Almost every AI governance deck claims a human in the loop. Few survive an inspector asking the operator what they actually did. Here is the difference.
I have read a lot of AI governance decks. Almost all of them contain a slide with a small human icon next to an arrow, and the phrase "human in the loop". Almost none of them would survive an inspector walking past the slide, finding the person behind the icon, and asking one question: what did you actually do on this case?
That question is the whole subject. Oversight is not a diagram, it is a set of conditions that either hold at the moment of the decision or do not: the reviewer has the authority to override, the time to review, the information needed to judge, the competence to interpret it, and a record proving they did. Remove any one of the five and you have decorative oversight, which is worse than none, because it manufactures a false assurance and, as we will see, a documentary trail against you.