Lemvigh-Müller, a 180-year-old Danish wholesaler of industrial building material, technical, and steel products, has built an AI use case that reads incoming business documents—automatically and within seconds.
For Lemvigh-Müller, an efficient supply chain isn’t a nice-to-have—it’s the business model. “Our company is low margin, and we are living from a very efficient supply chain,” says Frederik Aakerlund, CIO of Lemvigh-Müller. “We need to cut costs wherever we can, and we need to make sure our customers get our products as quickly as possible.”
Not every business partner connects via EDI (Electronic Data Interchange), the standard for exchanging business documents directly between IT systems. For Lemvigh-Müller, that means a steady stream of orders, delivery notes, and invoices arriving as PDFs and emails—documents that, until recently, had to be read and entered manually.
“Today we are receiving so many PDF files and emails that we don’t have the time to read them,” Aakerlund explains. “Basically, we don’t update our system, or we don’t find the deviations from what we expect, quickly enough.”
Letting AI read the mail
To close that gap, the team built a use case for receiving documents from business partners that can’t be exchanged via EDI and having them read automatically by SAP AI engines.
“This is exactly where AI is helping us,” Aakerlund says. “It’s reading 10 20-page documents in a few seconds, updating our system, and there’s no person involved.”
Behind the scenes, the solution combines a mix of SAP Business AI Platform, AI components, and SAP Fiori apps, integrated with Lemvigh-Müller’s core SAP system—SAP Cloud ERP Private, which the company adopted two years ago.
The shift in daily work is tangible. “Our users, instead of reading a lot of emails, are just working in a dashboard, finding the things they need to work on,” Aakerlund says. “We’re living the Autonomous Enterprise in a small part of our business.”
What started as a single use case has since become a template. “We’ve kind of made it a template for receiving business documents like orders, delivery notes, invoices, and so on,” Aakerlund notes. “Whenever we can’t get them digitally, we read them via this new system. It works for all kinds of PDFs and emails we receive from our business partners.”
Aakerlund’s advice to other companies considering AI projects: don’t start big.
“A good piece of advice could be to find the pockets of inefficiency in your company and apply AI there, instead of going for some really, really big project,” he says. The first version of the use case took just 200 hours over 10 weeks to build. “It turned out to be a reusable architecture, with reusable templates for a lot of other business processes.”
