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

# Test Your AI Agent

> Preview and refine your AI agent before going live

Test your AI agent in a live chat preview before deploying to customers. See exactly how your agent responds and refine its training until you're satisfied.

## The Test Console

Every AI agent includes a dedicated test console that simulates the customer experience. Use it to:

* Preview responses in real-time
* Identify gaps in your training data
* Verify actions work correctly
* Fine-tune answers before going live

## How to Test

1. Navigate to **AI Agent** → **Test**
2. Type a message as if you were a customer
3. Review the AI agent's response
4. If the response needs improvement, update your [training data](/ai-agent/train/overview)
5. Repeat until you're satisfied with the results

## What to Test

<CardGroup cols={2}>
  <Card title="Common Questions" icon="comments">
    Ask the questions your customers ask most often. Verify answers are accurate and helpful.
  </Card>

  <Card title="Edge Cases" icon="code-branch">
    Try unusual or complex questions to see how the AI handles uncertainty.
  </Card>

  <Card title="Actions" icon="bolt">
    If you've configured actions, test that they trigger correctly and return expected results.
  </Card>

  <Card title="Conversation Flow" icon="arrows-spin">
    Have multi-turn conversations to ensure the AI maintains context.
  </Card>
</CardGroup>

## Improving Responses

When you find a response that needs work:

<Steps>
  <Step title="Identify the issue">
    Is the answer wrong, incomplete, or poorly phrased?
  </Step>

  <Step title="Update training data">
    Add or edit content in [Files](/ai-agent/train/files), [Text](/ai-agent/train/text), or [Q\&A](/ai-agent/train/qna).
  </Step>

  <Step title="Re-test">
    Ask the same question again to verify the improvement.
  </Step>
</Steps>

<Tip>
  For critical questions, add a [Q\&A pair](/ai-agent/train/qna) to ensure the AI gives the exact answer you want.
</Tip>