Skip to main content

AI glossary Responsibility

What is human-in-the-loop (HITL) in AI?

← All glossary terms
In short

A design in which people review, approve or correct an AI system's outputs before they take effect, or step in when the system is unsure.

Explained

Human-in-the-loop (HITL) is a way of designing AI systems so that a person reviews, approves or corrects the output before it has an effect. The person may check every decision, only the uncertain or high-impact ones, or a sample. HITL is common in high-stakes decisions, such as credit, claims, hiring or medical support, and in agents that act on real systems. It also produces corrections that can be used to improve the model. The EU AI Act requires high-risk systems to be designed so that people can oversee them effectively.

In, on or over the loop

  • Human in the loop: a person approves or edits each output before it is used. Slower, but the safest for high-impact decisions.
  • Human on the loop: the system acts on its own and a person monitors it, with the ability to intervene or stop it.
  • Human over the loop: people set the rules, limits and thresholds, and review results periodically.

Many real systems mix the three: automatic handling for routine cases, human approval when confidence is low or the amount at stake is high, and regular review of the whole.

Where it is needed

  • Decisions with legal or significant effects on people, such as credit, insurance, employment or access to services.
  • AI agents that can send messages, pay, change records or deploy code.
  • New systems in their first weeks of production, before their behaviour is well known.
  • Cases where the model reports low confidence or the input falls outside what it was tested on.

Making oversight real

Human review only works if the reviewer can actually judge the output. That means showing the evidence behind a recommendation, keeping the workload manageable and making it easy to reject or correct.

The main risk is automation bias: people tend to accept what the system proposes, especially under time pressure. The EU AI Act names this risk explicitly. Track how often reviewers change outputs; if they never do, the review may not be working.

Frequently asked questions

What does human-in-the-loop mean?

It means a person is part of the decision process of an AI system. The system proposes, and a human reviews, approves or corrects before the result is used, or steps in when the system is unsure.

What is the difference between human-in-the-loop and human-on-the-loop?

In human-in-the-loop, a person approves outputs before they take effect. In human-on-the-loop, the system acts on its own while a person monitors it and can intervene or stop it. The second is faster; the first gives more control.

Does the EU AI Act require human-in-the-loop?

Article 14 requires high-risk AI systems to be designed so that natural persons can oversee them effectively while in use. That includes being able to understand the output, override it and stop the system. It does not always mean a person approves every single output; the measures must match the risk and context.

Does human-in-the-loop slow everything down?

It can if every output needs approval. Most designs route only uncertain, unusual or high-impact cases to people and let routine ones flow. The goal is to put human attention where it changes the outcome.

How does human-in-the-loop improve AI models?

Each correction a reviewer makes is a labelled example of what the right answer was. Collected over time, these corrections can be used to evaluate the system, adjust prompts or rules, and retrain or fine-tune the model.

Decision guides that use this term

How to put it to work

Where it shows up at Thinkia

Further reading

Sources

Updated: 2026-10-21

Want to see how this applies to your organisation?

Talk to an AI Expert