Building an open, AI-ready data foundation with Microsoft Fabric

Microsoft Fabric gives organizations a unified analytics platform. But as data teams take on more workloads and expand their AI ambitions, an important question remains: how do you build on Fabric today while keeping your data, business logic, and future options open?

That question was front and center at FabCon Barcelona 2026, where Fivetran + dbt Labs joined Microsoft and the broader data community to explore the future of data on Azure and Microsoft Fabric.

The opportunity goes beyond bringing data together. Organizations need reliable data pipelines, trusted business definitions, and an open foundation that supports analytics and AI without adding unnecessary complexity.

Several developments highlighted at FabCon point toward that future. Here are 5 takeaways for teams building on Microsoft Fabric, and what they mean for your data strategy. 

[CTA_MODULE]

1. Keep your data open and accessible across engines

As workloads multiply, so do copies of data. Every new engine or team can mean another pipeline, another copy, and another version of the truth. An open foundation avoids that by storing data once, in open formats, in storage you control.

Fivetran Managed Data Lake does this on Azure. It writes data from 700+ sources directly into Azure Data Lake Storage as open Apache Iceberg tables, with Delta Lake metadata available as an option. Because it integrates with the OneLake Catalog, those same tables are discoverable and governed inside Fabric.

Fabric, other Azure services, and any engine that reads Iceberg or Delta can query the same copy. You don't need to re-ingest data or build duplicate pipelines, so each engine can do the job it does best.

2. Make business logic reusable, not tied to one platform

Your metric definitions, transformations, and tests are some of your most valuable data assets. They shouldn't have to be rebuilt every time your compute strategy changes.

Microsoft builds and maintains the dbt adapters for Fabric Warehouse and Fabric Lakehouse, so existing dbt projects can target Fabric without a rewrite. You write business logic once, in SQL and dbt, and it stays portable wherever your compute runs. That means you can commit to Fabric today without locking your logic into a single platform.

3. Give AI access to trusted business context

AI agents are only as reliable as the context they work from. If an agent queries raw tables and guesses what "revenue" or "active customer" means, its answers won't match the numbers your business already trusts.

The dbt MCP Server, now generally available, closes that gap. It exposes dbt models, metrics, lineage, and tests to AI agents through the Model Context Protocol, an open standard. Agents, including copilots built on Microsoft's AI stack, work from the same governed definitions your analysts use. You define your business context once, and every person and agent can reuse it.

4. Improve developer productivity without starting over

At FabCon Barcelona, Microsoft announced that dbt Job in Microsoft Fabric is now generally available. Teams can build, schedule, and run dbt transformations natively inside Fabric. The GA release adds:

• Broader enterprise support for running dbt at scale.
• dbt logs that land in OneLake, so run history sits alongside the data it produced.
• More Fabric adapters, including Lakehouse.
• Variable library support in deployment pipelines, so the same project promotes cleanly from dev to test to prod.

Looking ahead, dbt v2 went GA at dbt Summit. It's the Rust-based engine that grew out of Fusion, built for faster parsing, smarter compilation, and a better developer experience. Both dbt Labs and Microsoft list Fabric support as coming soon. Because dbt v2 runs the same dbt projects, Fabric teams will be able to adopt the new engine without rebuilding what they already have.

5. Bring the pieces together with Open Data Infrastructure

Each of these developments matters on its own, but the real value comes from how they fit together:

• Fivetran moves data from hundreds of sources into open tables in your own Azure storage.
• dbt turns that data into tested, documented business logic that runs natively in Fabric and stays portable.
• The dbt MCP Server makes that logic available to AI agents.

The result is an Open Data Infrastructure: one copy of your data, one set of trusted definitions, and the freedom to choose the right engine and tools as your analytics and AI needs grow.

Putting it into practice: Gerflor's experience with Microsoft Fabric 

At FabCon Barcelona, French flooring manufacturer Gerflor shared how it is modernizing a 15-year-old on-premises data platform with Microsoft Fabric, Fivetran, and dbt. The transformation must support a global business spanning 40 subsidiaries and more than 400 applications, with an average of 4 new subsidiaries added each year.

Gerflor’s fragmented data landscape and inconsistent business definitions made information difficult to reuse and trust. Month-end consolidation alone required 10 employees and roughly 5 days of manual preparation. Meanwhile, its move to Microsoft SaaS created demand for consistent commercial reporting across CRM and Business Central for more than 1,700 users.

To address these challenges, Gerflor is building a cloud data foundation on Azure Data Lake Storage Gen2 and Microsoft Fabric. Fivetran automates change data capture and data movement, reducing manual pipeline maintenance and accelerating the onboarding of new sources and entities. dbt standardizes transformations and business logic so Gerflor can create governed, reusable datasets across the organization.

The foundation is designed to improve data consistency, shorten reporting cycles, and free teams to focus on analysis and AI innovation. Gerflor plans to expand its use of Fivetran connectors, scale its Microsoft Fabric capacity, standardize more transformations with dbt, and support AI services. The result is a scalable approach that connects infrastructure modernization directly to business value.

[CTA_MODULE]