Data Agent Now Supports Eventhouse Functions, Materialized Views, and Shortcuts (Preview)

Author: Shreyas Canchi Radhakrishna - Product Manager

New capabilities in Fabric Data Agent are now available to expand what's possible when data agent is integrated with Eventhouse KQL Databases. Data Agent now supports Eventhouse User Defined Functions (UDFs), Materialized Views (MV), and Shortcut tables enabling your AI-powered agent access to the full breadth of your Eventhouse KQL database.

Why this matters

Data Agent in Microsoft Fabric lets business users, analysts, and developers ask questions of their data in plain English. Behind the scenes, the agent translates natural language into precise KQL queries and returns results with no query authoring required.

Previously, Data Agent could query the tables in your KQL Database directly. But real-world Eventhouses are far richer than raw tables alone. Teams encode business logic in functions, optimize high-frequency dashboards with materialized views, and federate data across storage boundaries with shortcut tables. By extending Data Agent to understand and leverage these entities, we're closing the gap between what your Eventhouse can do and what your agent knows how to do.

What's new

Eventhouse user defined functions (UDFs)

What are UDFs: User defined functions are reusable KQL query fragments that encapsulate complex business logic. A function like `DetectAnomalyWindow(lookbackMinutes, minConsecutiveBreaches)`might join three tables, apply filtering, and return a clean result set.

Capabilities with Data Agent: Data Agent discovers the UDFs defined in your Eventhouse Database, understands their signatures and purpose, and calls them when they're the best way to answer a user's question. Creators can also provide additional descriptions and context for their UDFs, which improves the accuracy of the agent's responses.

Why it matters for data agent users:

• Consistency: The same validated business logic your engineering team built is now what the agent uses — no risk of the AI reinventing a calculation incorrectly.
• Accuracy: UDFs often encode domain-specific nuance (edge-case handling, time-zone logic, custom aggregations) that would be difficult to infer from raw table schemas alone.
• Simplicity: Analysts don't need to know the function exists or how to call it. They just ask their question, and the agent routes to the right function automatically.

Figure: An example GIF to demonstrate: "Which devices have had at least 3 critical sensor readings in the last 60 minutes?".

Data Agent recognizes that `DetectAnomalyWindow (60, 3)` is the right tool for the job, calls it, and returns the result — complete with the same filtering and deduplication logic your team defined.

• Learn more about Eventhouse User Defined Functions in the Eventhouse documentation.

Materialized views

What are Materialized views: Materialized views are pre-computed aggregations that Eventhouse maintains automatically as new data streams in. They're the backbone of high-performance dashboards. Instead of scanning billions of raw rows, queries hit a compact, pre-aggregated table.

Capabilities with Data Agent: Data Agent now includes materialized views in its schema discovery. When a user asks a question that aligns with a materialized view's aggregation, the agent queries the view instead of scanning the base table.

Impact on data agent users:

• Performance: Queries that would scan large sets of raw telemetry now return in milliseconds by hitting pre-aggregated views. This means faster answers from the agent, even on massive datasets.
• Cost efficiency: Fewer compute resources consumed per query translate directly to lower capacity utilization and cost.
• Real-time freshness: Materialized views in Eventhouse update continuously as data arrives, so the agent's answers reflect the latest state of the world, not a stale snapshot.

Figure: An example GIF to demonstrate: "Which device has had the highest peak temperature in any 5-minute window, and what was that peak?"

Instead of scanning 50 billion raw request logs, Data Agent queries `DeviceHealth_5min`, which is a materialized view that maintains hourly averages and returns the answer in under a second.

• Learn more about Eventhouse materialized views from the materialized view documentation.

Shortcut tables

What are Shortcut tables: Let your Database query data that lives outside the Eventhouse — in Azure Data Lake Storage (ADLS), Azure SQL Database, other Fabric Lakehouses, or even cross-cluster Eventhouse databases. They're the federation layer that connects your real-time analytics to the rest of your data estate.

Capabilities with Data Agent: Data Agent now sees external tables alongside native tables during schema discovery. When a user's question requires data that resides outside the Eventhouse, the agent seamlessly queries the external table without manual join instructions required.

Impact on data agent users:

• Unified experience: Users ask one question, and the agent handles the complexity of reaching across storage boundaries. There's no need to know where data lives, just what you want to know.
• Broader context: Many analytical questions require combining real-time streaming data with historical archives, reference data, or dimensional tables stored in a Lakehouse or data lake. External tables make this possible without data duplication.
• Zero data movement: The data stays where it is. Shortcut tables query in place, which means no ETL pipelines to build and maintain just to make data visible to the agent.

Figure: An example GIF to demonstrate: "Which devices have warranties expiring in the next 6 months, and what model are they?"

Data Agent queries `DeviceSpecifications` (an external table pointing to the Lakehouse) and return results.

• Learn more about Eventhouse shortcuts from the shortcuts documentation.

The bigger picture: a smarter, more complete agent

These three features share a common theme: they make Data Agent aware of the full richness of your Eventhouse KQL database, not just the raw tables. Together, this means:

For organizations that have invested in building a well architected Eventhouse with thoughtful functions, optimized materialized views, and connected shortcut tables, Data Agent now respects and leverages that investment. Your architecture becomes the agent's intelligence.

Getting started

These capabilities are available today with no additional configuration required. If your KQL Database already has UDFs, materialized views, or shortcut tables defined, Data Agent will automatically discover them in our Schema Browser. Just select these as an input to Data Agent and get started!

To get the most out of these features:

• Add descriptions to your entities. The richer the metadata, the better Data Agent can match user questions to the right function. Use data source instructions to add clear, natural-language descriptions.
• Review your materialized views and functions. Ensure your most common analytical questions are covered by materialized views and functions so the agent can take advantage of them.
• Connect your external data sources. If analysts frequently ask questions that require data outside the Eventhouse, attach the tables as shortcut tables to bring that data within the agent's reach.

What's next

This is part of our ongoing investment in making Data Agent capable of reasoning over real-time analytics. We're continuing to work on deeper integration with the KQL ecosystem, improved reasoning over complex schemas, and richer feedback loops that let creators fine-tune how the agent uses their data assets.

• We'd love to hear how you're using these capabilities. Share your feedback with us through the Fabric Community or your Microsoft account team.
• Learn more about Data Agent in Microsoft Fabric and Real-Time Intelligence with Eventhouse.