Human in the loop
Human in the loop means a person reviews, approves or corrects an AI system's work at defined points, rather than letting it act entirely on its own. Where the person sits is a design decision, made according to the cost of a mistake.
What it means
Not every step needs a person, and putting one everywhere defeats the point of automating. The useful approach is to decide, step by step, what a mistake would cost: a wrong suggestion someone can ignore needs no review, a payment or a message to a customer might.
Human review also produces the best evaluation data a team will get: every correction is a real case where the system was wrong, and a candidate for the evaluation set.
How we use it
Our method Deciding which parts are rules, which are a model, and where a person stays in the loop is one of the decisions we make in writing before any build, set out in how to scope an AI product.
Related terms
Published 28 September 2026. All terms