Capture AI model data

The license package Enterprise Portfolio Management is required to work with AI models. The use case AI Portfolio Management must be activated.

An AI model is a program trained on data to recognize patterns, make predictions, or perform tasks autonomously without needing explicit human instructions for each decision. Typical AI models are Chat GPT, Claude, or Gemini. An AI model provides various AI technologies.

  1. Go to AI Portfolio Management > AI Models and click New > AI Model. Define a unique name and basic attributes and click OK.

    Per default, the data workbench displays only a set of basic attributes. You can add more columns to capture other attributes directly in the data workbench or you can navigate to an AI feature's content area and define it in more detail there.

  2. Specify data in the data workbench or click the Navigate  Navigate button next to an AI model to open its content area. Specify the AI feature's attributes as well as the relationships that the AI model has to other assets in the repository.

Try to capture as much information as possible about the AI model because complete data considerably improves the results of business questions and other analytics.

Go to the AI model's content area > Overview.

Define the AI model's basic data. All mandatory fields must be defined to create the AI model and save it.

  • Name: (Mandatory) Enter a unique name for the AI model. The name should help others easily understand the purpose of the AI model.
  • Status: This is an approval status and typically indicates the level of quality of the information about the AI model. Possible values are:
    • Draft: The AI model has only mandatory data defined.
    • Under Review: The AI model is documented and being reviewed. An AI model with this release status cannot be deleted.
    • Approved: The AI model has been approved by the responsible stakeholders. An AI model cannot be deleted when it has an approved release status. An AI model with this release status cannot be deleted.
    • Declined: The AI model has not been approved.
    • Retired: The AI model is no longer valid.
  • Start Date and End Date: Specify the start and end date of the use of the AI model.
  • Model Type: Specify the AI model used to recognize patterns, make predictions, or make decisions.
  • Vendor: Specify the vendor that provides the AI model.
  • Description: Enter a meaningful description that will clarify the purpose of the AI model.

Specify the locations where the AI model is/is not allowed. Specify the locations where the AI model is permissible in the Geographic Limitations > Allowed Locations field. Specify where the AI model may not be used in the Disallowed Locations field.

Specify the AI features where the AI model is used in the Additional Details > Providing Application field..

Specify the vendor that provides the AI model in the Additional Details > Vendor field. In the Providing Component field, specify the component that supports the application providing the AI feature.

Specify which component provides the AI model in the Additional Details > Component field.

You can add multiple AI technologies to the AI model.

  1. Go to the AI model's content area > Overview > AI Technologies.
  2. Click the New plus sign button > Create AI technology for This AI Model. You can also select an AI technology from the repository via Add Existing AI technology for This Model.

You can specify determine whether the AI model is approved in terms of its performance, ethics, and data privacy issues and capture the details about the assessments. You can also document the AI model's known limitations and potential mitigations of such issues.

Go to the AI model's content area > Approvals page.

Assess the AI model's performance, ethics, and data privacy. Go to the use case's content area > Approvals. The Approval Responsibilities sections shows who is responsible for approving various aspects of the AI use case. Each responsible person should specify their aspect of the approval process in the Approvals section. The reasons for their decisions should be captured in the Approval Details section.

Capture known limitations for the AI model and if these can be mitigated. Document the AI model's weaknesses or flaws in the Known Limitations field and capture how to mitigate these issues in the Limitation Mitigation field.