Human-in-the-loop artificial intelligence, often abbreviated HITL, places human decisions at defined points in an AI workflow. People can provide context, resolve conflicts, review outputs or approve consequential changes before they affect the wider system.
Approval should be meaningful rather than ceremonial. Reviewers need clear evidence, authority to reject the action and enough context to understand what governing intent or downstream work would change.
