Microsoft Fabric Data Agent Lets Teams Query Data Without Coding

Data teams know the frustration: a business analyst submits a simple question about revenue trends, and three days later, the answer arrives—if it arrives at all. Microsoft’s Fabric Data Agent, currently in public preview, is designed to eliminate that gap by letting anyone ask questions in natural language and receive structured, accurate responses from enterprise data without writing a single line of code.

The agent is part of Microsoft Fabric’s OneLake ecosystem and uses Azure OpenAI’s Assistant APIs to interpret questions, determine the best data source, generate the right query, and deliver a human-readable answer. It connects to lakehouses, warehouses, Power BI semantic models, KQL databases, ontologies, and Microsoft Graph—all within Fabric’s governed environment. Within multi-agent architectures on Fabric, data agents act as the conversational analytics layer, bridging governed data and business users through natural language.

How Fabric Data Agent Processes Your Questions

Behind the scenes, the agent follows a structured pipeline to ensure both accuracy and security. When a user submits a question, it starts with validation:

• The agent uses Azure OpenAI Assistant APIs to parse the question in real time.
• It verifies compliance with security policies, responsible AI guidelines, and the user’s access permissions.
• All interactions are read-only; the agent cannot modify or delete data.

Next, the system identifies the most relevant data source. It checks schema metadata (not the actual data) using the user’s credentials to determine which connected source can best answer the question. Organizations can customize this routing with instructions such as, "Route financial metric questions to the Power BI semantic model and raw data exploration to the lakehouse."

Once the data source is selected, the agent translates the natural language question into the appropriate query language:

• Lakehouses and warehouses → Natural Language to SQL (NL2SQL)
• Power BI semantic models → Natural Language to DAX (NL2DAX)
• KQL databases → Natural Language to Kusto Query Language (NL2KQL)
• Microsoft Graph → Graph API queries

The generated query is validated for correctness and security before execution. Results are formatted into readable tables, summaries, or key insights and delivered back to the user.

Setting Up a Fabric Data Agent: A Step-by-Step Guide

Configuring a Fabric Data Agent resembles building a Power BI report: you design it, refine it, and then publish and share it across your organization. Here’s what the setup involves:

• Select data sources: You can connect up to five sources in any combination—lakehouses, warehouses, KQL databases, Power BI semantic models, ontologies, or Microsoft Graph.
• Choose relevant tables: For lakehouses, you define which tables the agent can access. Raw files like CSV or JSON must be ingested into tables first to be queryable.
• Add context with instructions and examples:
• Define behavior rules and routing logic for different question types.
• Clarify organizational terminology and domain-specific terms.
• Provide example question-answer pairs to help the agent learn common query patterns (note: examples aren’t yet supported for Power BI semantic models).

Security and governance are built into the agent from the start. The agent respects least-privilege access by inheriting the user’s credentials, meaning it only surfaces data the user is already authorized to see. It integrates with Microsoft Purview for data loss prevention, access restriction, insider risk management, and audit logging. Additionally, you can enable Azure AI Content Safety to filter harmful or out-of-policy responses.

Beyond the Chat Interface: Integration with Copilot Studio

Fabric Data Agents aren’t confined to the Fabric portal. The agent can be embedded into custom Microsoft 365 Copilot experiences, Microsoft Teams bots, or other applications through Copilot Studio. This makes it possible to bring governed data conversations directly into the tools your teams already use every day.

Why This Matters for the Future of Data Analytics

The Fabric Data Agent tackles a persistent challenge in enterprise environments: making data accessible to non-technical stakeholders without compromising governance or security. Most organizations have vast stores of data, but only a fraction of employees can extract insights from it without technical skills.

By enabling plain-English interactions with governed data, the agent lowers the barrier for business users, reduces the burden on data teams for ad-hoc queries, and supports a culture of informed decision-making. It keeps everything within existing enterprise controls, ensuring that insights are both timely and trustworthy.

As AI-driven analytics continues to evolve, tools like the Fabric Data Agent are paving the way for a more inclusive and responsive data culture—one where questions are answered in minutes, not days, and insights drive action at the speed of business.