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

# Task automation

> How StageWhisper detects, approves, and dispatches tasks to your AI assistant.

During a listening session, StageWhisper's AI doesn't just coach you. It also identifies moments where an action should happen: sending a follow-up email, logging something in your CRM, scheduling a meeting, or creating a task for someone on your team.

These actionable moments become tasks that flow to your connected AI assistant.

## How tasks are created

The AI monitors the conversation against your Playbook and looks for actionable patterns:

* A prospect asks you to send them something ("Can you share that case study?")
* A commitment is made ("Let's schedule a follow-up for next Thursday")
* A piece of information should be recorded ("Our budget is around 50k")
* A follow-up action is implied by the conversation flow

When the AI identifies one of these moments, it creates a task with a description, the relevant transcript context, and a suggested action.

## Approval flow

By default, tasks go through review before being sent to your assistant.

### Review-first (default)

Each task appears in the Assistant tab of your session panel during the call. You see:

* What the task is
* Why the AI created it (the trigger)
* The transcript excerpt that prompted it

You can approve or dismiss each task individually. Approved tasks are sent to your assistant immediately. Dismissed tasks are discarded.

This mode gives you full control. Nothing leaves StageWhisper without your explicit approval.

### Auto-dispatch

For task types you've learned to trust, you can enable auto-dispatch. Tasks matching your allowlist are sent to the assistant automatically, without interrupting your call for approval.

Configure auto-dispatch in the assistant settings. You can set it per task type, so high-stakes actions still require review while routine ones go through automatically.

## Task lifecycle

Each task moves through a lifecycle:

| Status           | What it means                                             |
| ---------------- | --------------------------------------------------------- |
| Created          | The AI identified an actionable moment and created a task |
| Pending approval | Waiting for you to approve or dismiss (review-first mode) |
| Approved         | You approved the task; it's being sent to your assistant  |
| Executing        | Your assistant is working on the task                     |
| Completed        | The assistant finished the task                           |
| Dismissed        | You chose not to send this task                           |

You can see the status of all tasks in the Assistant tab during and after a session.

## What makes a good task Playbook

The AI's ability to detect tasks depends on your Playbook. If you want task automation to work well, include guidance about what kinds of actions should be created:

```text theme={null}
Create tasks when:
- The prospect asks me to send them any materials or documents
- We agree on a specific next step with a date or deadline
- I learn something about their org structure, budget, or timeline
  that should go into the CRM
- They mention a pain point that maps to a specific feature
  we should demo later
```

Without this kind of guidance, the AI will still detect obvious actionable moments, but it won't catch the domain-specific ones that matter for your workflow.

## Data shared with your assistant

Each task includes:

* Task description and type
* Relevant transcript excerpt (usually 2-5 sentences of context)
* Signal summary
* Playbook context

Tasks do not include raw audio, full session transcripts, or tone data beyond what's directly relevant to the task. See [concepts](/agents/concepts) for the complete data boundary.