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

# Train the AI agent

> How to test, rate, and give feedback to your AI agent using Quick chat to improve its responses

**Quick chat** is the tool to test your AI agent inside the Mindo platform: you simulate queries from a real customer and rate the responses with a feedback system. Every negative feedback generates an internal ticket that the Mindo team uses to iterate on the agent's prompt and apply your corrections.

## Demo video

<iframe width="100%" height="400" src="https://www.youtube.com/embed/E6BsoYZeCt8" frameBorder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture" allowFullScreen />

<Warning>
  Corrections you send with **Dislike** do not modify the agent automatically. Each feedback generates an internal ticket: the Mindo team adjusts the agent's prompt to apply your improvement.
</Warning>

## Steps to train the agent

<Steps>
  <Step title="Open Quick chat">
    From the main menu, go to **AI Agents**. You'll see the list of every agent assigned to your account. Open the agent you want to test and launch the **Quick chat** tool to start a new conversation.
  </Step>

  <Step title="Talk to the agent">
    Begin the interaction with a greeting or a typical query a real customer would send. Try different scenarios to evaluate how the agent responds in each case.
  </Step>

  <Step title="Rate every response">
    As the agent replies, rate each message with one of the two buttons:

    * **Like**: the response is correct, precise, and aligned with what you expected.
    * **Dislike**: the response has errors, the tone is off, or it doesn't address the intent of the query.
  </Step>

  <Step title="Leave descriptive feedback when rejecting">
    When you click **Dislike**, the system asks you for a reason. Write a clear comment that includes two things:

    1. Which part of the message you didn't like and why.
    2. An example of how the agent should have replied.

    <Note>
      The more specific your feedback is, the faster the Mindo team can iterate on the agent's prompt.
    </Note>
  </Step>

  <Step title="Track corrections from Feedback tickets">
    To keep track of your corrections, go to the **Feedback tickets** section. You'll see every comment you submitted, contrasting the agent's original response with the correction you proposed.

    If you want to add more context, you can post new comments inside an open ticket. The admin team can also use this space to ask follow-up questions about the correction and better understand your need.
  </Step>

  <Step title="Confirm resolution">
    Once the Mindo team adjusts the agent and resolves the issue, they leave a notice and the ticket moves automatically to the **Resolved** tab, leaving a complete history of every improvement applied.
  </Step>
</Steps>

<CardGroup cols={1}>
  <Card title="Go chat with the agent" icon="arrow-up-right-from-square" href="https://app.mindosoftware.com/dashboard/ai-agents-management">
    Test your AI agent from Quick chat on the Mindo platform.
  </Card>
</CardGroup>