AI agents are only as good as the data behind them. At FabCon and SQLCon 2026 in Atlanta, Microsoft announced a wide set of Microsoft Fabric updates aimed at making enterprise data ready for AI. Here are the highlights, with the status Microsoft gave.
Generally available
- Fabric data agents: virtual analysts grounded in your data, such as lakehouses, warehouses and semantic models.
- Direct Lake on OneLake: fast Power BI analytics directly over OneLake data.
- Materialized lake views: simpler medallion architectures in Spark SQL and PySpark.
- Shortcut transformations: shape data automatically as it connects to or moves within OneLake.
- Mirroring for Oracle and SAP Datasphere: replicate these sources into OneLake.
- Snowflake interoperability.
Coming soon or in preview
- OneLake security: role-based, row- and column-level controls in one model. At FabCon, Microsoft said general availability was coming within weeks.
- Graph in Fabric: general availability announced as coming within weeks.
- Fabric IQ ontologies via an MCP server: preview.
- Mirroring for SharePoint lists and Dremio: preview, with Azure Monitor coming soon.
- Azure Databricks Unity Catalog reading from OneLake: public preview.
- Migration assistant for Azure SQL: public preview, and a savings plan for databases offering up to 35% savings against pay-as-you-go on selected services.
Why it matters
Many organisations want AI that answers questions like "what were our top-performing branches last quarter?" That only works if the data is unified, governed and described in business terms. Fabric's direction (one lake, shared security, semantic models and agents on top) is designed for exactly that.
Practical next steps
- Start with one domain, such as finance, sales or operations, and build a trusted semantic model.
- Define security once in OneLake rather than in every report.
- Pilot a data agent with a small group and a list of real business questions.
- Mirror before you migrate. Mirroring can bring Oracle or SAP data into OneLake without a big-bang project.
Our Cloud & Infrastructure and AI & Copilot teams can help you design a data foundation for AI.
Source: Microsoft Azure Blog — FabCon and SQLCon 2026 announcements (Arun Ulag, March 2026)



