A few weeks ago I was at an AWS roundtable in Auckland. A dozen heads of platform around one table, every one of them shipping AI agents into production, every one of them describing the same gap. Their agents could read AWS docs. They could call the AWS API. They could write Terraform. They could even, on a good day, propose a fix for a real incident. What none of them could do: tell on-call whether the API exposure on the billing service is a regression — or a 30-day risk acceptance the team signed off on last quarter. That gap is what this whole series is about. I'll show you the AWS DevOps Agent setup I built to close it. The companion implementation is Intent Guard — a demo I'll publish alongside this series (anonymised; I can't share my employer's copy of it). The thing that took me longest to internalise: the hard part of an org-aware agent isn't the AI .…