Glossary · AI engineering and governance
Human-on-the-loop
Also known as: HOTL, Supervisory control
German: Menschliche Überwachung der Automatisierung
In industrial AI and automation, human-on-the-loop is an operating mode in which the system makes and executes decisions autonomously while a person monitors its behavior and can intervene, override or stop it when necessary.
- Industrial AI
- AI
In one sentence
Human-on-the-loop means the system acts autonomously while a person monitors it and can intervene, override or stop it.
Example
An AI system adjusts conveyor speeds continuously; the operator watches a dashboard of its decisions and can switch the line back to fixed settings.
How it applies
- Engineering: Supervision only works if the person can see what the system is doing and why, and has an effective, tested way to intervene, such as switching to manual mode or a fallback strategy.
- Operation: Long periods without incidents reduce attention. Alarms and summaries should direct the operator to situations that need a decision.
- Documentation: Describe what the operator monitors, which signs indicate a problem, how to intervene and what state the system goes to after an override.
Human-on-the-loop vs. human-in-the-loop
In Human-in-the-loop (industrial AI), each decision waits for human approval. In human-on-the-loop, decisions take effect immediately and the human supervises. The choice depends on consequences, time constraints and how reliably a human can intervene in time.
Supervision of an AI function does not replace safety functions; hazards still need protective measures according to the machine's risk assessment.