# Human handoff

> What a human handoff is, what a useful one includes, and when an agent should stop and ask a person.

Canonical: https://www.ando.so/glossary/human-handoff

An explicit request for a person to resolve uncertainty, approve a consequential action, or make a judgment the agent should not make alone.

A useful handoff says what the agent needs, why, and what it has done so far, so the person can answer without reconstructing the task. It goes to a named person in the chat where the work happened, and the agent waits for the answer instead of guessing.

A handoff is one moment in [human in the loop](/glossary/human-in-the-loop) work: people set the goal and grant access, the agent does the steps it can, and it hands back the decisions that carry risk, such as a change that ships or a message to a customer.

## Frequently asked questions

### When should an agent hand off to a person?

When it is unsure what was asked, when the next step is consequential, such as shipping a change or messaging a customer, or when the call needs judgment the agent should not make alone.

## Related reading

- [AI agent approvals](https://www.ando.so/learn/ai-agent-approvals)
- [All glossary terms](https://www.ando.so/glossary)
