Notes from the Power BI Gebruikersdagen 2026
At the Power BI Gebruikersdagen 2026 I attended several sessions on modelling, governance, AI and the evolution of the Power BI platform. As with most events like this, the most interesting insights were not only in the big announcements, but in the presentations and demonstrations that show what you can already build with the tooling today. This page is about AI and the ways the Power BI platform is becoming increasingly programmable and interactive.
AI and Power BI in practice
Several sessions made clear that AI is not just an add-on to Power BI — it is also changing how you work with the platform, not only through chat interfaces but also through development tools, automation and agents. There are roughly three directions in which this is already developing.
Generating Power BI models via MCP servers
One of the more interesting demonstrations showed how you can use an MCP server and a language model to generate a Power BI model. Instead of manually building tables, relationships and measures, you provide a description of your dataset and the analysis you need, and the AI can:
- propose tables
- set up relationships
- generate measures
- produce an initial semantic model
This happens for example through Visual Studio Code integrations where a model is generated in code. The result is not a finished product, but it does accelerate the initial model design. For people who understand what a good model looks like this can save a lot of time; for people who do not, it can quickly become confusing.
Talking to data via Fabric and data agents
A second development that received a lot of attention is the use of AI agents within Microsoft Fabric, which allows you to interact with data through a chat interface. Fabric can then for example:
- answer questions about datasets
- summarise reports
- explain trends
- suggest analyses
These agents run on top of semantic models, which means the quality of the data model still determines how useful the answers are. A well-built dataset produces useful answers; a messy dataset produces messy answers. AI does not reduce the importance of modelling — if anything, it makes that importance more visible.
Building custom visuals with AI and Visual Studio Code
A third interesting development sits on the development side of the platform. Power BI has supported custom visuals for some time via TypeScript and the Power BI visuals framework in Visual Studio Code. With AI tooling that is now considerably more accessible: you can use AI to generate code for a visual, which in principle means you can build whatever you need, such as:
- new visualisations
- interactive components
- specialised interfaces
Technically you can even build small games inside a Power BI visual — demonstrations sometimes show examples like Tetris or Snake. That mostly shows how flexible the platform is, but it does not automatically mean every application is also useful.
AI makes things faster, not automatically better
What all these examples show is that AI speeds up the process of building solutions, but the quality of the result still depends on the foundation:
- a solid data model
- clear definitions
- governance
- understanding of the context
AI can generate a model, but it cannot always judge whether the model makes sense. AI can answer questions, but only based on the data that is available. That is why it remains important to understand where AI connects to the platform.
Where AI adds value
In practice the most value shows up in situations such as:
- prototyping datasets more quickly
- generating first drafts of DAX
- explaining data to users
- automating documentation
These are accelerators — they do not replace the thinking.
Reflection
The Power BI platform is becoming more flexible. You can generate models with AI, talk to data through agents, and even build your own software inside visuals. All interesting developments — but ultimately they all run on the same core: data, models and definitions. When those are solid, AI helps you move faster. When they are not, AI mostly accelerates the confusion.
Building that foundation is exactly what I focus on in the custom Power BI training I provide. Because however the tooling evolves — modelling, DAX and understanding data structures remain essential.
The other topics in this series are on the overview page.
A demonstration
Below is a small tic-tac-toe game. Not to prove anything, but to show that web technology is flexible enough to build something like this. The same applies to a Power BI custom visual.