ProcurementNewVerified Real deployment at a named or precisely described company, with a measured result from a primary source.

AI sorts every purchase line into spend categories so buyers see what they really spend with each supplier (Pentair, Barry-Wehmiller)

Added 2026-10-28 · Last updated 2026-10-28

Manufacturers in this case: Barry-Wehmiller · Pentair · all manufacturers using AI

Role

Procurement (spend analysis: knowing what the company buys, from whom, at what terms) · Typical users: purchasing manager, category manager or buyer, procurement analyst, controller, plant manager in a company with several sites or ERPs

The 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.

Step by step

Export 12 months of purchase lines (PO or invoice lines: supplier, item description, amount, site, payment terms) from every ERP or accounting system.
Pick a simple category list (10 to 30 categories you actually negotiate: raw materials, packaging, fasteners, MRO, freight, services) rather than a large standard taxonomy on day one.
Group supplier names that belong to the same company, and have a buyer check the groupings.
Let the AI classify the lines with a confidence level; a buyer reviews the low-confidence lines and the largest amounts, and the corrections become rules.
Look at the top categories and top suppliers by spend across all sites: duplicate suppliers, different prices or payment terms for the same item, spend with no contract.
Pick two or three actions (consolidate a category, align payment terms, renegotiate one supplier) and track the savings against a baseline.
Refresh the classification monthly so new suppliers and new descriptions do not drift into 'unclassified'.

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

In the AI tools index: Sievo Spend Analytics · all AI tools used by manufacturers

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).

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

Human review: cautions

The figures are published by the vendor (Sievo) with quotes from Pentair and Barry-Wehmiller; none is independently verified, and the savings also come from the negotiations people did with the new view.
Classification is only as good as the descriptions in your ERP: lines like 'misc' or 'see quote' need a human or better data at the source.
A buyer reviews the supplier groupings and the large or uncertain lines before anyone negotiates on the numbers.
Purchase data includes prices and supplier terms: use an approved company tool, not a public chatbot, and do not let an LLM add up amounts.

Try it yourself

You are a procurement analyst at a manufacturer.

I will paste: (1) our category list [category name, short definition, examples]; (2) purchase lines from last year [line ID, supplier name, item description, site, payment terms, amount].

Do this:
1. Group supplier names that look like the same company (spelling variants, Inc/Corp/Ltd, obvious subsidiaries) and show each group with its member names, marked 'BUYER CHECK'.
2. Assign each purchase line to one category from my list, with confidence high / medium / low and the words in the description that decided it. Lines that fit no category: 'UNCLASSIFIED' with a suggested new category.
3. List lines where the same item or category is bought from several suppliers, or with different payment terms, as candidates for consolidation or renegotiation.
4. Give me the classified lines as a table I can paste into a spreadsheet: line ID, supplier group, category, confidence.

Do not add up amounts or compute totals or savings yourself: I will build the totals in a spreadsheet pivot table from your table. 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.

Category list:
[paste]

Purchase lines:
[paste]

Want more like this? Join the newsletter

Related cases

All 12 Procurement use cases →