---
Title: "What Is Async Authorization for AI Agents?"
Url: "https://computer.io/answers/when-should-ai-agents-work-asynchronously"
Topic: "Agentic AI"
Description: "Async authorization lets an AI agent pause, request a person's approval out of band, then continue. When to use asynchronous agents and approvals."
---

# What is async authorization for AI agents, and when should agents work asynchronously?

Async authorization lets an AI agent pause a task that needs permission, send an approval request to the right person outside the conversation, such as by push notification, Slack or email, and continue only after approval. Agents should work asynchronously for long-running tasks, actions needing sign-off, and work that spans hours or several systems.

## Sync vs async agents

| | Synchronous | Asynchronous |
|---|---|---|
| User waits? | Yes | No |
| Typical task | Answer a question | Investigate, provision, reconcile |
| Approvals | Inline | Requested and awaited out of band |
| Duration | Seconds | Minutes to days |

## When to use async

- The task needs a manager or owner to approve (access, refunds over a limit).
- Steps depend on slow systems or other people.
- The agent works overnight or across time zones.
- Many items must be processed in bulk.

## Design tips

Make approval requests specific (who, what, why, risk), set expiry times, log the decision, and tell the requester the status.

## FAQ

### What is asynchronous authorization?

A pattern where an agent or app requests a user's approval through a separate channel and waits for the decision before acting.

### Why do AI agents need async authorization?

Because agents often act on behalf of people who are not watching in real time, and sensitive actions need explicit consent.

### Is async authorization the same as human in the loop?

It is one way to implement human-in-the-loop control for agents that run in the background.

## Related terms

- [Asynchronous authorization](https://computer.io/glossary/async-authorization)
- [Human in the loop (HITL)](https://computer.io/glossary/human-in-the-loop)
- [Autonomous agent](https://computer.io/glossary/autonomous-agent)

## How Computer helps

- [How computer works](https://computer.io/how-computer-works)
- [Agent studio](https://devrev.ai/agent-studio)

## Further reading

- [Case asynchronous ai agents](https://devrev.ai/blog/case-asynchronous-ai-agents)
- [Ai agent authorization](https://devrev.ai/blog/ai-agent-authorization)
- [Actions human in the loop](https://devrev.ai/blog/actions-human-in-the-loop)