Artificial intelligence agents are processing purchase orders and preparing collections work as businesses automate more of the steps between a B2B sale and payment.
PYMNTS Intelligence research with JPMorgan finds surveyed companies automate an average 73% of accounts receivable processes, compared with 55% in the previous year’s survey. Recent deployments extend into work that requires reading customer documents, checking records and deciding which unpaid accounts need attention.
Unilever’s NOVA system handles initial order processing in about 90 seconds across 11 pilot markets, down from 15 to 20 minutes, PYMNTS reported. The company reports that 95% of orders in those markets pass through without manual handling.
Order to cash starts with receiving a customer order and covers fulfillment, invoicing, collections and cash application, which matches incoming payments to outstanding invoices. In B2B transactions, purchase orders, delivery records, contractual terms and invoice disputes all form part of the work surrounding payment.
Much of the order-entry work is already digital. Unilever processes roughly 20 million orders annually, most of which pass directly between computer systems. About 2 million arrive through emails, PDFs, spreadsheets and photographs of paper forms, according to a Microsoft case study. Its customers include supermarket chains, wholesalers and distributors.
NOVA addresses those varied formats. Four agents read incoming messages, interpret their context and extract order information. A fifth checks the results. Missing information and exceptions go to employees for review. Unilever also reports using the technology for claims validation.
We’d love to be your preferred source for news.
Please add us to your preferred sources list so our news, data and interviews show up in your feed. Thanks!
For suppliers, accepting a customer’s existing documents leaves the conversion work inside their own operations. Employees previously keyed those orders into the system; agents now perform much of that task, with people handling unresolved cases.
Collections and Disputes
Receivables teams face another set of document and account decisions after an invoice goes out: which customer to contact, what’s preventing payment and whether a disputed charge is supported by the records.
PYMNTS’ coverage of Nestlé described how SAP’s Accounts Receivable Assistant can prioritize collections by risk and business context. Nestlé’s global head of IT innovation and enterprise architecture, Luca Dell’Orletta, said its SAP transformation across 140 countries provides the foundation for recommending collection actions and helping finance teams respond faster.
SAP’s Oct. 26 product update identifies seven receivables agents covering account analysis, collection preparation, email correspondence, dispute creation and resolution, and coordination. SAP says phased general availability is underway.
The PYMNTS Intelligence report “Time to Cash: What Two Years of Data Say About Faster Cash and Rising Risk” surveyed 100 U.S. chief financial officers in July 2026 at companies with annual revenue between $250 million and $2.5 billion. It compares their responses with a separate 2025 survey of 375 CFOs.
Integration complexity was cited by 90% of CFOs as a barrier to further AR automation, up from 67%. Nestlé’s emphasis on consistent processes and trusted data addresses the information requirements behind collections decisions.
Customers’ payment terms also affect collection timing. Only 38% of CFOs reported improvement in customer payment-term behavior, compared with 73% in 2025. The research found customers seeking longer terms even as finance teams accelerated internal processes such as forecasting and invoicing.
Add as Preferred Source
