What is human in the loop (HITL)?
Human in the loop is a way of running AI in which a person reviews, approves or corrects the system’s work at set points, instead of leaving every decision to the software.
Why it matters
Some decisions carry too much risk to automate fully, such as large refunds, complaints or anything touching a customer’s health or money. Keeping a person involved at those points catches mistakes before they reach the customer. It also gives the team a steady stream of corrections, which show where the AI’s instructions or source material need to improve.
How it works
The team decides which actions the AI may take alone and which need a person. For the second group, the AI prepares the work, such as a draft reply or a proposed refund, and places it in a queue. A staff member approves, edits or rejects it. Low confidence, sensitive topics or a customer’s request for a person can trigger review as well.
In practice
An online shop lets its AI agent approve refunds up to a set amount. A customer asks for a larger refund on a damaged sofa. The agent gathers the order details and photographs, writes a recommended decision and creates a ticket for a supervisor. The supervisor approves it in one click, and the agent tells the customer and issues the refund.
