Next upSF Pitch Night by the AI Collective - #SFTechWeek
News

Microsoft previews Fabric data engineering agent and expands Azure SQL tools

Microsoft is adding agent-assisted data preparation, governed business context and database operations across Fabric and Azure SQL, with some tools in preview and others generally available.

D
Oct 3, 2026 · 3 min read

Microsoft introduced a preview data engineering agent in Fabric alongside database, observability and semantic-model updates designed to give Copilot and enterprise agents governed data and operational context.

Availability varies across the releases. Fabric’s data engineering agent is in preview, as are Fabric IQ ontology, Fabric Observability Insights and Database Hub. Fabric IQ integration with Microsoft Copilot Chat and Cowork is generally available, while several Azure SQL and SQL Server capabilities have also reached general availability.

An agent for long-running data engineering

The data engineering agent builds on technology Microsoft acquired with Osmos. Microsoft says it can plan, execute and validate work over hours or days, including data preparation, platform migrations, lakehouse modernization, performance optimization and complex extract-transform-load jobs.

Microsoft’s Fabric update says the agent can create and test Spark notebooks, transform data, write Delta tables and keep running after a local client disconnects. Its controls combine Fabric permissions with settings for write safety, reruns, schema changes and generated artifacts. Colleagues with access to the hosting lakehouse can also see shared tasks.

Alongside that preparation layer, Microsoft is adding reusable business context. Fabric IQ integration with Copilot Chat and Cowork can use governed Power BI reports and semantic models without a separate AI-specific copy of the data. The company says Copilot can therefore use measures, relationships and definitions already maintained in Power BI.

The Fabric IQ ontology feature remains in preview. It maps business entities, relationships, rules and metrics to source data, giving applications and agents a shared vocabulary. Microsoft’s ontology agent can generate definitions and bindings, answer questions against that context and propose changes for human review before they are applied.

Real-time operations and database control

For real-time data, Microsoft announced natural-language creation of Real-Time Dashboards, Copilot help with eventhouse administration, and schema recommendations and optimization. An operations agent can investigate anomalies, correlate telemetry and recommend actions. Fabric Observability Insights, in public preview, combines Monitor Hub, workspace monitoring in a read-only Eventhouse database and AI-assisted investigation of failures and root causes.

Database Hub is also in public preview. The centralized Fabric view covers inventory, health, performance, security, risk, governance and optimization signals across SQL Server, Azure SQL, PostgreSQL, Azure Cosmos DB and SQL database in Fabric.

Microsoft also announced database agents for SQL and PostgreSQL on Azure, but its posts do not establish the same immediate availability. The broad announcement says the agents will soon be available in preview; the SQL-specific post says the preview experience begins with a SQL agent in Database Hub and a PostgreSQL agent in Visual Studio Code. Microsoft plans to bring both capabilities to both surfaces over time.

Azure SQL Database Hyperscale serverless auto-pause and higher-density elastic pools are in preview, supporting as many as 50 databases per pool. Premium-series Hyperscale service objectives at 160 and 192 vCores are generally available.

DiskANN vector indexes are generally available for Azure SQL Database Hyperscale, Azure SQL Managed Instance and SQL database in Fabric. Microsoft says full data-manipulation support keeps a vector index current when its underlying table changes. Agent Mode for GitHub Copilot in SQL Server Management Studio is also generally available for multistep work including dependency investigation, execution-plan analysis, and generating and evaluating data-definition language.

Microsoft also made SQL Server on Azure Local generally available for connected and disconnected operations. The disconnected configuration provides a local control plane for environments where cloud connectivity is unavailable, restricted or intentionally disabled.

More news