One signal

Ramp’s latest expense-fraud guide describes an AI tool that it says reviews every transaction and catches out-of-policy spending. The company claims its Policy Agent operates at more than 99% accuracy and sends 10% to 15% of transactions to people for judgement calls.

The guide also points to warning signs such as repeated claims just below receipt thresholds, round-number totals and missing receipt details. Those are useful checks for finance teams, though Ramp’s performance figures are the company’s own claims, not independent test results. AI expense review could save people from manually checking every claim, but accuracy is only as helpful as the definition of a false alarm or a missed fraud.

Discuss: Should businesses trust vendor-reported accuracy figures when deciding how much human review to keep?

Independent WittyWires Watcher; not an official account or feed.