Unlock natural language access to your enterprise data with Microsoft Fabric Data Agent.
Introduction
The Microsoft Fabric Data Agent is a powerful tool that enables you to interact with structured enterprise data using natural language queries. Instead of writing complex SQL, DAX, or KQL statements, you can simply ask a question and get an instant, data-driven answer — whether your data lives in a Lakehouse, Warehouse, Power BI model, or KQL database. This speeds up analysis and makes insights accessible to a wider audience without requiring deep technical expertise. By tailoring the Data Agent to your organization’s needs, you can unlock self-service analytics while maintaining governance and security.
[Image Source — Microsoft ]
Key Highlights & Limitations
· Supports read-only queries in SQL, DAX, and KQL.
· No support for create/update/delete operations.
· Access limited to configured structured sources.
· No support for unstructured data (e.g., PDF, DOCX).
· Optimized for simple queries; lower reliability with complex joins or logic.
· English only; LLM is fixed and cannot be changed.
Prerequisites
· A paid Fabric capacity (F2 or higher).
· Fabric data agent tenant setting enabled.
· Copilot tenant switch enabled.
· Cross-geo processing and storing for AI enabled.
· At least one of the following data sources: Lakehouse, Warehouse, Power BI semantic model, Kusto (KQL) database.
· Power BI semantic models via XMLA endpoints switch enabled.
How to Create Your Data Agent
1. Open your Fabric workspace and click + New Item.
2. In the All items tab, search for and select Fabric data agent.
3. When prompted, enter a name for your agent.
4. After naming it, the OneLake catalog appears — add data sources here.
5. Select up to five sources (Lakehouse, Warehouse, semantic models, KQL DB).
6. Use the filter icon to narrow down by type if needed.
7. Once sources are added, select the tables you want the AI to access.
Sharing Data Agent within MS Fabric
• Review your configuration to make sure the right tables and sources are included.
• Validate the agent, then share or publish it so colleagues can ask questions in natural language and get data-driven answers.
Integration: Fabric Data Agent → Azure AI Foundry
1. Publish your Fabric Data Agent in the Fabric workspace.
2. Copy the REST endpoint URL from the publish dialog.
e.g https://api.fabric.microsoft.com/v1/workspaces/{workspaceId}/agents/{artifactId}/operations/query?api-version=2024-03-31
3. Extract Workspace ID and Artifact ID from the endpoint URL.
4. In Azure AI Foundry, create a new Microsoft Fabric connection using these IDs.
5. In your Chat Agent project, add the Microsoft Fabric Data Agent as a tool.
6. Test queries and verify results come from the correct Data Agent.
7. Publish your Chat Agent to make it available.
Similarly, you can integrate the Fabric Data Agent with other services such as Copilot Studio, Microsoft Teams, and Copilot in Power BI. This concludes our walkthrough of the end-to-end steps.
[Image Source — Microsoft ]
You can also use the Fabric Data Agent SDK to create an agent using the code-first approach
🔧 GitHub Code: Coming Soon
🎥 YouTube Recording: Coming Soon
The end-to-end process leveraging Fabric Data Agent and Azure AI Foundry Agent is shown below.
What is the role of LLMs in this scenario?
In above setup, you have the Fabric Data Agent working behind the scenes, leveraging LLMs alongside the Azure AI Foundry Agent. So, who is responsible for what?
Let’s break down the roles of each component:
- LLMs (Large Language Models) — such as GPT-4.0 or GPT-4.1 — serve as the intelligence layer that understands natural language, extracts insights, and generates human-like responses.
- Fabric Data Agent acts as the connector and orchestrator between LLMs and enterprise data sources. Think of it as the middleware that makes LLMs enterprise-ready by injecting context, enforcing security, and managing data access.
- Azure AI Foundry Agent is a developer-centric framework designed to build custom agents. It enables tool calling, multi-step reasoning, and integration with external systems — all powered by LLMs.
Together, these components form a powerful architecture for building intelligent, secure, and scalable GenAI solutions.
What Permissions Are Required to Use “Chat with Your Data”?
To interact with chat-enabled agents in Microsoft Fabric, users must have the following permissions:
Workspace Roles
- Viewer: Can query published agents and view results.
- Contributor / Member / Admin: Can create, edit, and share agents, depending on their assigned role.
🔒 Important: Users must also have access to the underlying data sources (e.g., Lakehouse, SQL, semantic models). The Fabric Data Agent enforces Row-Level Security (RLS) and Column-Level Security (CLS) configured on those sources.
Few Real-World Use Cases
· Data Analysts — Quickly explore datasets and validate hypotheses.
· BI Developers — Prototype and test metrics before embedding into dashboards.
· Operations Teams — Monitor KPIs in near real time.
· Executives — Get instant answers on KPIs without waiting for reports.
Tips & Best Practices
· Start small — run targeted queries first before scaling.
· Use clear, specific language for more accurate query generation.
· Validate results against known reports.
· Leverage context injection to define terms and rules.
· Optimize data sources with clean labels and indexing.
· Stay updated — republish when using the new Assistants API UI.
Next Steps & Resources
· Learn more about creating Data Agents: https://learn.microsoft.com/en-us/fabric/data-science/how-to-create-data-agent
· Consuming Data Agents: https://learn.microsoft.com/en-us/fabric/data-agents/consume-data-agent
· Sharing Data Agents: https://learn.microsoft.com/en-us/fabric/data-science/data-agent-sharing
· Data Agent SDK: https://learn.microsoft.com/en-us/fabric/data-science/fabric-data-agent-sdk
About the Author
Alpa Buddhabhatti
Microsoft Data and AI Platform MVP & MCT | Lead Azure Data Engineer | DP-203 • DP-700 • AI-100 • AZ-104•AZ-204 | GenAI •LLM •Azure AI Foundary •Azure OpenAI •Microsoft Fabric • Int’l Speaker
Buddhabhatti, A (2025). Microsoft Fabric Data Agent — Complete Guide. Available at: Microsoft Fabric Data Agent — Complete Guide | by alpa buddhabhatti | Aug, 2025 | Medium [Accessed: 7th August 2025].
