#sqlsaturday

Mar 21, 2026

From Power BI to Conversational AI: Building Data Agents That Scales

Your Power BI semantic models already contain trusted business logic, metrics, and relationships. What if those same models could answer questions in natural language across Teams, Excel, and M365 Copilot without rebuilding everything? In this session, you'll see the complete journey from a standard Power BI semantic model to a conversational data agent deployed across multiple surfaces. We'll walk through: Preparing your semantic model for AI: Adding descriptions, AI instructions, and verified answers directly in Power BI Desktop Creating Fabric Data Agents: Turning your semantic model into an intelligent assistant with proper guardrails Deploying across M365: Making your agent accessible in Excel, Teams, and M365 Copilot Through a live demonstration using a real semantic model, you'll see how to handle time intelligence, variance calculations, and business-specific definitions so the AI understands your metrics the way your users do. We'll also cover what actually works (and what doesn't) when moving from visual analytics to conversational experiences. This isn't about replacing Power BI reports. It's about extending the work you've already done into new interaction patterns where they add real value.