Configuring AI-powered incident categorization

Edit online

Configure the multi-class classifier (MCC) model to enable AI-powered incident categorization in the Tickets application. The MCC model analyzes incident descriptions to suggest appropriate incident types and categories.

Before you begin

Ensure that the AI service is available and running.

Prepare your training data by identifying historical incidents with accurate categorizations. For more information, see Preparing data for AI-powered incident categorization.

About this task

After AI-powered incident categorization is enabled, the recommendation feature is accessible to all users who have access to the Tickets application and whose incidents meet the inference filter criteria.

After a configuration is activated, to change settings, you must deactivate it, edit the configuration, and then activate it again.

You can follow these steps to activate the incident categorization configurations PLUSGINCIDENTGROUP and PLUSGINCTYPE.

Procedure

  1. In side navigation, open the Application administration > AI configuration application.
  2. Select PLUSGINCIDENTGROUP configuration.
  3. Review and edit the AI configuration.
    1. Click Actions > Edit.
    2. In the Template version field, ensure that the latest version is selected.
    3. In the Additional details for AI explained section, provide information specific to your organization to help users understand the model and its output.
      An AI icon appears alongside the model's output. Users can click the icon to access the information you specify here.
    4. Click Save.
  4. Optional: Set up arguments to adjust model behavior.
    You can configure threshold values and other parameters that affect how suggestions are generated and displayed.
  5. Click Actions > Check data requirement.
    The system reviews the training data to determine whether it contains sufficient incidents and categories. If the data check fails, add or improve the quality of data in your training filter.
  6. Click Actions > Activate.
    Activating the AI configuration indicates that the configuration is prepared and the model is ready to be trained.
  7. Click Actions > Train model.
    Training begins according to the configured schedule. Training can take several hours depending on the volume of data.

    Monitor training progress in the Model training log table or in the Model status dialog.

    When the Ready to inference field shows Ready, the AI feature is ready to use.

    The model accuracy score measures how the model performs on the training data. The closer to 1.0, the more accurate the output likely is based on the training data.

  8. Verify that AI suggestions are enabled for incidents that meet the inference filter criteria.
    Follow the same steps to configure PLUSGINCTYPE categorization.

What to do next

To retrain the model on new data or edit arguments, make the changes and then click Actions > Re-train model.

To change other configuration settings, deactivate the configuration first, edit it, activate it again, and then train the model.