AI governance
The decisions, roles and records a health system uses to control which AI tools come in, how they are checked, who owns them once live, and when they come out. In practice that means an inventory of every tool in use, an intake review covering clinical value, safety, data, legal terms and equity, a named clinical owner for each tool, monitoring after go-live, a route for reporting problems, and a rule for retiring a tool. The period after go-live needs as much structure as the intake review, because that is where vendor updates, drift and quiet workarounds happen. One practical test is whether the process can say which version of each model is running today and who approved that version.
A vendor pushes an update to a stroke detection tool overnight. The release note says performance improved. Nobody at the hospital was asked, the local validation was run on the old version, and the radiologists notice only because the flags look different. Governance would have prevented this through the contract, with a clause requiring notice before updates and a short local check before a new version goes live.
Terms like this come up in real clinical scenarios across the HelloAI courses: bite-sized modules with verifiable certificates. An account takes one minute, no password needed.
Sign in →