Create a similar records prediction model in Task Intelligence for ITSM

  • Release version: Xanadu
  • Updated August 1, 2024
  • 2 minutes to read
    • Summarize
    Summarized using AI
    This content was generated using new OpenAI-powered functionality. Results are provided on an as is basis and are not guaranteed to be accurate or complete.

    Summary of Create a similar records prediction model in Task Intelligence for ITSM

    This guide explains how to create and manage a similar records prediction model within Task Intelligence for ITSM. The model compares fields between two types of tables—prediction and training tables—to identify and predict similar records, such as matching incident descriptions to relevant knowledge base articles. This helps agents quickly find pertinent information and resolve incidents more efficiently.

    Show full answer Show less

    Key Features

    • Customizable Prediction and Training Fields: Admins can select which fields to use for making predictions and training the model, tailoring it to the organization’s data.
    • Configurable Prediction Behavior: Options include enabling recommendations, running the model in monitor-only mode, or turning off predictions entirely.
    • Model Training and Testing: The model uses historical case records for training and allows admins to view test results to assess performance before deployment.
    • Base System Models Included: Out-of-the-box models for similar incidents, similar open change requests, and similar open problems come pre-installed and auto-deployed when the Task Intelligence for ITSM plugin is installed.
    • Performance Monitoring: Admins receive email notifications with analytics dashboard links and can review model performance in the Task Intelligence for ITSM Monitoring interface.
    • Easy Editing and Redeployment: Models can be modified from the Setup page and redeployed without manual configuration of rules in Recommended Actions for ITSM.

    Practical Benefits for ServiceNow Customers

    • Accelerates incident resolution by surfacing relevant historical or knowledge base records automatically.
    • Enables admins full control over what data is used, when predictions occur, and how results are presented to users.
    • Reduces manual effort in finding similar records, improving agent productivity and service quality.
    • Leverages automatic training on existing data, requiring minimal setup while providing actionable insights.
    • Supports continuous improvement through monitoring and easy adjustment of models based on performance metrics.

    Set up a training model to help it recognize similarities between two types of tables by comparing their fields.

    The model looks at the prediction fields of a prediction table and the training fields of a training table. It uses the similarities in these fields to predict similar records.

    For example, the prediction fields might be the short description and description of an incident record, while the training fields could be the short description and title of a knowledge base (KB) article.

    The model compares these fields to find KB articles that are similar to the current incident. This makes it easier for agents to find relevant information and resolve incidents quickly.

    If you're an admin, you can decide when you want it to make predictions, what you want it to predict, and what data to use for those predictions.

    You can select the following information: 
    • Prediction fields using which similar records are predicted for the model.
    • Training fields using which the similarities are compared to prediction fields to predict similar records.
    • Conditions used to choose the set of records to train the model.
    • The prediction behavior: Recommendation, Monitor only (run in the background)  or turn off predictions.

    After customizing your model, you can view the test results and deploy the model.

    The model uses previous case records to train. You can see how the model would have predicted similar records for past cases, providing you with an estimate of how the model performs when deployed.

    Base system model 

    Task Intelligence for ITSM comes with a base system which has following models based on similarity:
    • Similar Incidents model for predicting the similar incidents for incidents table.
    • Similar open Change Requests for Incidents model for predicting the similar change requests for incident table.
    • Similar open Problems for Incidents model for predicting the similar problems for incident table.
    When you install the Task Intelligence for ITSM plugin in your production instance, the model is installed and is trained on your data automatically. This model is auto-deployed. The model starts predicting data in the background and doesn’t impact the Incident form. 
    Note:
    As an admin user, you should receive an email within four weeks with a link to your analytics dashboard.

    The prediction data is then stored in the prediction tables. You can navigate to All > Task Intelligence for ITSM > Monitoring to view the performance results of this model. Based on these results, you can edit the model to show the predictions in Service Operations Workspace.

    To edit and redeploy the base system model, you can navigate to the Setup page. From the model list, select the required similarity model(For example, Similar Incidents) and edit the model. For more information on how to edit the model, see Edit an incident prediction model in Task Intelligence for ITSM.

    This model is linked with a base system rule in the Recommended Actions for ITSM configuration. An admin isn't required to go to the Recommended Actions for ITSM application to configure any rules for this model.

    To set up a new model using a template, see Set up similar records prediction model.