Power BI semantic views let teams define business context and metrics in a Fabric Lakehouse, then reuse those definitions across reports and agents. The public preview is aimed at data teams and Power BI developers who want shared, governed metrics to support both analytics and AI experiences. Microsoft Events demonstrated the feature as part of its Power BI session on the product’s direction in the agentic era.
Semantic views provide a Lakehouse entry point for defining a lightweight semantic layer: the business terms, relationships and metrics that help users and software interpret data consistently. Microsoft describes the goal as defining semantics where the data lives and reusing them across business intelligence and AI scenarios.
The preview supports an open format in YAML for describing datasets, relationships, business metrics and their meaning. It also supports defining metrics with DAX, Power BI’s expression language for calculations. That combination is intended to make business logic reusable instead of leaving each report or agent to interpret raw data independently.
The demonstrated workflow brings together locally defined metrics and definitions inherited from another location. OneLake shortcuts can bring in files containing canonical metric definitions without copying them, so a team can use shared definitions while adding metrics of its own. Microsoft showed the resulting semantic model being queried from Power BI and an agent, with matching metric results.
Semantic views are relevant to Power BI developers, data engineers and analysts working with Fabric Lakehouses and shared business metrics. They address the overlap between analytics and data engineering: teams may need to maintain data context close to the data while also making the same definitions available to reporting and AI tools.
The feature is in public preview. The demonstrated starting point is a Lakehouse, where users can create a semantic view or upload files to begin defining one. DAX experience is useful for teams creating calculated metrics, while the YAML format provides a way to describe the datasets, relationships and business meaning.
Shared metric ownership is another practical scenario. A team can use OneLake shortcuts to inherit files maintained by the owners of canonical definitions, then add additional metrics where needed. This can help avoid maintaining duplicate copies of the same business logic across teams.
The preview’s basic flow starts in a Lakehouse and results in a semantic model that can be queried by different experiences. The demonstration used both files created for the semantic view and files made available through shortcuts.
That shared result is the main operational point: reports and agents can use a common definition rather than each implementing its own version of a metric. The semantic model also provides business-friendly names and relationships that help explain what the data represents.
Microsoft also described public APIs for organizations building their own agentic experiences. The session did not walk through API calls, but their availability was mentioned as a way for custom development to use the semantic context.
Semantic views are a public-preview feature, so teams should treat the experience as a preview rather than assume it is a finished production workflow. The demonstrated definition format is YAML, with DAX for metrics. Microsoft previewed additional work, including TMDL support, SQL metrics, importing and exporting semantics from existing semantic models, and agentic skills for creating semantic views. Those items were presented as coming next, not as capabilities already shown in the preview.
The workflow also depends on teams having meaningful, maintained definitions to share. A semantic view can make definitions reusable, but teams still need to decide which metrics are authoritative and who owns updates. Where another team owns canonical measures, shortcuts provide a way to reference those files rather than copy and independently maintain them.
Before using the preview, identify a Lakehouse and the metrics your reports or agents need to share. Check whether the definitions are already maintained elsewhere, and verify that the resulting metrics return the expected values in the consuming report or agent. This keeps the focus on validating shared business meaning, not just creating the semantic view.
Semantic views were announced as a public-preview feature.
The preview supports YAML to define datasets, relationships, business metrics and their meaning. It also supports DAX for metric calculations.
Yes. The demonstrated semantic model was queried from Power BI and an agent, returning matching metric results.
The demonstrated workflow used OneLake shortcuts to reference metric definition files from another location without copying them.