Microsoft Fabric IQ adds ‘semantic intelligence’ layer to Fabric

With Fabric IQ, Microsoft is adding new semantic intelligence capabilities to its unified data and analytics platform, Fabric, that it says will help enterprises maintain a common data model and automate operational decisions.

We’ll be hearing a lot more about “IQ” from Microsoft, which has also just introduced Work IQ, the semantic intelligence layer for Microsoft 365, and Foundry IQ, a managed knowledge system for grounding AI agents via multiple knowledge repositories, that together with Fabric IQ form what Microsoft calls a “shared intelligence layer” for the enterprise.

Fabric IQ will offer five integrated capabilities, Microsoft said: the ability to create an ontology or shared model of business entities, relationships, rules, and objectives; a semantic model extending business intelligence definitions beyond analytics to AI and operations; a native graph engine to enable multi-hop reasoning with data; virtual analysts called data agents that can answer business questions; and autonomous operations agents that can reason, learn, and act in real time.

The ontology is the foundation of all of this, a living structure capturing key business concepts, the company said. It contrasted it with traditional data modelling, saying that it can be created and evolved by business experts using a no-code tool (still in preview), without support from engineers — yet still offering IT managers the control to secure, approve, version, and manage it.

Constellation Research principal analyst Michael Ni was skeptical: “There is upfront work for IT. Ontologies don’t build themselves,” he said.

IT teams may be able to capitalize on work they have already done: Organizations using Power BI can import its data models as the basis for their ontology, Microsoft said.

From ontology to autonomy

For organizations not already invested in the Fabric ecosystem, adoption of IQ will likely take longer, warned Suhas AR, associate practice leader at HFS Research. “Key hurdles include agreeing on shared business definitions, ensuring that data permissions are carried through to AI agents, and maintaining clean and reliable real-time data feeds for automation. Teams also need some new skills and processes to govern these agents as they evolve.”

The endgame of all this describing of data is to ensure that workers, both human and AI, have a shared understanding of what the data means in the real world, so that they can analyze and act on it together.

Operations agents “monitor the business in real time, reason over live conditions, evaluate trade-offs, and take actions automatically to advance business outcomes,” Microsoft said in a blog post announcing Fabric IQ.

Such agents are capable of much more than just alerting or simple workflow automation, Microsoft said, enabling decisions at scale in seconds, without the need for interminable meetings.

Who benefits?

While successfully implementing Fabric IQ is likely to involve IT teams in some up-front heavy lifting, they’ll benefit from a longer-term reduction of operational effort, analysts say.

Microsoft-first enterprises with strong central governance stand to gain the most, Suhas said, cautioning enterprises with low data-maturity, or not committed to Microsoft’s data stack, to “wait and watch.”

Constellation Research’s Ni sees good reasons for adopting Fabric IQ: “These benefits include consistent semantics, fewer one-off models, less duplicated logic, and a shared decision layer that lowers downstream maintenance as enterprises ramp up iteration on decision automation and AI-driven automation,” he said.

Stephanie Walter, practice leader of AI Stack at HyperFrame Research, doesn’t expect IT teams’ data modelling workload to disappear with the introduction of Fabric IQ’s ontology, she does see it shifting towards controlling security, and approving the changes made by business users.

Other analysts have reservations, fearing complex, time-intensive deployments and vendor lock-in.

While Fabric IQ’s ontology will provide a clear basis for communication between employees and autonomous agents it will also, according to Moor Insights and Strategy principal analyst Robert Kramer, tie the enterprise to it.

“The more an enterprise builds on this semantic layer, the harder it becomes to move that logic elsewhere.,” Kramer said.

Fabric costs

Suhas, too, pointed to the heavy switching costs enterprises would face if they wanted to move to another platform that didn’t support the Fabric IQ ontology.

And if, after spending on the creation and governance of the ontology, and all the attendant Fabric IQ services, an enterprise was unable to drive meaningful agent adoption, then all that investment would be for nothing.

Predicting or measuring that investment will be a challenge in itself. Fabric IQ is treated as a workload in Microsoft Fabric, just like Data Factory, Analytics, Databases, Real-Time Intelligence, and Power BI, the company said.

“It uses the same unified Fabric licensing model and runs on Fabric capacity. There is no separate SKU or add-on fee,” said Yitzhak Kesselman, Microsoft CVP for Messaging and Real-Time Analytics Platform.

Those costs are tracked via a bewildering array of Fabric capacity invoice meters for underlying infrastructure usage.

Microsoft hasn’t yet published the billing meters for the Ontology item, but plans to do so later this week, with billing beginning in the first half of 2026, he said. Billing for other Fabric items will remain unchanged.