Microsoft Fabric IQ: Check GA status by component
Fabric IQ includes components with different release statuses, so Fabric teams should plan around each item rather than the workload label.
Key insights
- Fabric IQ is a workload of related items, not a single feature with one release status.
- The demonstration describes ontology as still in preview and Graph and Operations agents as GA.
- The old ontology experience is scheduled to retire on January 31, 2027.
- An ontology generated from a semantic model can use its tables, columns, relationships, measures, and calculated columns.
- Graph is optional and does not load instance data by default when generated from ontology.
Microsoft Fabric IQ combines related data and AI capabilities, but its components do not all share the same release status. Fabric administrators, data teams, and project leads should check the status of each item before building a production plan around it. Pragmatic Works demonstrates how to assess the components and describes a practical starting point: trusted Power BI semantic models.
What's new
Fabric IQ is a workload—a group of related items—not one feature that has a single on-or-off switch or release status. The components discussed include ontology, Power BI semantic models, planning, Graph, data agents, and operations agents. The demonstration characterizes Fabric IQ as a mix: some parts are generally available (GA), while others remain in preview.
In the status picture presented, Graph and Operations agents are GA, while ontology remains in preview. Power BI semantic models have already been production-ready, and planning is described as having arrived after the GA announcement. The practical implication is that “Fabric IQ is GA” is not enough to establish that every capability you want to use is ready for production.
Ontology is a shared, machine-understandable representation of a business. It describes concepts such as customers, products, and orders, how they relate, and the rules and metrics that give them meaning. It connects those definitions to data in places such as a lakehouse, warehouse, Eventhouse, or semantic model; it is not described as another copy of that data.
Who it is for and what you need
This matters to Fabric and Power BI teams evaluating business context for analytics and agents, as well as administrators responsible for permissions and rollout planning. The demonstration does not specify a required license, capacity SKU, region, or tenant-wide availability, so those details cannot be established from the material here. It does mention that tenant settings are required for ontology, but does not provide their names or values.
For teams with existing Power BI semantic models, those models can provide a starting point for ontology work. A generated ontology can bring in tables and columns from a semantic model, as well as its relationships, measures, and calculated columns. The DAX remains owned by the semantic model; ontology metrics point back to those measures.
The release status described is not uniform. Ontology is called preview in the demonstration, while Graph and Operations agents are described as GA. Treat those descriptions as the source’s account, and verify current status and requirements in Microsoft’s product documentation before making a deployment decision; no official Microsoft documentation URL was supplied among the available sources.
How it works
A semantic model can hold established relationships and measures that already reflect business definitions. Connecting an ontology to those definitions can help agents use existing measures instead of relying on copied formula text in prompts. If a measure changes in the semantic model, the ontology reference can use the updated measure, according to the demonstration.
- Review the release status of each Fabric IQ item you intend to use; do not treat the workload label as the status of every component.
- Identify a trusted Power BI semantic model or lakehouse as a possible starting point for ontology generation.
- Map business concepts, relationships, rules, and metrics to the data and measures they represent. The demonstration says ontology rules can express business logic in plain language.
- Use Graph when following relationships across entities is central to the question. Graph is described as an optional execution layer; generating it from ontology does not load instance data by default.
- For agents, check how permissions on the ontology and underlying data apply. The demonstration says Fabric permissions and OneLake, row-level, object-level, and column-level security remain relevant.
The video also describes three agent experiences. Data agents support domain-specific question-and-answer experiences grounded in semantic models and ontology; Operations agents monitor real-time data, detect anomalies, and recommend or trigger actions; and the ontology agent helps create definitions and bindings. For the ontology agent, proposed changes are reviewed before they are applied.
Limitations and open questions
The clearest planning constraint is ontology’s preview status. The demonstration reports that the old ontology experience is scheduled to retire on January 31, 2027, and says a copy can be created in the new experience. That copy is a separate item: agents and dashboards must be reconnected, and rules must be recreated.
Those migration steps can add work, particularly for teams that have already built dependencies on the old experience. The material does not specify a licensing requirement, a rollout schedule by region, or a complete list of tenant settings. It also does not establish that every described capability is available to every organization.
Graph is not presented as necessary for every query. The demonstration says ordinary lookups and aggregations can remain against source systems, and Graph can be used when traversing relationships is the key task. For a critical production process that depends on ontology, the source recommends prototyping while it is in preview and planning production around its eventual GA status rather than assuming that status has already been reached.
Security is another design consideration, not a reason to skip access checks. The demonstration says agent requests through Foundry run as the signed-in user and describes Fabric permissions, workspace access, OneLake security, and semantic-model row-level security as relevant. The exact configuration should be checked against the organization’s implementation and current product documentation.
Sources
Frequently asked questions
Is Microsoft Fabric IQ generally available?
The material describes Fabric IQ as a workload whose components have different statuses. Graph and Operations agents are described as GA, while ontology is still in preview.
When does the old Fabric ontology experience retire?
The demonstration gives January 31, 2027, as the retirement date for the old experience. It says the new-experience copy is a separate item, so agents and dashboards must be reconnected and rules recreated.
Can I generate a Fabric ontology from a Power BI semantic model?
Yes. The demonstration says generation can bring over tables, columns, relationships, measures, and calculated columns. DAX remains in the semantic model, and ontology metrics point back to its measures.
What license or tenant settings does Fabric IQ require?
The material does not specify a license or capacity SKU. It mentions required tenant settings for ontology but does not identify them.