The short answer
Human-in-the-loop means the AI prepares an action and a person approves it before it happens. It is the safest way to get value from AI on tasks that have consequences.
What Anchor does about it
In Anchor, suggestions such as a follow-up message or restock order are written as pending actions. You review, approve or reject; approving is what triggers the result.
Limits to know
- Approval adds a step. The trade-off is intentional: speed for low-risk reading, control for real-world actions.
Common questions
- Is there an audit trail?
- Pending actions and their decisions are recorded so you can see what was proposed and approved.
Related topics
Updated 2026-09-25. Something unclear or wrong? Tell us.