From Limited Open-Source to an Agile, AI-Ready Data Catalog in Two Months

The Client

Bertel O. Steen (BOS) is one of Norway’s largest service and trading groups, founded in 1901 and headquartered in Lørenskog. The company is best known as the importer and retailer of Mercedes-Benz, Kia, Peugeot, Opel, Citroën, DS and Smart, and also operates in real estate, agriculture, sports and leisure. BOS manages data in both Norwegian and English, with an analytics stack built on Databricks and Microsoft Fabric (Power BI).

​

The Problem

BOS had already invested significant effort into building a data catalog on DataHub with their open-source platform offering. As the data team scaled their work across Databricks and Microsoft Fabric (Power BI), the limitations started to show.

The critical trigger was the increased operational effort to keep the open-source platform up to date and operational on the DataHub stack. This is part of a broader trend where vendors with an open-source offering gradually retire features to push customers toward their paid cloud offerings. For BOS that meant that the open-source path stopped keeping up with the needs as they scaled, either forcing a move to a paid tier cloud version of the tool or to find a platform that met the requirements.

BOS needed:

• A catalog that could migrate existing DataHub work without losing business glossary, data products, ownership, or lineage
• End-to-end lineage across Databricks, Microsoft Fabric (Power BI)
• A business glossary that works natively in both Norwegian and English
• A sensible total cost of ownership
• Speed: a working solution before summer

​

The Solution

BOS engaged Dawiso at the end of March 2026. Within roughly two months, a fully customized, production-ready environment was in place.

The migration from DataHub was the centerpiece. BOS exported their existing business glossary and data product documentation, and the Dawiso team designed a custom import path (partly automated through the Dawiso MCP server) that brought the content into Dawiso without losing structure or relationships. Where DataHub’s metamodel was too rigid, Dawiso’s was extended to match how BOS actually thinks about their data.

The biggest piece of customization was the bilingual content model. Many objects in BOS’s environment exist in both Norwegian and English, so Dawiso was configured to carry two parallel sections per object, with a single-click toggle between languages. The business glossary was also extended with a custom Business Entity layer above the standard Business Term, giving BOS room to model concepts like Geography → Region → Country → Municipality → Postcode the way their information architecture demanded.

Connectors for Databricks and Microsoft Fabric brought technical metadata into Dawiso and automatically generated lineage across the systems. An MCP integration for Databricks is also in place so that AI agents can reach into governed metadata through a controlled-access layer. Single Sign-On ties Dawiso into BOS’s existing identity setup, so onboarding new users is a matter of permissioning, not provisioning.

​

Main features of the Dawiso solution for BOS:

End-to-End Lineage Across the Modern Stack

Lineage spans Databricks and Microsoft Fabric (Power BI), including column-level detail. Impact analysis works across both systems in one view.

​

Migration Without Loss

BOS arrived with significant pre-existing work in DataHub. Dawiso preserved that investment: every business term, data product, ownership relation and link came across, and the metamodel was adapted to fit BOS’s structure.

​

AI-Ready via MCP

A custom Databricks MCP integration lets AI agents query governed metadata directly. BOS can connect their own tooling and copilots to the same context layer that human users rely on.

​

The Impact

In two months, BOS moved from a constrained open-source deployment to a customized, production-ready catalog. All technical metadata from Databricks and Fabric is now in one place, alongside a bilingual business glossary and a documented portfolio of data products.

For everyday users, searching, reading and navigating data assets is faster and easier than in DataHub. For architects and engineers, lineage across Databricks and Fabric (Power BI) removes the guesswork from impact analysis. For the data team, Dawiso is now the single source of truth they can build the next phase of their data strategy on.

​

“What changes the day-to-day is that our governed metadata is now something agents can reach, not just people - through the Databricks MCP integration, analysts can query lineage and definitions in Dawiso directly from their workflow. The close dialogue with the Dawiso team is what made the pilot work: when our model didn’t fit the tool, they adapted the tool to us.”

Daniel Fallenius, Information Architect, Bertel O. Steen.

​

Next Phase

BOS continues to expand the platform. Discussions are underway on data quality documentation, KPI management, and a deeper rollout of Data Products as Dawiso ships the next iteration of its Data Products module, including data contracts and access workflows. The user community will also be progressively onboarded through Single Sign-On as the platform graduates from PoC into long-term use.