Building and Deploying Smarter Data Agents in Microsoft Fabric
Description
Learn how to build and deploy smarter data agents in Fabric. We’ll show how to add richer context, connect new data sources, and use the latest intelligent capabilities. You’ll also see how to enable your agent in MCP, Teams, M365, Foundry, and more—making insights available to users wherever they work.
Key Takeaways
- Fabric Data Agents combine natural language understanding with structured data access — bridging the gap between AI and enterprise analytics
- Deployment considerations: agents need clear instruction sets, well-documented schemas, and curated knowledge sources to avoid hallucinations
- Smarter agents use multiple data sources in concert — OneLake lakehouses, warehouses, KQL databases — with Fabric managing unified access
- Amir Jafari (Microsoft) brings platform-insider perspective on agent architecture; Misha Desai brings implementation experience from real deployments
- Key to agent quality: grounding the agent with domain-specific context, example Q&A pairs, and explicit constraints on what it should/shouldn't answer
- Production deployment requires monitoring agent queries, logging interactions, and building feedback loops to improve accuracy over time
- Agents integrate with Copilot Studio for enterprise distribution and with Teams for end-user access
My Notes
Action Items
- [ ]