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

# Run your first test conversation

> Validate replies in the Playground before going live.

Before an employee touches real customer traffic, run at least one conversation in the **Playground** — a live chat panel attached to every AI Employee card. Nothing sent here reaches a real channel.

<Tip>
  Select a **test contact** before sending a message. Without one, the employee has no contact data to read, share, or update, so any guideline that depends on it will not behave the way it does live.
</Tip>

## Start a test conversation

1. Open an employee in [**AI Employees**](https://app.heyy.io/ai-employees).
2. Open the **Playground** panel from the toolbar on the right (it opens full-screen on smaller windows).
3. Select a **test contact** from the dropdown.
4. Send a message as if you were that contact.

<Frame>
  <img src="https://mintcdn.com/heyy-c8bd5c9e/1coCVJUJuVs6SA9x/assets/ai-employees-playground-test.png?fit=max&auto=format&n=1coCVJUJuVs6SA9x&q=85&s=38024736512f2029c9305b03c32feafb" alt="Playground chat conversation with an AI employee replying in a playful tone" width="1570" height="774" data-path="assets/ai-employees-playground-test.png" />
</Frame>

The Playground uses whatever settings are currently saved across every tab, and the conversation persists as you switch between tabs — so you can adjust a guideline and continue the same thread without losing context.

## What to check

Review the reply against three things:

| Check           | What you are confirming                                                                                                        |
| --------------- | ------------------------------------------------------------------------------------------------------------------------------ |
| **Content**     | The reply draws on your actual [knowledge base](/help/ai-employees/content/add-your-website), not a generic answer.            |
| **Actions**     | Any [action](/help/ai-employees/capabilities/actions-overview) the message should trigger actually fires.                      |
| **Exit points** | A message that should resolve or escalate triggers the right [exit point](/help/ai-employees/capabilities/create-exit-points). |

Use **Reset chat** to clear the thread and start clean when testing a specific scenario from scratch.

## What's next

Run through the edge cases you are most worried about — a boundary you told it never to cross, a question outside its knowledge base — before moving on to [connect to real customer traffic](/help/ai-employees/getting-started/connect-to-real-customer-traffic). For the full testing reference, including how test contact data is used and what to verify before going live, see [Test in the Playground](/help/ai-employees/going-live/test-in-playground).