Human-in-the-Loop (HITL)
Human-in-the-loop (HITL) is a design where a person reviews or approves an AI system's output before it takes effect, keeping human judgment over decisions that carry risk.
It keeps a person in control of consequential actions while the AI does the heavy lifting of drafting, analyzing, or proposing. In practice, an agent can prepare an action, such as updating a record, sending a message, or advancing a deal, but a person signs off before it commits.
It is one point on a spectrum: human-in-the-loop (review every action), human-on-the-loop (monitor and intervene), and human-out-of-the-loop (full autonomy with after-the-fact audit). Teams often start with HITL for higher-risk tasks and relax oversight as reliability is proven.
Frequently Asked Questions
It means a person reviews and approves an AI system's output before it takes effect, keeping human judgment over consequential actions.