of measure, due date]; (3) our receiving rules [quantity tolerance, items that need batch or expiry dates, certificates required].

Do this for each delivery note:
1. Extract supplier, delivery note number, date, PO number(s), and each line: item, description, quantity, unit, batch or lot, production and expiry dates.
2. Match each line to a PO line and mark Match / Quantity difference / Unknown item / Missing batch or date / Wrong unit, quoting both values.
3. Prepare the goods receipt lines for the clean matches only, in a table I can check.
4. List the exceptions with the reason and the question to ask the supplier or buyer.

Never round or change a quantity, and never assume a batch or expiry date that is not printed on the note.
Use only the data I paste in; if something is missing, write 'not found' or list it as a question instead of guessing. Do not send, order, post or change anything in any system: produce drafts for my review and approval.
```

## Related use cases

- [Agent confirms supplier PO acknowledgements in the ERP](https://aiagentsformanufacturing.com/use-cases/agent-po-confirmation-erp/)
- [AI accounts payable invoice processing](https://aiagentsformanufacturing.com/use-cases/ai-accounts-payable-invoice-processing/)
- [AI orchestrates a plant-attached warehouse: more loads shipped straight from the plant, fewer shuttle moves (P&G Lima)](https://aiagentsformanufacturing.com/use-cases/ai-plant-warehouse-orchestration-pg-lima/)
- [AI predicts when shipments will really arrive, so planners act before a late delivery stops a line (Faurecia, Magna)](https://aiagentsformanufacturing.com/use-cases/ai-predictive-eta-shipment-tracking-line-stops-faurecia-magna/)

## 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

---
Please cite the canonical URL: https://aiagentsformanufacturing.com/use-cases/ai-delivery-note-goods-receipt-erp-siemens-raumedic/