Human-in-the-Loop

ELI5 - The Vibe Check

Human-in-the-loop is the 'are you sure?' dialog for AI agents. Instead of letting the agent fire off a destructive command, send a real email, or spend real money entirely on its own judgment, it pauses and waits for a human thumbs up before doing anything irreversible. The agent still does the work, a person just still holds the button for the scary parts.

Real Talk

Human-in-the-loop (HITL) describes an agent design where specific actions, such as destructive commands, financial transactions, or external communications, require explicit human approval before execution rather than running fully autonomously. It's a core mitigation for agent safety and a common configuration setting in coding agents and automation platforms, letting teams dial autonomy up or down per action type.

When You'll Hear This

"Anything that touches production stays human-in-the-loop, no exceptions." / "We removed the approval step and immediately regretted it."