Recent News in Analytics and AI: February 2025 Edition

This month's summary of recent news in analytics and AI touches on improvements to the Copilot experiences in Power BI and Microsoft Fabric, articles that provide some practical applications of Copilot, the launch of AI Foundary Labs, the announcement of an updated AI security framework from Databricks, and more. Each section presents three analytics and AI news highlights.

Read on and get up to speed.

Power BI

• The Copilot experience for DAX queries in Power BI received significant improvements. Copilot has started using additional context derived from descriptions, synonyms and sample values in your semantic models to translate your natural language queries into DAX. What this means for you is the results are more likely to present what you want. Learn more.

• If you frequently access your Power BI reports from an iPad or Android tablet, you'll be pleased to know that Copilot is available in the respective mobile app. Learn more.

• The report view Copilot chat pane may not work correctly after 4th April 2025 on Power BI Desktop versions older than the February 2025 update. If you use or plan to use this feature, you'll need to ensure you upgrade before then to continue using it without disruption. Learn more.

Microsoft Fabric

• The OPENROWSET function arrived in Fabric Data Warehouse. If you've previously worked with Azure Synapse Serverless SQL pools, this function will feel familiar to you. If not, a summary of what it does is allow you to quickly query data in Parquet or CSV format stored in Azure Blob storage or ADLS Gen2 without needing to create any database objects. The feature is currently in preview status but, regardless, is a handy tool in your T-SQL toolbox. Learn more.

• Copilot and AI Skill in Microsoft Fabric were both enhanced to improve the accuracy and relevance of their responses, and introduce context knowledge retention. This means the AI can not only recall details from your previous conversations but in the background apply chain-of-thought reasoning to produce richer and deeper responses to your prompts. Learn more.

• One of the most significant points of frustration when working with Fabric Data Warehouses from Spark notebooks in Fabric is the inability to write to it. Fortunately, that's an issue no more, and now you can both read and write to Fabric Data Warehouse from Spark using the Spark Connector for Fabric Data Warehouse. Learn more.

Microsoft 365 Copilot and Copilot Studio

• The Microsoft Copilot team published an article that explores how you can use Copilot to perform a technical analysis of a stock using stock market data. The article presents an example with prompts, the results they produced, and the rationale behind each of the steps taken. Learn more.

• Have you wondered how you could use Copilot to help you organise projects and tasks? A recent post on the Copilot blog explores precisely that. It offers a few examples of how, through prompting, you can describe the task at hand and then break it down into smaller, more manageable chunks. Learn more.

• There were many updates to Copilot Studio in February, and one of the major ones is the ability to import and export agent components. Although it doesn't sound as exciting as many other features, this one means you can create Component Collections and use them as templates for new Copilot agents, enabling you to standardise some agent characteristics while accelerating the creation process. Learn more.

Azure Analytics and AI

• AI Foundry Labs launched, a place for you to be inspired by the art of the possible in bleeding-edge AI. There, you will find models and agentic frameworks created as a result of Microsoft's research-driven AI efforts. From models that can forecast weather and air pollution to multi-agent systems that solve complex problems by orchestrating multiple agents. You'll be sure to find something in the AI Foundry Lab assets catalogue that makes you think, 'Wow!'. Learn more.

• Muse is a generative AI model for gameplay ideation with a detailed understanding of the 3D game world, physics, and player inputs through controllers. Ultimately, this new AI technology will empower game creators to try new ideas faster and bring new gaming experiences to the market quicker. What's even more impressive is that this technology is available to you on Azure AI Foundry now. Learn more.

• Databricks recently announced the Databricks AI Security Framework 2.0. It builds on existing standards and presents 12 AI system components, which each have their own risks and mitigation controls. Identifying 62 technical security risks and 64 recommended controls, it includes mappings to industry standards such as ISO 42001, ISO 27001:2022, and the EU AI Act. Learn more.

Open-Source Analytics and AI

• Velox is an open-source library created by Meta that can be used to optimise the distributed processing of data in frameworks like Apache Spark and Presto. For example, Microsoft uses it as a core component in its Native Execution Engine for Fabric Spark. The team behind the library has started a short blog series that will introduce you to the internal structures and concepts in Velox. Learn more.

• DuckDB can security read and write to Google Sheets using a DuckDB Community Extension. In a recent article, the DuckDB team show you what this looks like and you might find it useful if you happen to be using Google Sheets today. Learn more.

• Polars Cloud is still under development and holds lots of promise. The team behind it published an article that explores how it addresses a gap in scalable DataFrame implementations, how Polars Cloud will be multi-cloud, its support for running a query remotely, its out-of-core design, including CPU and CPU support, and a lot more. Learn more.

Did You Find This Useful?

Get notified when we post something new by following us on X and LinkedIn.