AI agents can now ask to pay, prove they’re authorized, and check out using new protocols. The protocols cover the request, but settling the payment is left to whoever processes it.
Before a merchant has the funds, five questions need answers: who the agent is and who it acts for, whether the payment is within its limits, which asset and rail carry the value, which compliance checks apply, and how the outcome is recorded so finance can reconcile it.
We wrote The Execution Gap on what happens after an agent decides to pay. It walks through four transaction flows, from a pay-per-call API to an enterprise agent spending against a budget, and what has to happen at each step.
→ Read the report: