Fabric Apps as a replacement for corporate Power BI: what works and where it breaks

Fabric Apps has been in preview since June 2026 and is already enough to raise an uncomfortable question: whether a corporate dashboard still has to be a Power BI report, or can be an application built on the same semantic model. Several weeks of real use, on one BI case and on an internal one that has nothing to do with BI, are enough for a provisional answer.

First, a word about how the work is done, because it is the part that draws the most noise: Fabric Apps is code-first. There is no canvas and no drag and drop. The app is written in code, today with an assistant such as Claude applying the logic it is asked for and iterating in conversation. Vibe coding, in short. The detractors already have their argument; what matters is what comes out the other side.

And what comes out is a level of experience and visualisation that Power BI almost never fully reaches, or simply cannot given the nature of the tool.

What already works

• A multilingual app on a single semantic model. Four languages, one logic, four interfaces. No duplicated reports or measures.
• Production changes in seconds. Redeploy and done. No .pbix to publish, no Desktop, no waiting for anything to refresh.
• Charts from open-source libraries. Tooltips synchronised across charts, heatmaps with clickable cells, waterfalls, native light and dark mode. What in Power BI means a custom visual, a licence or a no.
• Interactive SVG floor plans. Every area of the plan is clickable and paints its figure on top. For a plant, a warehouse or a store it is a different way of looking at the business.
• Users enrich the data. They interact and contribute information that stays in the system. They do not just read: they write on top of what they see.
• Roles and access from the app itself. No portal, no admin ticket every time someone new joins.
• Export without artificial limits. Excel or copy and paste of the table behind every chart, PNG of the visual, PDF at full resolution.
• A real menu. With hierarchy, sections and navigation, not chained bookmarks.
• And room to iterate with no ceiling, because in the end it is code and the limit is whatever there is to build.

The caveats, for now

• It is in preview. Nobody knows what SKU cost an app like this will carry in production. Without that number, nobody should recommend betting on it.
• The app does not leave Fabric. Today there is no external link that works with a connection to the semantic model. Whoever wants to see it has to be inside.
• No outbound API connection. There is no way to plug in a language model API with the context of the measures and business logic to build a chat that tells each role what they need to know. It is the most promising piece and the one that is not there today.

Where it would break

And then there is the classic doubt of any early adoption: nobody can yet be seen running this in production at their company. It makes sense, it is in preview. But there it is.

The obvious debate, Power BI yes or Power BI no, matters little. What matters is where it would break if a super-dashboard in Fabric Apps replaced classic Power BI at corporate level. Three candidates:

1. Governance. A .pbix has a known lifecycle: workspace, certification, lineage. An app written in code needs a different kind of control, and it is not clear who provides it.
2. Cost. Until the SKU is known, any return calculation is a bet.
3. Adoption. Business users know what a Power BI report is. A new app, even a better one, has to be explained, trained and sustained.

The provisional answer: as an operational layer on top of a BI that already exists, Fabric Apps is already worth it. As a replacement for corporate Power BI, it depends on those three things, and none of them is solved with more code. This article will be revised when there are answers.