Human-on-the-Loop

A design paradigm where humans monitor AI system operations and can intervene when necessary, but the system operates autonomously by default.

The AI runs on its own, but a person watches over it like a supervisor and can step in if something goes wrong. Think of a factory floor manager overseeing automated machinery.

Understanding the spectrum of human oversight models helps organisations design proportionate governance. Human-on-the-loop is appropriate for moderate-risk AI applications where full human review of every decision is impractical.