> ## Documentation Index
> Fetch the complete documentation index at: https://docs.cogniagent.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Working without you

> Your AI coworker can start work on a timetable, keep going on a long job, come back to check on something, and answer people while you're away.

The point of a coworker isn't that it's fast while you watch. It's that things get done while you don't.

## The four ways

Each answers a different question about *when* work happens.

<CardGroup cols={2}>
  <Card title="Schedules" icon="calendar" href="/cowork/autonomy/schedules">
    Recurring work on a timetable. A morning briefing, a Friday report. **You decide, in advance.**
  </Card>

  <Card title="Keep working" icon="repeat" href="/cowork/autonomy/keep-working">
    Push through one long job in a single run instead of check-in-sized pieces. **You decide, in the moment.**
  </Card>

  <Card title="Self-scheduling" icon="alarm-clock" href="/cowork/capabilities/self-scheduling">
    Come back to a task later — when a reply lands, when a process finishes. **The coworker decides, mid-task.**
  </Card>

  <Card title="Message channels" icon="inbox" href="/cowork/capabilities/message-channels">
    Someone messages it in Slack or email and it just gets on with it. **Someone else starts it.**
  </Card>
</CardGroup>

## What stays true when you're not there

This is the part worth being clear about, because "autonomous" makes people nervous, and it should mean *unattended*, not *unchecked*.

**Approvals still pause.** Anything leaving your workspace waits for your yes, however long you take. A coworker working overnight doesn't send the email at 3am because you were asleep — it waits.

**Permission rules still apply.** Nothing is relaxed because nobody's watching.

**Everything is on the record.** Every step it took while you were away is in the conversation, in order, with a notice explaining why it woke up and what it did.

<Frame caption="A quiet line in the conversation saying why it picked the work up again. Expand it for the full account.">
  <img src="https://mintcdn.com/glorium/JBtd_griZWPnHu68/images/cowork/autonomy/overview/01-wake-notice.webp?fit=max&auto=format&n=JBtd_griZWPnHu68&q=85&s=1fd8a63bace58463d4b28868ba97715f" alt="A notice in the conversation explaining that the coworker resumed because a deploy finished" width="1088" height="78" data-path="images/cowork/autonomy/overview/01-wake-notice.webp" />
</Frame>

**It stops rather than guesses.** When it hits something it can't resolve, it waits for you instead of picking an answer.

## Building up to it

Nobody should hand a coworker unattended work on day one. A reasonable progression:

<Steps>
  <Step title="Watch it do the job">
    Run the task manually a few times and read what it does. You're learning where it's reliable.
  </Step>

  <Step title="Tighten the brief">
    Whatever you found yourself correcting goes into its [instructions](/cowork/coworkers/instructions) or a [skill](/cowork/skills/creating).
  </Step>

  <Step title="Set the boundaries">
    Write the [permission rules](/cowork/control/permission-rules) that matter — especially the "never" ones.
  </Step>

  <Step title="Schedule it">
    Now let it run on its own, and read the first few results properly.
  </Step>
</Steps>

<Tip>
  The step people skip is the third one. Ten minutes writing "never delete anything" and "never email outside the company" is what makes the rest comfortable.
</Tip>

## Good to know

<Note>
  A paused coworker doesn't do any of this — no schedules fire and no channels listen. It's the single switch that stops everything.
</Note>

<Note>
  Work done while you were away is normal work. It shows up in [your tasks](/cowork/chat/tasks) and on the coworker's Activity tab like anything else.
</Note>

## Common questions

<AccordionGroup>
  <Accordion title="Can it act on something important without me?" icon="shield-halved">
    Only where you've allowed it. The default pauses sends, changes and deletions for your approval — see [Approvals](/cowork/control/approvals).
  </Accordion>

  <Accordion title="How do I know what happened overnight?" icon="sun">
    The conversation has everything, and a notice explains why it woke up. Have it message you the summary with [Talking to people](/cowork/capabilities/talking-to-people) if you'd rather not go looking.
  </Accordion>

  <Accordion title="What if it goes wrong while I am away?" icon="triangle-exclamation">
    It stops and says so rather than pressing on. You'll find the task waiting with an explanation.
  </Accordion>

  <Accordion title="Does unattended work cost more?" icon="coins">
    Only in the sense that more work happens. Waiting is free — a coworker sitting between check-ins costs nothing.
  </Accordion>
</AccordionGroup>

## Next steps

<CardGroup cols={2}>
  <Card title="Schedules" icon="calendar" href="/cowork/autonomy/schedules">
    Work on a timetable.
  </Card>

  <Card title="Permission rules" icon="shield-halved" href="/cowork/control/permission-rules">
    Set the boundaries first.
  </Card>

  <Card title="Keep working" icon="repeat" href="/cowork/autonomy/keep-working">
    One long run.
  </Card>

  <Card title="How a coworker gets better" icon="arrow-trend-up" href="/cowork/autonomy/getting-better">
    Why it needs you less over time.
  </Card>
</CardGroup>
