Microsoft Purview Data Governance
Microsoft Purview is the company’s platform for data governance. In this blog, we discuss the use of Microsoft Purview to catalog AI use cases. We also demonstrate the use of Tavro’s AI Agents to auto-classify AI use cases based on the EU AI Act.
Cataloging AI Use Cases in Microsoft Purview
As shown in the figure above, we cataloged a Benefits Assistant Chatbot in Purview as an AI use case with the following custom attributes:
- Category of AI Use Case (Internal, External)
- Complexity (Low, Medium, High)
- Detail on the use case
- Name of the use case
- Pattern (e.g. extend an existing LLM)
- Sponsoring department
- Type (e.g., process efficiency)
- Use Case ID
- Vendor names
Tavro AI Agent for EU AI Act Risk Classification
The EU AI Act requires organizations to classify AI use cases as Prohibited, High, or Other. The risk classification drives various downstream activities. Use cases with a risk classification of “other” require minimal compliance. However, use cases with a “high” classification need rigorous AI governance compliance and documentation. The Tavro AI Agent auto-classified the Benefits Assistant Chatbot as High Risk based on Article 5 & 6 of the EU AI Act (see Figure above).
Tavro Agent Generated Risk Classification in Seconds and Significantly Reduced Cost
For a fraction of the manual effort, the Tavro Agent can improve the productivity of the AI governance leader. For organizations with hundreds of AI use cases, this can result in meaningful productivity improvements.
Human-in-the-Loop (HITL)
The AI governance lead still acts as the Human-in-the-Loop (HITL) and provides oversight for the results of the Tavro AI Agent.
