What is human in the loop?

Human in the loop means a person reviews and approves each AI output before it is sent or acted on. The AI drafts the reply, report or recommendation, and a person checks it and signs it off. In the lighter human on the loop setup, the AI acts and a person checks a sample afterwards. In human out of the loop, software checks every output and holds back what fails.

Why it matters for a business

Human in the loop is the safest of the 3 arrangements and the slowest, because a person stays in every single loop: the typing goes, the judging stays. It fits where one uncaught mistake would be expensive or impossible to undo, such as paying a supplier or rejecting a candidate. Its protection is real only while the reading is real. Someone approving a draft every 3 seconds is not reviewing anything, and the process looks the same while the protection is gone.

Example

Let’s say a support team of 6 answers 400 emails a day with AI drafting every reply. Each person reads their queue, fixes what is wrong and sends. A draft with a wrong delivery date costs one correction, because someone was always going to read it first.

Go deeper

Human in the loop and guardrails: how much checking AI needs

Related terms

• AI guardrails
• AI governance
• AI agent
• Responsible AI

All terms, A to Z