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

# Autopilot

> The work Whappy does on your campaigns without being asked, and how much of it you allow.

Autopilot is the part of Whappy that works on your campaigns rather than on your leads. It can build a campaign, rehearse it against invented customers, watch how real conversations go, propose improvements, and — if you let it — apply them.

<Note>
  **Two different things are called "AI" in Whappy, and they do different jobs.**

  Your **AI assistant** talks to leads on WhatsApp. It follows the funnel you built, asks your questions, and books appointments. Everything under [Campaign](/campaign/context) configures it.

  **Autopilot** works on the campaign itself — building it, testing it, and improving it. It never talks to a lead. Everything on this tab configures Autopilot.
</Note>

## What Autopilot is made of

<CardGroup cols={2}>
  <Card title="Permissions" icon="shield-halved" href="/autopilot/trust-levels">
    Five levels controlling what Autopilot may do on your behalf, from read-only to running live campaigns.
  </Card>

  <Card title="Coach" icon="chart-line-up" href="/autopilot/coach">
    Grades finished conversations, spots repeated weaknesses, and proposes fixes backed by quotes from your own leads.
  </Card>

  <Card title="Rehearsal" icon="masks-theater" href="/autopilot/rehearsal">
    Runs six invented customers through your funnel before real ones arrive. Nothing is sent to anybody.
  </Card>

  <Card title="Weekly briefing" icon="newspaper" href="/autopilot/briefing">
    The two or three things from the past week you would not have known to look for.
  </Card>
</CardGroup>

## Nothing is on by default

A brand-new account starts with Autopilot at its lowest setting:

* Permissions are set to **Read only**. Autopilot can answer questions about your campaigns and change nothing.
* Coach is **Off** on every campaign. Conversations are still graded — that costs you nothing and builds up evidence — but no proposals are made.
* Rehearsal only runs when you press **Rehearse**.

You raise each of these when you are ready, one at a time.

## Permission is not the same as readiness

The two highest permission levels need more than your consent. You can grant **Launch**, but Autopilot stays at **Build** until your account has also demonstrated it is ready — three campaigns launched by hand without problems, and a rehearsal that passed.

This is deliberate. It means nobody reaches "the system messages real customers unattended" by clicking through a dialog on their first day. See [Permissions](/autopilot/trust-levels) for exactly what is required.

## How the pieces feed each other

Autopilot's parts are connected, and the connections are worth knowing:

* A **passing rehearsal** is one of the two things that unlocks the **Launch** permission level.
* **Coach** can only apply changes by itself when permissions are at **Operate** — the highest level.
* The **weekly briefing** reports what Autopilot did, including changes Coach kept and changes it reverted.
* If a watchdog notices something wrong — your WhatsApp sender quality falling, for instance — Autopilot is **paused automatically** back down to Build until you clear it.