From On Prem SQL to Natural Language Queries with Fabric Mirroring & Data Agents

In this video, I walk through how to expose on prem SQL Server data through a natural language interface using Microsoft Fabric Mirroring and OneLake—no ETL pipelines, no refresh schedules, and no duplicate copies to manage. We start with a traditional SQL Server 2022 on prem instance and mirror a small set of AdventureWorks tables into Fabric using Change Data Capture (CDC) and the On Premises Data Gateway. From there, the data lands automatically in OneLake as Delta tables, ready for analytics and AI. Once the data is mirrored, I build a Fabric Data Agent and show how users can ask plain English questions like “Number of orders and total sales by territory for fiscal year 2012”—and have Fabric handle the SQL generation, execution, and formatting behind the scenes. What you’ll see in this demo: • How Fabric Mirroring works for on prem SQL Server • When and why the On Premises Data Gateway is required • Enabling CDC for SQL Server mirroring • Creating secure gateway connections into Fabric • Mirroring only the tables you need into OneLake • Monitoring near real time replication • Using a Fabric Data Agent for natural language queries • Reformatting results with plain English—no SQL rewrites The goal here is simple: turn existing SQL Server data into something your users, Copilot, and AI agents can query naturally—without re architecting your environment. If you’re exploring Fabric as an analytics or AI layer over existing relational systems, this is one of the cleanest on ramps. 👍 If this was helpful, hit Like 🔔 Subscribe for more Fabric, SQL, and AI architecture walkthroughs 💬 Drop questions or follow up ideas in the comments Link to On Premise Data Gateway installation docs: https://learn.microsoft.com/en-us/fabric/data-factory/how-to-access-on-premises-data 0:00 Introduction 0:57 Configuring CDC on SQL Server 2022 1:50 Create a Fabric SQL User 2:14 Configure On Prem Data Gateway 3:59 Connect Fabric to SQL 5:01 Create a Workspace 5:20 Begin Mirroring Data 6:25 Create a Data Agent