Human-in-the-Loop (HITL)

Human-in-the-loop (HITL) is a design pattern in which an AI system pauses at defined points and waits for a person to approve, edit, or reject its proposed action before it continues. The system reaches a checkpoint, such as a tool call that sends an email or moves money, saves its state, and resumes only after a human decision is recorded. It differs from human-on-the-loop, where a person monitors a running system and can intervene but does not block each step, and from full autonomy, where no checkpoint exists. In agentic systems, HITL is usually applied selectively to irreversible, costly, or regulated actions, while low-risk steps run without review.

Agent frameworks implement the pattern as interrupts or deferred tool calls: the run stops, the pending action and its arguments are shown to a reviewer, and the reviewer's answer is fed back into the same run. A related mechanism, agent-to-human handoff, passes the whole conversation to a person instead of asking for a single decision. For high-risk AI systems, Article 14 of the EU AI Act requires that they be designed so people can oversee them effectively.

Related terms

• Agent-to-Human Handoff
• Reinforcement Learning from Human Feedback (RLHF)
• HubSpot Agent AI
• HVAC AI Agent

Related services: Agentic AI consulting.

← Back to the AI Glossary