The human in the loop is the mechanism by which a person validates, corrects or approves the action proposed by the AI. It is particularly important in high-impact processes, where the decision may have economic, production, legal or reputational consequences.