Danish wholesaler Lemvigh-Müller deployed a three-agent AI workflow built on SAP Business AI—implemented with NTT DATA Business Solutions—to automatically process more than 100,000 supplier order confirmations each year, cutting handling from hours to minutes.
'We have previously tried both RPA and traditional automation approaches without really achieving the desired effect. The key difference this time is that we broke the task down into multiple independent AI agents, each responsible for a specific part of the process. Together, they now handle what previously required manual review,' IT director Frederik Aakerlund said. Klaus Heinemann, head of SAP ERP at Lemvigh-Müller, added: 'Previously, when order confirmations were handled manually, it could take hours or even days before changes were reflected across the organization. Today, the AI agents update the data almost immediately, allowing customers to receive a much more accurate picture of deliveries far sooner.'
Lemvigh-Müller is one of Denmark's largest wholesalers of steel, plumbing, heating and electrical products. It sends roughly 175,000 purchase orders a year to more than 2,000 suppliers, but about 60% of supplier order confirmations still arrive as unstructured PDFs. Even small discrepancies in price, quantity or delivery dates previously triggered extensive manual procurement work.
The project began with an internal experiment in which AI-savvy project manager Jess Frederiksen successfully used ChatGPT to match an order confirmation against a purchase order. From that initial test to live production took 10 weeks—an unusually short implementation window for a procurement-grade automation. The deployment has continued to mature, with the company expecting to free up 3-4 FTE equivalents over time.
Three specialised AI agents are orchestrated into a single SAP-integrated workflow. The first (e-mail agent) receives and triages supplier messages and attached PDFs. The second (data-extraction agent) extracts prices, quantities and delivery dates from PDFs and structures the data for SAP comparison. The third (matching agent) compares the extracted data against the purchase order in SAP and either continues automatically or flags deviations. The whole workflow runs on SAP Business AI with NTT DATA Business Solutions integration.
More than 100,000 supplier order confirmations per year are processed end-to-end (out of roughly 175,000 annual POs). The system identifies price discrepancies before invoicing, surfacing them for resolution with suppliers.
Lemvigh-Müller expects the AI agents to keep absorbing routine work, with freed capacity redeployed to complex, exception-driven orders. Master-data quality (Incoterms and similar reference data) is a flagged focus area.