Human in the loop (HITL) is a design where a person reviews, corrects, or approves an automated system's output before it takes effect.
What HITL means
In a human-in-the-loop system, automation does the bulk of the work and a person makes the decisions that carry risk. The person might label data, correct a model, or approve an action. The term comes from control systems and machine learning, and it now describes most responsible uses of AI agents.
There are degrees:
- Human in the loop: nothing happens until a person approves.
- Human on the loop: the system acts, and a person watches and can stop it.
- Human out of the loop: fully automatic.
An example
An agent drafts a post announcing a pricing change. It lands in a review queue. The marketing lead fixes one number, approves it, and only then does it publish.
Why it matters for social publishing
A public post cannot be quietly recalled. Screenshots last. HITL puts a person at the one point where a mistake becomes permanent, without asking that person to write every post.
SocialHandoff is built around this step. Drafts from agents go to an approval workflow inbox on desktop or a swipe stack on your phone, where you approve, reject, request edits, or edit inline. Once an agent has a clean record, you can move it up the trust ladder so it publishes on specific networks without review, which is a move from "in the loop" to "on the loop." The approving posts guide shows the flow.