Create dataset assets

Create AI assets to track and manage the life cycles of your datasets.

Before you begin

Role required: AI steward [sn_ai_governance_ai_steward] or AI asset owner [sn_ai_asset_mgmt.ai_asset_owner]

Note:

Users with the AI asset owner [sn_ai_asset_mgmt.ai_asset_owner] role can only create AI assets and submit them for life-cycle review. They can't start or complete any life-cycle reviews.

About this task

A Model Context Protocol (MCP) server is an external service or integration endpoint that AI agents connect to retrieve context, execute tools, or access data.

Procedure

1. Navigate to All > Al Control Tower > Home > Inventory.
2. On the Inventory page, select Add AI asset.
   The Add AI asset dialog box opens.
3. In the dialog box, select Enter asset details.
4. From the list of available asset types, select Dataset.
   The dialog box closes and you're automatically redirected to the Add dataset asset form.
5. On the form, fill in the fields.

   Table 1. Add dataset asset form

   Field	Description
   Name	Name of the dataset.
   Provider	People or organization that developed the dataset.
   Acceptable usage	Acceptable use for the dataset. You can specify more than one use.
   Version	Version of the dataset.
   Data type	Type of data that the dataset contains. You can specify more than one data type.
   Source	Source of the dataset, such as internal customer data within an organization or publicly available data from a government agency.
   Description	Brief description of the dataset.
   Asset state	Current state of the dataset. Select one of the following options: Design Build Available Deployed Note: A concatenation of asset state and asset status appears on each asset record page The available UI actions on the asset record page vary based on the current state and status.
   Asset status	Status of the dataset in a workflow. If this field is empty, that indicates the dataset is not going through any workflow. A value indicates if this dataset is going through a workflow, is approved or is rejected. Note: A concatenation of asset state and asset status appears on each asset record page The available UI actions on the asset record page vary based on the current state and status.
   Managed by	User who is assigned to manage the dataset. This field is automatically set to the user who creates the dataset asset. Note: This field is editable only if you have the AI steward [sn_ai_governance_ai_steward] role. If you have the AI asset owner [sn_ai_asset_mgmt.ai_asset_owner] role, this field is read-only.
   Creation type	Method by which the dataset was created. Select one of the following options: Curated: The dataset was created through careful selection, organization, and refinement of data. Derived: The dataset was created by processing, transforming, or combining data from existing datasets. Synthetic: The dataset was created by using artificial data that mimics real-word data.
   Base datasets	Base dataset that the dataset was built from.
   Department	Department that the dataset belongs to.
   Dataset creation date	Date and time that the dataset was created.
   Dataset card	Brief document that describes important information about the dataset, including the context, intended use, limitations, and potential biases of the dataset.

6. Select Save draft to save the details in the form or select Cancel that provides you with the following options:
   • Keep editing: to continue editing the form.
   • Save draft and exit: saves the form and you exit the form.
   • Discard and exit: your unsaved changes are discarded and you exit the form.
7. Select Submit for review.
   The Asset record page opens and your asset is in an unmanaged state. You can edit details related to your asset on the Details tab by using the pencil icon in the sections titled About this asset and Use and purpose.
8. If you have the AI steward [sn_ai_governance_ai_steward] role, select the Actions menu and then select Start lifecycle review.
   Important:

   Only users with the AI steward [sn_ai_governance_ai_steward] role can start the life-cycle review.
   The Start AI steward review dialog box opens.
9. In the Managed by field of the dialog box, search for and select the user that you want to assign the life-cycle review to.
   Important:

   You must select a user with the AI steward [sn_ai_governance_ai_steward] role.
10. Select Start review.

    The asset automatically changes from an unmanaged asset to a managed asset and starts the onboarding workflow. The Lifecycle tab now contains three sub tabs: Onboard, Maintain, and Retire. The Maintain, and Retire are disabled and are enabled once onboarding is complete.

    The Onboard sub tab contains the Onboarding playbook with tasks displayed. If tasks are not already present in the Onboarding playbook, users with the AI steward [sn_ai_governance_ai_steward] role can create tasks by selecting New and assign them to other AI stewards.

11. If you're assigned to the life-cycle review, select Mark complete to complete each activity in the onboarding playbook.

    If your tasks are open and you select Mark complete, those tasks automatically get closed with the Closed skipped state and you move to the next stage of onboarding. Once you move forward, you can no longer create tasks in the previous stage.

    You can select the Actions menu and Reject request to cancel the onboarding playbook and mark this asset as unmanaged.

    When all stages in the onboarding playbook are marked as complete, the playbook status updates to Completed. The Maintain and Retire tabs show a status of Not Started.

    Note:

    The end-to-end onboarding process from start to finish may take a few days.