Skip to main content

Create AI model

You can create AI models in Atlan to govern and manage AI applications. AI models help you track model versions, configure training datasets, and maintain compliance with your organization's AI policies.

Prerequisites​

Before you begin, make sure you have:

  • Admin permissions in Atlan. Only admins can create AI apps and configure their settings. Alternatively, if a member has access to an AI policy persona with appropriate permissions, they can create and manage AI models and apps. Learn more about admin roles and AI policy.

Enable AI governance​

Before you can create or manage AI models, you need to enable AI Governance in your Atlan workspace.

  1. In your Atlan workspace, click the Admin icon in the left sidebar.

  2. Navigate to the Labs section.

  3. Locate the AI Governance toggle and enable it. If you don't see the AI Governance toggle, contact Atlan Support.

Once enabled, the AI Governance features are now available in your workspace.

Create AI model​

Once AI Governance is enabled, you can create your AI model in Atlan. The creation process registers both the model and its first version together.

  1. In your Atlan workspace, open AI from the left sidebar.

  2. From the top right corner, click New and select AI model.

  3. Enter a Name for your model (for example, Customer Churn Predictor) and a Description of what it does.

  4. Select one or more Owners to manage the model.

  5. Set the Model status (for example, Active).

  6. In the Version details section, enter the first version's information:

    • Version name: A name for this version.
    • Stage: Select None, Staging, Production, or Archived.
    • Description: What this version does.
    • Metrics: Add key-value pairs like accuracy: 0.95 or f1_score: 0.88.

  7. In the Select datasets section, choose the data assets this version uses. Mark them as inputs, outputs, or other types to track lineage.

  8. In the Additional details section, add custom metadata if configured.

  9. Review your model and version configuration.

  10. Click Publish to create the model and version. If a New Entity Creation workflow is active, the model publishes after workflow approval.

Once published, your AI model appears under the AI section of your workspace. You can view versions, manage their lifecycle, and connect them to AI apps as part of your AI governance framework.

Create additional model versions​

After an AI model is published, you can add more versions to track iterations over time.

  1. Open the AI model's profile page.

  2. Navigate to the Versions tab.

  3. Click Create version to open the version creation modal.

  4. Fill in the version details:

    • Version name (required)
    • Stage: Select the deployment lifecycle stage (None, Staging, Production, or Archived).
    • Description: Describe this version's changes or purpose.
    • Metrics: Add key-value pairs for performance and evaluation metrics.

  5. Optionally, select datasets to link to this version and add custom metadata.

  6. Review and Publish the version.

See also​

  • AI policy: Learn how to configure AI policies to control access to AI models and apps.
  • Create AI app: Learn how to create AI applications using your models.