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

# Set up AI-based categorization

> How to train and configure categorization in SuperOffice AI

Categorization uses machine learning to suggest categories for incoming requests based on their content. Before this feature can be used, the AI must be trained using a selection of existing requests.

<Note>
  Categorization requires the SuperOffice AI Platform and a **Categorization license**. SuperOffice AI is available for CRM Online only.
</Note>

## 1. Create a selection of requests

You must first create a selection of requests in SuperOffice to train the AI model.

1. Click <i class="ph ph-list-magnifying-glass" aria-label="Find icon" /> **Find** or go to **New** > **Selection**.

   ![New selection options in Sales client -screenshot][img1]

2. Choose **Request** as the selection type. Use **Find Request** or pick from **Typical searches**.

   ![Typical searches for request selection -screenshot][img2]

3. Define search criteria:

   1. Add filters to match the requests you want to use for training.
   2. Click **Find**.

   ![Request selection criteria -screenshot][img3]

4. Save the selection:

   1. Click **Save as selection**.
   2. Choose static or dynamic.
   3. Enter a name and click **Save**.

   ![Select selection type -screenshot][img4]
   ![Enter selection name -screenshot][img5]

<Note>
  Only the first 5000 requests in the selection will be analyzed.
</Note>

## 2. Choose the selection in Settings and maintenance

1. Open **Settings and maintenance**.
2. Select **AI services** from the Navigator.
3. Under **Categorization**, open the dropdown and search for the selection name.
4. Select the correct request selection.

   ![Choose request selection in AI services -screenshot][img6]

## 3. Analyze the request selection

Once a valid selection is chosen, the **Analyze requests** button becomes active.

<Note>
  Large datasets may take hours or even days to analyze.
</Note>

1. Click **Analyze requests** to start the training process.
2. When training is complete, the **Latest training** status will show *successfully trained model*.
3. If errors occur, review the message and create a new selection if needed.

   ![Training results and feedback -screenshot][img7]
   ![Training summary examples -screenshot][img8]

## 4. Enable suggestions

After successful training:

1. Select the checkbox **Use AI to suggest categories** to activate the feature.

   ![Checkbox to enable AI suggestions -screenshot][img9]

## 5. Result: category suggestions in requests

When enabled, users will see a suggested category in the request footer. Clicking the link applies the category to the request and removes the suggestion.

If there is no match, no suggestion will be shown.

## Re-train or change selection

You can re-analyze a new or updated selection at any time:

1. Go to **Settings and maintenance** > **AI services**.
2. Under **Categorization**, choose a new selection.
3. Click **Analyze requests** to train again.

[img1]: /media/loc/en/search-options/choose-new-selection.png

[img2]: /media/loc/en/search-options/typical-request-searches.png

[img3]: /media/loc/en/ai/find-request-search-criteria.png

[img4]: /media/loc/en/ai/select-selection-type-dialog.png

[img5]: /media/loc/en/ai/name-selection-input.png

[img6]: /media/loc/en/ai/select-category-menu.png

[img7]: /media/loc/en/ai/categorization-analysis.png

[img8]: /media/loc/en/ai/categorization-summary.png

[img9]: /media/loc/en/ai/use-ai-suggested-categories-checkbox.png


## Related topics

- [How to create a selection](/en/search-options/selection/learn/create.md)
- [Request categorization](/en/ai/learn/categorization.md)
- [Other AI services](/en/ai/learn/index.md)