Principles, written down

Domain intelligence in a regulated institution is accidental, leaking, and missing. It exists where projects happened to be won, it walks out when people leave, and whole domains have none. Knowledge management worked on this for decades and did much of it well. What changed is the forcing function: once AI agents act on institutional knowledge, that knowledge has to be governed at the level of the claim, at machine speed, under supervisory scrutiny.

The manifesto defines governed intelligence as the operational discipline this requires. The unit is the claim, with provenance, an epistemic tier it has earned, temporal validity, scope, and governance status. The lifecycle runs continuously: ingest, consolidate, curate, expand, apply. Curation is the immune function, defending the substrate against pollution, staleness, fragmentation, amnesia, cascade failure, and structural distortion.

Six values

Twelve principles

Four integrity preconditions

Numbered 13 to 16 in the text for cross-reference. A substrate cannot be called operational without them.

An enterprise that uses agents is not yet an agentic enterprise. The manifesto reserves the term for an organization where governed agents, operating on governed intelligence, exercise collective initiative: they perceive what matters, reason about institutional consequences, and raise opportunities for action before being asked. It depends on the Intelligence Governance Manifesto for the substrate and inherits the engineering principles beneath it, and it needs both in place.

Six values

Sixteen principles