Power BI: PBIP, PBIR & Agentic AI

Key insights

• PBIP, PBIR and TMDL: PBIP is a folder-based Power BI project format that stores report and model definitions as plain-text files; PBIR holds report definitions and TMDL defines semantic model metadata.
  
• Agentic AI workflow: Microsoft bundles curated Skills and tools so AI coding agents (for example, GitHub Copilot) can read, modify, and validate PBIP projects using natural language.
  
• Platform stack: authoring uses a combination of the MCP Server (authoring), Skills, the PBIP file format and the IPC Bridge so agents can perform deterministic edits inside Power BI and Fabric.
  
• What changed: Microsoft is shifting report authoring from UI-only editing to open, file-based, agent-readable development, and PBIP/PBIR are moving to GA as the default project format.
  
• Key benefits: expect reliable AI authoring, better Git workflows, end-to-end automation for model and report creation, and simpler natural-language development that reduces repetitive work.
  
• Practical limits and next steps: current tooling targets PBIP/PBIR-based projects (not legacy PBIX workflows), the capability is a curated capability set for agents rather than a simple toggle, and teams can create custom skills and naming standards as agent support expands.
  

Keywords

PBIP, PBIR, agentic AI Power BI, Power BI Fabric, Fabric Insider Ep 14, agentic AI development, Power BI automation, AI-driven analytics