Explicit Measures Podcast 494: MCP Your Fabric Data
Jan 16, 2026
Mike and Tommy dive into the world of Microsoft's MCP servers and AI agents for Power BI and Fabric. They discuss whether teams should quickly embrace these integrations or proceed with caution. The hosts highlight the critical balance between agent capabilities and the need for human oversight. Listeners will learn how agents can enhance workflows by catching exceptions and organizing data effectively. Security risks and the importance of pilot testing are key concerns, making this an essential listen for anyone involved in data management.
AI Snips
Chapters
Transcript
Episode notes
Agents Become Actionable Model Editors
- MCP servers give AI agents direct tools to inspect and modify semantic models, not just retrieve information.
- This moves agents from passive search helpers to actionable collaborators that can run DAX, edit measures, and change model metadata.
Govern Agents Deliberately, Not Immediately
- Treat agents as tools that multiply outcomes but still need human judgment and governance.
- Design agent deployments with memory, exception handling, and economic tradeoffs in mind before wide rollout.
Use Agents For Creation, Not Continuous Runtime
- Agents excel at creation tasks where they generate artifacts once, not at repeated runtime costs.
- Build once with an agent, then deploy efficient runtime artifacts to avoid continuous agent expense.