This page describes what the system may do on its own and where the limit sits. Who sets those limits and how they are audited is under operational boundaries; the place in the architecture is under overview.
Autonomy Within Boundaries
The autonomy engine governs self-organizing behavior, adaptive response calibration, and operational decision boundaries. Critically, it operates within configurable constraint boundaries set by the deploying organization itself — Henri doesn't decide how far its own authority extends; the operator does.
Classification and Action
When pattern confidence crosses a threshold, Henri selects a response from its action library — depending on the action class, it either executes autonomously or flags the case for operator review. That distinction is the concrete implementation of operational boundaries: low-risk, well-understood responses run autonomously, everything outside that scope goes to a human.
When Confidence Isn't Enough
Henri logs its own known limitations and tracks hypothesis accuracy internally. Where confidence is insufficient to act, the system explicitly flags this rather than guessing or executing an uncertain action. That's a deliberate design principle, not a shortcoming: an open question to the operator beats a wrong autonomous decision.