> For the complete documentation index, see [llms.txt](https://docs.cekat.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.cekat.ai/en/ai-and-platform/ai-agent/evaluation-ai.md).

# Evaluation AI

This feature lets you correct and retrain your AI whenever an answer misses the mark or isn't relevant. By reviewing or revising the AI's replies directly, you help it learn to be more accurate next time.

1. **Open it from the Chat page**
   * Hover over a reply the AI sent.
   * Click the evaluation icon (next to the reply icon).

     <figure><img src="https://3693109682-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FZNZlvyOVkZXyd3pz2Cql%2Fuploads%2FWsRuzErPpNmM010I5QwH%2FScreenshot%202025-07-03%20at%202.06.32%E2%80%AFPM.png?alt=media&#x26;token=b9c11810-94e9-4e07-ba13-ad1581fa5274" alt=""><figcaption></figcaption></figure>
2. **The evaluation pop-up**
   * Once you click the icon, the AI message evaluation pop-up opens.
   * It's filled in automatically with the AI's reply and one customer message as starting context.

     <figure><img src="https://3693109682-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FZNZlvyOVkZXyd3pz2Cql%2Fuploads%2FFNvrS6lD4nZiWDFBskYT%2FScreenshot%202025-07-03%20at%202.07.33%E2%80%AFPM.png?alt=media&#x26;token=ec7e28ae-48aa-4e67-927c-f9217e749b9c" alt=""><figcaption></figcaption></figure>
3. **Adjust the conversation context**
   * If you need to, you can add more messages as context.
   * Click the icon in the top-left corner to show the conversation bubbles, then tick the messages you want to include.
4. **Edit the AI's answer**
   * Edit the AI's answer so it says what you think is right.
   * To split the answer into several bubbles, click **+** to add a new message line.
5. **Revert to the original answer (optional)**
   * Changed your mind? Click the **revert response** icon to bring back the AI's original answer.
6. **Save your changes**
   * When you're done editing, click **Save** to keep your revision.
7. **Manage your evaluations**

   * Go to **AI Agent > Setting > Evaluation**.
   * There you can see your full evaluation history, edit evaluations again, or delete ones you no longer need.

     <figure><img src="https://3693109682-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FZNZlvyOVkZXyd3pz2Cql%2Fuploads%2FYlkBYylYdeuKj8tfO5v4%2FScreenshot%202025-07-03%20at%202.07.57%E2%80%AFPM.png?alt=media&#x26;token=1a11e533-e6e9-4273-921d-79836fde0ce9" alt=""><figcaption></figcaption></figure>

   **Also:**

   You can evaluate replies from the **AI test box** on the **AI Agent Settings** page too, while you're testing the AI's answers yourself. It works the same way, so you can correct the AI's output the moment you spot something off.

   <figure><img src="https://3693109682-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FZNZlvyOVkZXyd3pz2Cql%2Fuploads%2FQevOUxLnp4FK7iqEAhzt%2FScreenshot%202025-07-03%20at%203.50.04%E2%80%AFPM.png?alt=media&#x26;token=08085694-aed6-4fed-8c53-a3a890c5508f" alt=""><figcaption></figcaption></figure>