Unilever and Nestlé Use AI to Handle Orders and Chase Late Payments

Big companies are handing more of the paperwork between a sale and a payment to AI. Unilever says its order system now takes about 90 seconds to start processing an order, down from 15 to 20 minutes. Nestlé is using SAP tools that help finance teams decide which unpaid accounts to chase first.

What changed

Unilever's system, called NOVA, reads customer orders that arrive as emails, PDFs, spreadsheets and photos of paper forms. Four AI agents read the messages and pull out the order details. A fifth checks the work. Anything missing or unusual goes to an employee. Unilever reports that 95% of orders in 11 pilot markets pass through without manual handling.

Unilever handles roughly 20 million orders a year. Most move directly between computer systems. About 2 million arrive in those messier formats, according to a Microsoft case study.

Chasing payments

After an invoice goes out, someone has to decide which customer to contact, what is blocking payment and whether a disputed charge is backed by the records. Nestlé is using SAP's Accounts Receivable Assistant to rank collections by risk and business context. SAP's Oct. 26 update lists seven receivables agents, covering tasks like account analysis, emails and dispute handling. SAP says a phased rollout is underway.

What the survey found

PYMNTS Intelligence and JPMorgan surveyed 100 U.S. chief financial officers in July 2026. Their companies have annual revenue between $250 million and $2.5 billion. The results were compared with a 2025 survey of 375 CFOs.

• Companies automate an average 73% of accounts receivable work, up from 55%.
• 90% of CFOs say connecting systems is a barrier, up from 67%.
• 38% saw customers improve payment-term behavior, down from 73% in 2025.

What still gets in the way

Faster internal work does not make customers pay faster. The research found customers asking for longer payment terms even as finance teams sped up forecasting and invoicing. Nestlé's focus on consistent processes and trusted data points to the other hurdle: AI can only act well on records that are clean and connected.

Who it affects

Any business that sells to other businesses deals with orders, invoices and late payers. Suppliers who accept customers' own documents still have to convert them, and AI is now doing much of that work, with people stepping in on unresolved cases.