Microsoft Fabric Data Agents introduce a new way to interact with data: asking questions in natural language and getting answers from a Lakehouse, a semantic model, an Eventhouse, or mirrored databases.
In this technical, demo-first session, we will explore:
- How does it work? The architecture, entry points, and a live overview of the feature in action.
- How do you shape agents? Using markdown, the spec-kit, and hands-on demos to understand how to guide, specialize, and control agent behavior.

As a bonus, we will conclude with a demonstration of MCP Servers, showing how a semantic model and DAX KPIs can be generated directly from a prompt.

You will leave with a clear understanding of what is possible today, the current limitations, and what you can start preparing now (Prep for AI) to fully benefit from this transformation.