Microsoft Fabric

    Microsoft Fabric: data from every system in one place.

    At seven o'clock, orders from the ERP, projects from Dataverse and plant readings sit in OneLake, and the management report reads straight from them. We set up Microsoft Fabric so that reports and data agents read the same figures.

    For controlling and IT whose Power BI reports draw on more and more sources and years of history.

    • OneLake
    • Lakehouse
    • Warehouse
    • Data Factory
    • Mirroring
    • Fabric Data Agent
    • Power BI
    • Microsoft Purview
    An embroidery mill in St. Gallen looking out at the cathedral, a woman examines a piece of lace, data flows into a report on the laptop, spools of thread, fabric samples and the robot agent beside her
    01

    When Fabric pays off

    A Power BI model carries many analyses on its own. Fabric joins when several sources come together, history reaches back years or several teams prepare the same data. Everything sits in OneLake in the open Delta format: Data Factory loads the sources, mirroring keeps databases such as SQL Server, Oracle or Snowflake current in near real time, lakehouse and warehouse prepare the data, and Power BI reads it in Direct Lake mode without a copy of its own. Fabric data agents answer questions about the same tables, with the rights of the person asking.

    02

    What we build in Fabric

    We start with one area and one decision, such as contribution margin per project, and build the data layers for it so that the next area builds on top. The building blocks:

    • Loading and mirroringMirroring for databases that Fabric supports directly, pipelines and dataflows in Data Factory for ERP, files and interfaces, shortcuts to data that stays in Azure or another cloud.Result: every source has a load path, a schedule and a person who is told about failures.
    • Lakehouse and warehouseRaw data, cleaned tables and business views in separate layers, in the lakehouse with notebooks, in the warehouse with T-SQL for teams that know SQL.Result: a measure is calculated in one place and used the same way in every report.
    • Power BI on FabricSemantic models in Direct Lake mode over the tables in OneLake, with roles for regions or cost centres; existing reports are carried over.Result: reports show large volumes of data without an overnight import and read the same state as the analysis.
    • Fabric data agentsOne agent per business area with up to five sources, instructions on your terms and example queries, reachable from Microsoft Copilot, Copilot Studio or Microsoft Foundry.Result: sales asks for revenue per region and gets the figure from the checked model, only from data it may see.
    • Governance with PurviewSensitivity labels on Fabric items, DLP policies for lakehouses, warehouses and semantic models, activities in the Purview audit; plus domains, endorsement and lineage in the OneLake catalog.Result: sensitive tables are labelled, and it is traceable who opened or shared them.
    • Capacity and operationsAn F SKU sized to the load, workspaces per area and environment, Git integration and deployment pipelines from development to production, monitoring of utilisation.Result: changes reach production after checks, and overload shows before reports slow down.
    03

    Capacity, licences, region

    Fabric runs on a capacity from Azure, an F SKU billed per second that can be paused; a 60-day trial capacity is available for trying it out. Anyone who creates and shares Power BI content needs Power BI Pro; readers with the free licence open reports from F64 upwards with the Viewer role. Premium Per User does not create a Fabric capacity. Switzerland North and Switzerland West offer the Fabric workloads; because Switzerland lies outside the EU Data Boundary, Copilot and data agents there need approval for processing and storage outside the region. Plus read access to the sources and one person from controlling for the definitions.

    04

    Reconciled with your figures

    One area in Fabric whose figures your controlling team has reconciled with previous reports for a closed period, with deviations explained. Load paths with logs and a deliberately broken source whose failure reaches the person responsible. Roles tested with accounts from every group, and a data agent that answers a list of real questions with the expected figures. Plus the capacity utilisation from the pilot before further areas join.

    Sound familiar? Tell us where you stand today.

    Contact

    What's new.

    As of: October 2026

    • As of 15 June 2026

      From F64, people with a free licence can read along.

      Fabric capacities range from F2 to F8192 and are billed through Azure per second, with a one-minute minimum. Below F64, every person reading Power BI content needs a Pro or Premium Per User licence; from F64 the free licence with the Viewer role is enough. Premium Per User does not create a Fabric capacity.

      Source: Microsoft Learn: Microsoft Fabric licenses and capacity
    • As of 28 August 2026

      Mirroring from SQL Server, Oracle, SAP and Snowflake.

      Fabric continuously mirrors, among others, Azure SQL, SQL Server, Azure Database for PostgreSQL, Azure Cosmos DB, Oracle, SAP Datasphere, Snowflake and Google BigQuery into OneLake; SharePoint lists are in preview. Each purchased capacity unit includes one terabyte of storage for mirroring, and replication does not consume the capacity.

      Source: Microsoft Learn: Mirroring in Microsoft Fabric
    • As of 29 September 2026

      Fabric data agents are generally available.

      A data agent answers questions from up to five sources: lakehouse, warehouse, semantic model, KQL database, ontology (preview) or Microsoft Graph. It only reads, with the rights of the person asking, and follows Purview policies. It needs a paid capacity from F2 and can be called from Microsoft Copilot, Copilot Studio and Microsoft Foundry.

      Source: Microsoft Learn: Fabric data agent
    • As of 25 September 2026

      All Fabric workloads in Switzerland North.

      Switzerland North offers Power BI and all Fabric workloads, as does Switzerland West, except the Fabric Apps preview. Workspaces on a capacity in one of these regions store their content there.

      Source: Microsoft Learn: Fabric region availability
    • As of 29 September 2026

      Data agents on Swiss capacities need approval.

      Copilot and data agents in Fabric use Azure OpenAI. If the capacity sits outside the EU Data Boundary and the US, as in Switzerland, administrators must allow in the tenant settings that data is processed and stored outside the region. When a data agent is called from other services, their rules apply as well.

      Source: Microsoft Learn: Fabric data agent tenant settings

    Reports Fabric can build on.

    Power BI analyses from our projects, built on Power Platform and ERP data: building data at AEW, ERP, PIM and CRM data at REV Ritter and quality documents at Inselspital. Fabric joins when models like these grow.

    What controlling and IT ask about Fabric.

    Do we need Fabric if we have Power BI?

    Not necessarily. A Power BI model carries many analyses on its own. Fabric pays off when several sources come together, queries slow the model down, history reaches back years or agents should use the same data. We decide that from your sources and data volumes.

    Which capacity do we need?

    The smallest one that carries your load; in the pilot we measure utilisation with the Capacity Metrics app. From F64, people with the free licence open the reports; below that, every reader needs Pro or Premium Per User. A capacity can be scaled up, scaled down and paused.

    Does the data stay in Switzerland?

    Workspaces on a capacity in Switzerland North store their content in that region; some metadata sits in your tenant's home region. Copilot and data agents may process outside Switzerland through Azure OpenAI and only work once Fabric administration allows it. When a data agent is called from Microsoft Copilot, Copilot Studio or Microsoft Foundry, the rules of those services apply as well. We record which features you switch on.

    Do data agents understand German?

    As of October 2026, Microsoft supports only English for data agents and recommends questions, instructions and example queries in English. Answers are capped at 25 rows and 25 columns, and the agent does not change data. In the pilot we test with the questions of your business area how well this works for your team.

    What happens to our existing reports?

    They keep running. Reports and semantic models in a workspace on a Fabric capacity work as before; where a model benefits from Direct Lake, we switch it once the tables are in OneLake.

    How does Purview protect the data in Fabric?

    Sensitivity labels can be set on all Fabric items and stay in place through supported exports. DLP policies detect sensitive data in lakehouses, warehouses and semantic models and restrict risky sharing, and every activity is recorded in the Purview audit. Data agents follow these policies.

    Agree on a useful next step

    Tell us about your current workflow, the systems involved and the result you need. In our first conversation, we identify what else we need to know to plan a useful next step.