Human-in-the-loop
Human-in-the-loop describes an automated process that pauses at a defined point for a person to approve, correct or reject the system's proposed action before it takes effect.
The distinguishing feature is where the pause is
Every automation involves people somewhere. What makes a design human-in-the-loop is that the pause is a defined step in the process rather than an informal habit: the system prepares the work, a named role reviews it, and the action does not proceed until that review happens.
Why a review step is added
Two reasons, and they are different. The first is consequence: some actions are expensive or impossible to reverse, so a person confirms them regardless of how reliable the system is. The second is confidence: the system is uncertain about this particular case, so it escalates rather than guesses.
How it changes over time
A review step is not permanent by definition. Where the reviewed decisions are logged, the record shows which categories the system gets right consistently and which it does not, and the boundary can be moved deliberately on that evidence rather than on optimism.