A global analysis by Basware suggests large organisations are losing visibility over nearly half a trillion dollars each year due to invoice errors and missing purchase orders. Of 272 million invoices processed through the company’s systems in 2023, around 40% arrived without an associated purchase order — creating what Basware calls a $549 billion blind spot in uncontrolled spend.
For accounts payable (AP) teams, those non-purchase-order invoices represent a persistent drain on time and accuracy. Each requires manual checking, coding, and approval before payment — a process that Basware estimates consumes about 10 per cent of AP staff time, or more than 200 hours annually per person.
The Finnish software group believes its new “Agentic AI” system can restore control. Integrated into its InvoiceAI platform, the technology reads and validates both PO and non-PO invoices automatically, routing them for approval without manual input. Basware describes the system as acting like “a skilled AP clerk who thinks, decides, and solves problems”, capable of recommending actions rather than waiting for instructions.
“Finance teams are under massive pressure to reduce costs while improving operations,” said Perttu Nihti, Chief Product and Technology Officer at Basware. “Regular AI is like a calculator — you have to key in an equation to get an answer. Agentic AI doesn’t wait for instructions. It sees a non-PO invoice, understands what needs doing, and recommends a course of action.”
Early adopters are reporting material gains. At global cable manufacturer Nexans, Basware’s SmartPDF function — part of the same automation suite — has shortened invoice processing from days to hours. Cécile Barrère, the company’s AP Director, said the rollout brought “better visibility and traceability” while cutting the risk of error and fraud. Within seven weeks, Nexans achieved a 92 per cent reliability rate in automatic data capture.
Basware positions Agentic AI as a response to growing pressure on CFOs to maintain control over complex supplier networks without expanding headcount. As finance automation moves beyond rule-based processing, the technology could mark a shift towards systems capable of interpreting intent as well as data — turning what was once a blind spot into a source of operational insight.
