You put a gateway in front of your agents, wrote the policies, and tested them. Then a support agent issues a refund far outside its normal range, and nothing stops it. The gateway never saw the call because the action didn’t travel the route the policy was watching. Now you’re explaining to your VP of AI why the controls you were sure about only covered part of what the agent could do.
Where you place governance decides what it can stop, and most teams don’t find the boundary until something crosses it. This guide shows how gateway and in-flow governance differ, what each can actually see before an action runs, and why the answer changes for approvals, sensitive data, and audit evidence. You’ll work through five everyday scenarios that expose the gaps, get twelve questions to put to any agent governance vendor, and finish with a pilot-to-production checklist you can use this quarter. Download it to see where your coverage really ends.