Your EA data in Power BI, Microsoft 365 Copilot, Microsoft Fabric or your data warehouse
From model to Power BI in a few hours instead of weeks. Four ways to get there, and the one that fits depends on where your repository lives and who needs to read it.
A copy of your repository, laid out the way your business talks
A reporting database is a copy of your Sparx EA repository organized the way your business talks about it: a table of applications, a table of capabilities, a table of which application supports which capability. Each column is something you would say out loud, such as criticality, owner or regulatory scope.
The repository itself is organized the way EA stores models, as generic objects and connectors in a handful of tables. That layout is right for modeling and wrong for reporting, which is why nobody can report from the repository directly, and why Power BI, Microsoft 365 Copilot and your warehouse need the business layer in between.
From weeks to a few hours
Getting EA data into Power BI used to be a project. Someone had to learn how EA stores a model, write the SQL that turns generic objects and connectors into tables a business reader would recognize, define every relationship by hand, and rebuild it all when the modeling language changed. Power BI’s own relationship detection made it worse: on one 40-table model it proposed 63 relationships where the model needed 38, inventing 28 and switching off one that was right. Most teams never got past the first report.
With AI-Augmented Architecture the architect describes what the business needs to read, and AI builds the business layer from the modeling language that governs the model: one table per stereotype, one per allowed connector combination, every column named the way people say it, every cell checked against an independent read of the repository. It then generates the Power BI project with the relationships declared, so Power BI never has to guess. The architect opens it in Power BI Desktop and starts on the visuals the same day.
In our experience that is the difference between a few hours and several weeks, and it holds whether the model has forty tables or four hundred. On the Westbrook Bank reference model we use for teaching, the business layer is 18 entity tables and 23 connector tables, built in under three seconds, with 1,456 cells verified against EA. A small model, so treat it as the shape of the result rather than a figure to expect.
Pick the path that fits where your repository lives
The free AI Power Tools for Sparx EA offer three paths to the same business layer, so a report built on one works on another. Kernaro AI Hub from Sparx Systems is a fourth, and the one we recommend when the budget allows. Which one fits depends on your database, your access to it, and whether readers need live data or a refreshed snapshot.
- Live views in your EA database
SQL views built in their own schema inside your EA repository’s SQL Server database, with EA’s own tables untouched. The data is live: Power BI reads it in DirectQuery, or takes a snapshot in Import mode. Fits when you host the repository on SQL Server and can grant rights to create views.
- A separate reporting database
The same views in a separate SQL Server database that AI Power Tools fills with a copy of EA’s tables, rebuilt in full on every refresh. Report queries never touch the EA database. Fits when the repository must stay isolated, or when reporting readers should not have access to it.
- Parquet files
One Parquet file per table in a folder, a snapshot rewritten on every refresh, read by Power BI in Import mode. Fits when you have no SQL Server, when the repository is hosted elsewhere, or when the files are headed for a data warehouse. A published model cannot refresh from local files.
- Kernaro Connector for Power BI
Kernaro AI Hub exposes your Sparx repository through a read-only GraphQL endpoint, and the Kernaro Connector for Power BI reads it live: no intermediate database and no batch refresh, with the same endpoint serving Microsoft 365 Copilot. It is a licensed Sparx Systems product, and it is our recommended route for Power BI and Copilot for any organization that can afford it.
A refresh on the AI Power Tools paths runs as one operation that a scheduled AI agent can start and check, with EA open on that machine. Whatever the path, every answer carries a statement of how complete the data behind it is, because a roll-up looks equally authoritative at 12% fill and at 100%.
The model answers, and architects stop being the reporting bottleneck
The same answer every time it is asked. Each answer is read from the model, not rebuilt from a spreadsheet, so last quarter’s question and this quarter’s give comparable results.
No architect in the middle. Risk, finance and portfolio teams ask in the tools they already use and get the answer without a request to the architecture team.
How complete the data is. A roll-up looks authoritative whether a field is filled in on 100% of applications or 12%. You get the answer and a statement of how complete the data behind it is.
We build the reporting database and get the result where it is needed
Build it for you
Consulting for organizations that are capacity constrained or need results quickly. We choose the path with you, build the business layer on your Sparx EA repository, generate and finish the Power BI semantic model, build the dashboards, and hand it all over with the refresh in place.
We build and update your architecture model for youGet it beyond Power BI
Microsoft 365 Copilot answers, read-only, from the published Power BI model. The reporting database or its Parquet files can feed Microsoft Fabric, Snowflake or your data warehouse, so EA data sits next to the finance and operations data it is read against. We set up whichever route fits.
Use the Kernaro route when you can
Kernaro AI Hub is our recommended route for Power BI and Microsoft 365 Copilot when the organization can fund it: live reads through its GraphQL endpoint, no intermediate database, no batch refresh. We help you set up Hub, the Kernaro Connector for Power BI and its dashboards rather than building a second path.
Kernaro AI HubTrain and mentor your architects
Training and mentoring that teach your architects to build and run the business layer and the Power BI model themselves, on your own repository, so the next report does not need us.
Training and mentoring for architects