Meet the New Fabric Data Agent (Preview)

The Fabric Data Agent is a conversational AI feature in Microsoft Fabric that lets users ask natural language questions about organizational data and get accurate answers from a coworker who actually reads documentation. It works across data in lakehouses, warehouses, Power BI semantic models, KQL databases, and Fabric ontologies. The mission is to make data exploration simple because SQL is brilliant but not everyone wants to write a CTE during a meeting.

To use it, you need paid Fabric capacity, the right tenant settings, and at least one supported data source such as a lakehouse, warehouse, semantic model, KQL database, or ontology. Azure OpenAI authentication is handled for you, which means no keys to lose, misplace, or accidentally commit to GitHub.

The Data Agent checks schemas in connected data sources using user permissions, interprets the question, picks the right data source, and decides whether to write SQL, DAX, or KQL. It switches languages faster than a multilingual tourist ordering breakfast. It validates the query, runs it in read only mode, and returns a clean human friendly answer. Users can view the underlying query for transparency much like checking the source code of a website you definitely did not build by copying Stack Overflow.

Creating and Improving a Data Agent

Creating a Data Agent is simple. You choose up to five data sources, select the tables to include, and stay focused to keep it accurate. To improve responses, you can provide instructions and example question and query pairs to help the agent understand your organization's vocabulary, logic, and preferred data sources. Configure and publish the Data Agent and users can interact with it instantly without needing a cape or a superhero origin story.

Copilot Extensions

The Fabric Data Agent extends into Microsoft 365 Copilot. In Teams, users can chat directly with the Data Agent by mentioning it in conversations and receive real time answers. In Copilot for Power BI, the agent supports deeper governed questions. In Copilot Studio, custom copilots can use the Fabric Data Agent as a connected knowledge source for multi agent solutions where the bot relies on enterprise answers instead of internet myths.

OneLake Matters

OneLake is the unified data foundation behind Fabric. It brings lakehouse tables, warehouse structures, semantic models, mirrored databases, and ontologies into one logical lake. The Data Agent reads from OneLake, giving consistent governed data without juggling connectors or random storage systems.

Improved Data Agent Capabilities

The Fabric Data Agent has leveled up with expanded data source support, tighter Microsoft 365 Copilot integration, public endpoint connectivity for developer scenarios, improved multi agent orchestration through Copilot Studio, better performance, stronger reliability, enhanced safety controls, improved natural language understanding, and expanded semantic reasoning powered by Fabric IQ ontologies. The agent is smarter and requires far fewer late-night patches.

The Fabric Data Agent delivers a conversational layer over enterprise data, lowers the barrier to insights, improves decision making, and brings data directly into Teams, Power BI, and custom copilots. It is an evolving capability that is reshaping how users interact with organizational data.

Create a Fabric data agent (preview) - Microsoft Fabric | Microsoft Learn