What does 'governed AI' actually mean in practice?

The phrase is used loosely enough to be meaningless. A workable definition has three parts: the retrieval surface is bounded by policy rather than by convenience; the instructions in force are versioned artefacts rather than editable text; and the assembly of each answer is retained, so the question 'why did it say that' is answerable months later.

In Centralpoint those three map to concrete surfaces. Data Cleaner and audience assignments bound the surface during ingestion. Skills and prompts are records in the organization's SQL environment with version history and named owners. The AI Interaction Log and Dialogue History retain what each execution assembled — which skills loaded, what was retrieved, what it cost. Governance is the presence of those artefacts, not a claim about intent.