Episode 312 – Power BI November 2025 features: Copilot Everywhere, MCP Servers & Modernized Visuals
Recording Date: December 10, 2025
Back from Ireland and Ready for November’s Updates
Jason and John had just returned from ESPC (European SharePoint Conference) in Ireland, where they delivered a brand-new session on Power BI that left a lasting impression. One attendee approached them afterward with a powerful observation: “I was scared of Power BI before. I’m not scared anymore.” It’s a reminder that even after 10 years in the market, Power BI remains unfamiliar to many professionals, and there’s real value in making it more approachable.
Both hosts also hinted at exciting announcements coming in 2026—something they’re actively planning that will be worth watching for on LinkedIn and the blog. But for now, they’re diving into the Power BI November 2025 feature summary, and there’s plenty to unpack.
This month’s coverage starts with Power BI instead of Fabric, which represents a substantial update. Note that the Power BI section will be covered in this episode, while Fabric features will span the next episodes, so stay tuned for the complete picture.
A Quick Clarification: ARM 64 Native Power BI Desktop
Before diving into the official announcement blog post, John wanted to clarify something from the October update that caused some confusion: the ARM 64 native Power BI Desktop announcement wasn’t quite what it sounded like.
Here’s what actually happened: Windows improved its Prism X 64 emulator to better support the Analysis Services engine that powers semantic models in Power BI Desktop. In other words, Power BI Desktop still runs in emulation mode on Windows on ARM, but it runs noticeably better now. The emulation improvements mean features that previously didn’t work smoothly are now performing well.
The bottom line for Mac users running Windows VMs in Parallels? You don’t have a native ARM 64 Power BI Desktop option, but the newer Windows versions are plenty fast enough, and you’re not missing an experience. If you’re currently using Power BI Desktop on ARM Macs via emulation and haven’t been running the latest Windows updates, an upgrade will likely improve your experience.
The Service Is Becoming a Viable Workbench
Here’s something worth paying attention to: the Power BI service now has a complete Get Data experience with full Power Query capabilities built in. This means you can create a semantic model and report entirely in the Power BI service without ever opening Power BI Desktop.
John emphasized that while there are still some things easier or more available in Desktop, the gap has narrowed significantly. The service now provides a full, end-to-end experience. This is especially important for Mac users who’ve been waiting years for a native Mac version of Power BI Desktop. That’s not coming, but the need for it just decreased substantially.
One thing John hasn’t yet found: a way to turn off automatic time intelligence when creating reports this way in the service. That’s something worth investigating, and the team will explore it in upcoming sessions.
John noted that this capability hasn’t received nearly as much attention as it deserves. Expect a bigger announcement around this in the near future, because it genuinely changes the accessibility story for many users.
Copilot Expands to Mobile and Beyond
Ask Anything Anywhere: Copilot on Mobile
The November update brings Copilot to the Power BI mobile apps (both iOS and Android) in preview. You can now ask natural language questions directly from your mobile device and get answers from your Power BI environment. On iOS specifically, you can also dictate your questions using the microphone feature—a convenience feature not yet available on Android.
This replaces the older Q&A feature, which has been deprecated. The Copilot approach is simply better. Rather than asking questions of a flat repository of information, you’re asking it to build and execute queries against your actual data model, whether that’s a warehouse, lakehouse, or Cosmos database.
Service Copilot Gets Smarter
In the Power BI service, Copilot continues to improve. The homepage is being redesigned with Copilot as a primary entry point. Jason appreciates that you can still flip back to the classic homepage view if preferred. Meanwhile, the Copilot experience itself is getting smarter about figuring out what data to use and asking clarifying questions before returning answers. You can now attach existing items to queries to narrow scope, giving the copilot better context to work with.
John highlighted an important principle: “I’ll put about as much effort into my answer as you’ll put into your question.” Blanket questions without refinement aren’t holding Copilot wrong—they’re holding it right, just without expecting perfect results.
Report Copilot Improvements
Report Copilot (the feature that generates reports from semantic models) has received improvements focused on visual selection and overall performance. The blog post description is brief, but the enhancements are noted as “bigger, better, faster”—a significant update that makes report generation more reliable.
Verified Answers Get Better
Verified answers now inherit the full visual state from your queries, allowing you to drill down and explore further instead of just getting a single answer and stopping. You can also use the new card visual for verified answers, and Azure Maps is now supported as a visualization option for location-based answers.
App Copilot Launches
When you publish a Power BI app, users can now ask natural language questions directly within that app’s context without jumping back to reports, workspaces, or standalone copilot experiences. Copilot is increasingly becoming a first-class feature throughout the Power BI ecosystem.
Understanding MCP Servers: Two Different Tools
The November update brings two MCP (Model Context Protocol) server announcements that serve different purposes. Understanding the distinction matters.
Remote MCP Server for Semantic Models
First, Microsoft is releasing a remote MCP server (hosted by Microsoft) that allows any AI tool—Copilot, Claude, Gemini, or others—to better reason over your semantic models. Think of MCP as an API for AI: it gives AI agents more context about your data.
Here’s what this server enables: it provides access to your semantic model schema, can generate queries based on natural language, and can execute those queries. The fundamental difference from other AI interactions is that this is specifically designed for querying data models rather than searching flat repositories of information. Whether your data lives in a warehouse, lakehouse, Cosmos database, or elsewhere, the MCP server translates natural language into appropriate queries.
Practically speaking, you could now fire up Claude or another external AI tool, supply it this MCP server, and ask it questions about your Power BI semantic model directly—no need for the standard copilot experience.
Local MCP Server for Semantic Model Development
The second MCP announcement is entirely different: a local Power BI modeling MCP server that runs in Visual Studio Code and allows developers and AI applications to modify semantic models using natural language. Want to rename measures across your entire model? Just ask. Need to add a year-over-year analysis dimension? Describe it in natural language and the server handles the technical implementation.
Jason raised an important consideration: this capability lives in the hands of developers using tools like GitHub Copilot programmatically, not end users casually creating semantic models. That said, the governance implications are worth thinking about. If an external tool can programmatically create semantic models, you’ll need clear governance and cleanup processes to manage the inevitable proliferation of experimental models.
Visual Updates: The Card Gets Its Moment, the Image Finally Modernizes
Card Visual Goes GA
The new card visual is now generally available in both the Power BI service and Desktop. It’s a significant upgrade from the original card visual, offering multiple display options and much more flexibility. John noted it’s now prominently displayed in the visuals pane instead of hiding at the bottom, making it more discoverable. The original card visual is still available if you need it—you can restore default visuals to bring it back.
Image Visual Gets a Complete Modernization
This is genuinely exciting: the image visual, which hasn’t received meaningful updates since Power BI’s inception 11 years ago, has been completely modernized. Previously, it was static—you could place a logo or image on a page, but that was about it. The image wasn’t even available in the service.
Now? The image visual is data-aware and fully data-bindable. You can hover over images to trigger state changes, borders can respond to filters, and the entire visual behaves like a modern Power BI component. It’s a substantial upgrade that opens new possibilities for dashboard design.
Reporting and Modeling Updates
Matrix Columns Auto-Expand
Following last month’s ability for tables to automatically expand to fill available space, matrices now get the same treatment. This improves readability and layout efficiency, especially when combined with the growth-to-fit option. The blog post references documentation on this behavior and how it interacts with features like hidden columns and expansion controls.
UDFs and Trans-Policy Task Flows Integration
Power BI’s integration with User-Defined Functions (UDFs) continues to improve. You can now select UDFs from the One Lake catalog when building trans-policy task flows in Power BI Desktop. This makes the write-back experience smoother.
That said, there’s opportunity for improvement here. John and Jason appreciate the capability but acknowledge that the user experience could be more unified. Currently, you’re building UDFs outside of the Power BI context (requiring Python knowledge), then calling them from Power BI. If Microsoft could develop a more integrated workflow—similar to how Power Query’s Get Data experience works—this feature could see broader adoption. For now, it’s a useful bridge between Power BI and Fabric capabilities, and the groundwork is solid.
Semantic Model Version History GA
Semantic model version history is now generally available in the service. Since every change you make in the service is immediately committed, having version history provides crucial undo and audit capabilities. This is particularly valuable as semantic models become more complex and multiple people work on them simultaneously.
TMDL Visual Studio Code Extension Goes GA
The Tabular Model Definition Language (TMDL) extension for Visual Studio Code is now generally available in the Visual Studio Marketplace. If you’re working with semantic models programmatically or building them as code, this extension should be part of your toolkit. It’s a powerful way to version control and collaborate on model definitions.
Connectors, Visualizations & Deprecations
Connector Updates
As with most months, there are multiple data connectors moving to GA or preview status. Check the official blog post for the specific list, as new connectors arrive regularly based on customer demand and platform priorities.
Third-Party Visualizations
The visualization marketplace continues to grow. While there are excellent third-party options available, the general recommendation remains to use native visuals unless you have a specific need. The decomposition tree remains a popular choice for exploratory analysis, but native visuals are the default for a reason—they’re performant, well-integrated, and receive ongoing Microsoft support.
R and Python Visuals Deprecation for Embedded Solutions
The November update includes notification of an important deprecation: R and Python visuals will no longer be available in Power BI Embedded (the solution for ISVs embedding Power BI in their applications). This is part of a broader effort to deprecate these visual types across the platform.
The reasoning makes sense from multiple angles. Usage of R and Python visuals has been relatively low. When organizations need advanced statistical or data science capabilities, notebooks in Fabric provide a more efficient and modern approach. Those same capabilities that people might have used R or Python visuals to achieve are now better served through other tools in the Microsoft ecosystem. For Python users specifically, there are more performant environments than Power BI visuals for implementing complex logic.
What’s Next: Fabric Features Coming
This episode covers Power BI’s November updates comprehensively, but the feature summary spans multiple products. Fabric updates will follow in upcoming episodes—the hosts aren’t committing to a specific number of episodes for Fabric coverage, as it depends on the depth of discussion needed. Expect at least two episodes for Fabric, potentially more given the breadth of announcements from Ignite and the monthly updates combined.
The Bigger Picture
The Power BI November updates reflect a platform that’s evolving thoughtfully. Copilot is becoming deeply integrated, the service is becoming more capable for people who don’t use Desktop, visuals are being modernized, and developer tools are improving. It’s not flashy, but it’s substantial.
The one through-line Jason and John consistently emphasize: talk to your colleagues about what’s interesting to your specific scenarios. Not every feature matters to every organization. But having these conversations—understanding what’s available and how it might apply to your work—that’s where real value comes from.
Related Episodes
- Power BI November 2025 Feature Summary
- Episode 311 – Microsoft Ignite Recap: Fabric Databases, Data Agents & What Actually Matters
- Episode 302 – Power BI August 2025 Feature Summary
- Episode 304 – Microsoft Fabric August 2025 Feature Summary
Music: “Indie Rock” by Scott Holmes, shared under Creative Commons
One Reply to “Episode 312 – Power BI November 2025 features: Copilot Everywhere, MCP Servers & Modernized Visuals”