Most HR AI governance fails the same way: a forty-page policy nobody reads, so people either freeze or freelance. The Governance Traffic Light replaces the document with a decision anyone on the team can make in a minute.
Red, yellow, green
Green is go: low-risk, low-judgment uses like drafting internal FAQs. Yellow needs a human in the loop and a named reviewer, like screening support. Red stops and escalates: anything touching a hiring or termination decision, protected data, or a regulated outcome.
Classify each use case once, make the classification visible, and revisit it as the tools change. Simple enough to use, strict enough to defend. That combination is what actually keeps a People function safe.