Definition
Human-in-the-loop is a design where AI prepares or recommends an action and a named person approves it before it takes effect, with the decision logged.
Why regulated organisations use it
Approval gates make automation auditable and keep accountability with people. They are also a form of human oversight expected for high-risk uses under the EU AI Act.
Typical controls
- Approval gates at defined steps.
- Confidence thresholds that route uncertain cases to people.
- A full log of drafts, approvers and decisions.
See AI automation.