Most government organisations have now piloted AI. Far fewer can evidence what it delivered, or sustain it without ongoing supplier support. Drawing on Phoenix Software's public sector delivery experience, this session shares three practical lessons: how to measure AI so that outcomes stand up in a business case, how a peer community is helping organisations share use cases and lessons across government rather than solving the same problems in isolation, and how a champion-led, build-with delivery model turns AI adoption into a repeatable in-house capability. Delegates will leave with a measurement approach, a collaboration model and a clear view of how to make AI skills stick.