Power BI Weekly

Issue #360 - 2nd June 2026

It was Microsoft Build this week, with a few announcements from the Power BI team, mainly focused on making the semantic model the foundation for AI agents.

There are two posts from the team worth reading together. Building in the Agentic Era with Power BI and Fabric covers two new developer-facing features: Agent Skills for Power BI (preview), which lets you use an AI agent to build semantic models and reports end-to-end via Power BI Projects, and Fabric Apps for Semantic Models, which lets developers build Fabric-native web apps on top of semantic models. The second post, Power BI at Microsoft Build 2026: The Agentic Era of analytics, covers all four Build announcements and is worth a skim for the full picture. Both posts make the case that without a well-structured semantic model, AI agents are just guessing, which is a useful framing even if agentic workflows aren't something you're thinking about yet. The new Org Apps experience is also worth a look if you're rolling out analytics at scale.

Also this week: DAX User-Defined Functions are now generally available. UDFs have been in preview since September 2025 and are clearly catching on. With GA, you can write a calculation once and reuse it across measures, calculated columns, and visuals, with inline documentation that surfaces in IntelliSense and type safety at runtime. They'll be on by default in the June 2026 Power BI Desktop release, coming soon. Also out this week is the on-premises data gateway May 2026 release, which adds admin consent controls for gateway diagnostics and plugs gateway diagnostics into the Dataflow Gen2 run experience to make troubleshooting a bit less painful.

On the community side, my colleague Carmel Eve has started a new series on the endjin blog. Optimising DAX: A Series Introduction sets out what's coming: VertiPaq storage, compression, how the storage and formula engines work together, and the model design decisions that underpin all of it. The content comes from a full-day SQLBits 2026 workshop by Alberto Ferrari of SQLBI, and Carmel has written it up to share what she learned. The second post, Optimising DAX: How VertiPaq Stores Your Data, gets straight into the column-store basics: VertiPaq stores data column-by-column rather than row-by-row, which is why aggregations are fast and cross-column queries need more care. Worth bookmarking as the series continues. Finally, Ned Charles has put together a video on setting up custom Power BI deployment workflows, showing how you can build on top of deployment pipelines using notebook widgets to customise your deployment gates and add things like test suites before content gets promoted.