Human-in-the-loop describes a system design where AI-generated output must be reviewed and approved by a person at defined points before it takes effect.
Why it's needed. AI produces plausible-sounding but incorrect content with some regularity.
Placing checkpoints. Where output goes external or has financial/legal weight.
Separation of responsibility. The business stakeholder owns technical content, procurement owns commercial terms.
Audit value. Every approval logged.