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

# AI feedback

> Flag AI messages and actions to improve accuracy and identify areas for improvement

When an AI-generated message or action is incorrect or could be improved, you
can flag it with feedback. This helps improve the AI Operator over time and
allows admins to identify patterns in the
[Insights Hub](/documentation/insights-hub/overview).

This is different from
[conversation rating](/documentation/tickets/conversation-rating), which scores
the whole conversation from 1 to 5 rather than flagging a specific message or
action.

***

## What can be flagged

You can flag both:

* **AI messages**: Replies sent by the AI Operator to customers
* **AI actions**: Actions performed by the AI Operator (e.g., updating records,
  processing requests)

***

## How to flag

<Steps>
  <Step title="Find the AI message or action">
    In the ticket conversation, locate the AI-generated message or action event
    you want to flag.
  </Step>

  <Step title="Click the flag button">
    Click the flag icon that appears when you hover over the item.
  </Step>

  <Step title="Select feedback reasons">
    Choose one or more reasons that describe what was wrong.
  </Step>

  <Step title="Add additional context (optional)">
    Provide any additional details that might help understand the issue.
  </Step>

  <Step title="Submit feedback">Click **Submit** to save your feedback.</Step>
</Steps>

***

## Feedback options

| Option                                               | When to use                                                  |
| ---------------------------------------------------- | ------------------------------------------------------------ |
| **Answer not relevant**                              | The response doesn't address the customer's question         |
| **Answer included incorrect information**            | The response contains factually wrong information            |
| **Answer included confidential or internal content** | Internal-only information was shared with the customer       |
| **Wrong or incorrect writing style**                 | The tone or style doesn't match your brand guidelines        |
| **Did not follow instructions**                      | The AI ignored specific instructions from your Specialists   |
| **Wrong language**                                   | The response was in the wrong language                       |
| **An action should have been performed instead**     | The AI should have taken an action rather than just replying |
| **Other**                                            | Use this for issues not covered by the other options         |

***

## How feedback is used

Your feedback serves two purposes:

1. **Improves the AI Operator**: Feedback helps identify patterns and improve AI
   accuracy over time
2. **Identifies weaknesses**: Flagged content appears in the
   [Insights Hub](/documentation/insights-hub/overview) under "Flagged content",
   grouped by Messages, Actions, and Cases

<Tip>
  Review the flagged content regularly in Insights Hub to identify patterns. If
  you see the same issues repeatedly, consider updating your
  [Specialists](/documentation/behavior/specialists) or [Knowledge
  sources](/documentation/knowledge/overview).
</Tip>

***

## Impact on billing

Flagging any message or action in a ticket immediately marks the entire ticket
as **"Not resolved by AI"**. This means the ticket will not be billed as a
resolution.

For details on resolution types and metering, see
[Resolution status](/documentation/tickets/resolution-status) and
[Usage & limits](/documentation/billing/usage-and-limits#how-resolutions-are-counted).

***

## See also

* [Conversation rating](/documentation/tickets/conversation-rating): Rate the
  overall quality of a conversation (1–5)
* [Insights Hub](/documentation/insights-hub/overview): View flagged content
  analytics
* [Resolution status](/documentation/tickets/resolution-status): How resolution
  affects billing
* [Usage & limits](/documentation/billing/usage-and-limits): How resolutions are
  metered in Billing
* [Handling wrong AI answers](/guides/best-practices/handling-wrong-ai-answers):
  Tips for improving AI accuracy