Thank you for your feedback.
Form temporarily unavailable. Please try again or contact to submit your comments.

Create an NLU model

Log in to subscribe to topics and get notified when content changes.

Create an NLU model

Create an NLU model that the system uses to recognize and process user utterances, intentions (intents), and objects of, or contexts for, an action (entities). Train the model dataset iteratively using utterance examples so that the system predicts the optimal intent output for a new utterance.

Before you begin

  • Make sure that the NLU Model Builder - Core plugin, NLU Model Builder plugin, and Predictive Intelligence plugin are all installed and activated.
  • Role required: Admin or Delegated Developer role (with permission of All File Types)

About this task

To create an NLU model, you follow these high-level steps.
  1. Create a custom application record for your NLU model in Guided Application Creator.
  2. Use the NLU model builder to create, train, test, save, and publish your NLU model in ServiceNow Studio.

In this first-access scenario, you haven't previously built anything in Studio, so the system directs you to Guided Application Creator first. After you establish an application record of your model there, you're redirected back to Studio, where you continue to build your model iteratively until it's ready for publication.

In this example procedure, you're building an NLU model to help the system understand human-expressed intent regarding user requests for access to systems, data, roles, equipment, and other entities.

Note: As you build your NLU model and its component intents, utterances, and entities, make sure that you click the Train button so that your update is validated and captured in the model. See Train and test your NLU model.


  1. Navigate to System Applications > Studio.
    You're temporarily routed to Guided Application Creator.
  2. Click Let's get started.
    This image shows the Guided Application Creator landing page where you're guided to click the 'Let's get started' button. When you hover over the 'Can I still use Studio link?', a message appears confirming that you can still use Studio if you click a link on the next page of Guided Application Creator.
  3. In the Let's get started on your new app page, fill in these fields.
    Name Enter a unique name for your NLU model. In this example scenario, you enter NLU for Access Requests. You'll need to reenter this name again later in this procedure. Both names should match.
    Description (Optional) Summarize your model's purpose. For example, you could say something like This model helps the system to understand human-expressed intent and actions regarding user access requests.
    Advanced settings As you enter your model's name, this field, which designates scope, automatically populates with a system-assigned name that's similar to your Name value. The default scope for custom applications is Global.
  4. Click Create.
  5. In the Let's create some roles for your app page, ignore the Continue button.
  6. Click Continue in Studio (Advanced).
    This image guides you to ignore the Continue button and instead click the Continue in Studio (Advance) link, which takes you to Studio.
    The Welcome to Studio page appears.
  7. Above the Application Explorer, click Create Application File.
    This image guides you to click the Create Application File button on the Welcome to Studio landing page. When you click the button, the Create Application File window appears.
  8. In the Create Application File window, enter NLU in the search filter.
    The NLU Model application file type appears, associated with the Natural Language Understanding category.
  9. Click Create.
    This image shows you where to create your NLU Model: under the Natural Language Understanding category.
  10. In the Create NLU Model window, enter a unique name for your new NLU model in the Model Name field.
    In this example scenario, you enter the same name you entered in Step 3 of this procedure: NLU for Access Requests. You choose that name as it represents a grouping of language around a particular subject that you want the system to learn and understand.
  11. In the Confidence Threshold(%) choice list, select the minimum confidence score above which you want the model to return its predictions. For example, if the threshold is 60, the model returns the top intent predictions from those with a score that's 61% or higher. You can use the default threshold score of 60 or you can choose a value that's higher or lower.
    This image shows the Create NLU Model screen, where you name the model and use the default Confidence Threshold(%) value of 60 or set it to a higher or lower value between 0-101.
    Note: When you test your NLU model prediction later, you can return to this screen and set a higher or lower confidence threshold.
  12. Click Save.
    Your new model appears in the Application Explorer and the NLU Model screen, including sections where you can configure your model intents, entities, and synonyms.
    This image shows you how the system displays your newly created model on the NLU Model tab after you've saved it in Studio.
    Going forward, each time you access Studio, your model appears in a list of NLU models.
    This image shows you what you see the next time and all subsequent times that you access Studio: your NLU model appears in a list with other NLU models.
  13. (Optional) To change your NLU model name, click the Properties button to make your edit and click Save. To discard your NLU model draft and start again, click Delete.

What to do next

Create one or more intents for your NLU model per the instructions in Create an NLU intent.