Microsoft business data AI took a specific step toward the enterprise on September 28, 2026: Fabric IQ became generally available in Copilot Chat and Cowork. Microsoft announced the development at its Fabric and SQL community conference in Barcelona, alongside previews of new data-sharing and database-management capabilities.
Why put so much emphasis on data definitions rather than the model itself? Within a business, the question “What was our revenue last month?” may have several answers depending on when a sale is recorded, whether returns have been deducted and which units are included. An assistant can produce a fluent response without resolving those disagreements. Microsoft’s approach is to link Copilot to the governed business definitions already used in Power BI reports.
Fabric IQ gives Copilot a reference point
Fabric IQ uses Power BI semantic models, including established metrics and relationships, to ground answers in company-specific context. According to Microsoft’s September 28 announcement, the generally available integration lets Copilot Chat and Cowork draw on those definitions while preserving existing governance and access controls.
The difference may sound subtle. A general-purpose model can explain a financial term; a model grounded in an approved reporting system can work with the version of that term the business actually uses. This is useful when an employee needs a forecast or a regional performance figure, provided the underlying report is accurate and the employee is authorized to see it.
Microsoft says this capability is enabled by default for relevant Fabric and Power BI customers. That does not remove the need to validate reporting models. If a metric or data pipeline is faulty, the assistant may simply communicate a faulty result more conveniently.
Sharing business knowledge is a different problem
IQ Sharing, which Microsoft describes as a preview, is intended to share governed data together with the business context that makes the data meaningful. A supplier receiving a spreadsheet may otherwise have to guess whether an inventory figure represents stock on hand, reserved units or goods in transit. Sharing definitions and permissions together could reduce ambiguity.
Preview status matters. Organizations should not assume every proposed workflow is ready for deployment, and access policies still need legal and security review. Microsoft also discussed cross-workspace observability and tools that help teams spot problems across data workloads. Those systems are relevant because dependable AI answers ultimately depend on dependable infrastructure.
Database Hub shifts attention to operations
Database Hub in Fabric entered public preview, providing a central view of databases across technologies such as SQL Server, Azure SQL, PostgreSQL and Cosmos DB. The aim is to make monitoring and governance less fragmented. Microsoft also outlined specialist database agents for investigating performance issues, with preview access planned rather than broadly available at the announcement.
That distinction between current functionality and future capability is particularly important for IT teams. A database agent suggesting an optimization is not the same as a fully autonomous system safely changing production infrastructure. Administrators still need controls for access, change review, rollback and incident response.
The company also announced general availability of SQL Server on Azure Local, while disconnected operations were described as a preview. That gives organizations with regulatory or location requirements more options for where certain workloads run, though the exact operational characteristics depend on the configuration and Microsoft’s documentation.
What the rollout means for businesses
Microsoft says Fabric serves more than 40,000 customers worldwide in its Azure conference briefing. This is a company-reported platform customer figure. It should not be interpreted as the number of businesses already using the latest Fabric IQ features successfully.
A large organization that already relies on governed Power BI models may find a straightforward place to test Copilot against known reporting questions. A smaller firm with inconsistent accounting data may gain more by cleaning records and agreeing on basic definitions first. The right order of investment differs, even if both organizations eventually use AI assistance.
One practical test is to compare responses with approved reports. Track which questions Copilot answers correctly, whether it respects permission boundaries, how much employee time it saves and how often a human needs to correct an interpretation. Costs deserve a separate review: a claim about no extra AI token fees for one integration does not mean the complete Fabric deployment is free.
BitCodeSolution’s WordPress security checklist addresses a related principle for online systems: access controls, monitoring and routine review matter as much as individual features. For enterprise data, that principle applies to every dataset an agent can query.
Microsoft’s announcement offers businesses a clearer proposition than the idea of AI that simply “knows the company.” Fabric IQ is available for certain Copilot experiences, while sharing and operational tools remain in preview or development. The practical outcome will depend on whether governed information is accurate, accessible to the right people and connected to decisions worth improving.
Sources: Microsoft Official Blog; Microsoft Azure Blog. September 28, 2026.
