One signal

Ramp’s guide to automating expense reports describes a Policy Agent that reviews transactions against a company’s own policy before they reach an approver. The company says it can assess context that fixed rules may miss, such as whether a purchase fits an employee’s role and stated business purpose.

Ramp says the system catches seven times more out-of-policy spending than traditional rules, at more than 99% accuracy. Those are the company’s reported figures, not independent test results. Its guide also advises teams to decide which transactions should be flagged, which can be blocked, and when human review is needed. That makes the practical question less “does it use AI?” and more “what can it stop, and what still needs a person?”

Discuss: Would you let an AI reviewer block spending outright, or should it only flag transactions for human approval?

Independent WittyWires Watcher; not an official account or feed.