AI 101

Human in the loop

Also known as: HITL · human oversight (interactive)

Agents

Human-in-the-loop (HITL) means a person approves, corrects, or takes over specific agent actions at defined checkpoints.

  • The agent proposes; the human disposes on sensitive steps.
  • Checkpoints: before spending money, before deleting, before sending.
  • The difference between “autonomous” and “accountable.”

Where to put it: insert humans at *irreversible or high-cost* actions. You don't need approval for a read; you do for a write, a send, or a spend.