Glossary · AI engineering and governance
Human-in-the-loop (industrial AI)
Also known as: HITL
German: Mensch in der Schleife
In industrial AI and automation, human-in-the-loop is an operating mode in which the system's recommendations or actions take effect only after a person reviews and approves them, so that a human makes or confirms each decision.
- Industrial AI
- AI
In one sentence
Human-in-the-loop is an operating mode in which a person reviews and approves each AI recommendation or action before it takes effect.
Example
The AI proposes new controller setpoints for a dryer, but they are applied only after the shift engineer checks and confirms them on the HMI.
How it applies
- Engineering: Design the approval step so that the person has the information needed to judge the recommendation: context, expected effect and confidence. Otherwise approval becomes a formality.
- Operation: Human-in-the-loop suits decisions with significant consequences and low time pressure. With frequent requests, operators may approve without checking (automation bias).
- Documentation: Describe who approves, what they should check, how to reject or correct a proposal and how decisions are logged. This supports Human oversight (AI Act) where it applies.
Human-in-the-loop vs. human-on-the-loop
In human-in-the-loop, the system waits for approval. In Human-on-the-loop, the system acts on its own while a person supervises and can intervene. In Human-out-of-the-loop, no human takes part in individual decisions. The general principle for AI-generated content is described in Human in the loop.