Recent News in Analytics and AI: June 2025 Edition

This month's summary of recent news in analytics and AI showcases two new Copilot experiences in Power BI, updates to AI functions in Microsoft Fabric, two key Model Context Protocol (MCP) launches to take note of, the announcement of MLflow 3.0, and more. Each section presents four analytics and AI news highlights.

Read on and get up to speed.

Power BI

• Two new Copilot experiences have been launched in Power BI. As announced at Build 2025, the 'Chat with your Data' experience is now available as a preview. You can also access Copilot in your securely embedded Power BI report. It's still early days and you can expect these Copilot experiences to get better and better from month to month. Learn more.

• Power BI Desktop is now able to handle any semantic model in Fabric. This is a huge milestone as previously there were some limitations that meant you'd have to use the web experience for some Direct Lake related changes. Now, even if you have models with multiple partitions and tabular data sources, you can open them directly in Power BI Desktop. This gives you the ability to add new tables or edit query expressions. Learn more.

• Power Query editing in the web for import models is slightly delayed. Issues came up during the rollout, meaning it is now likely to be available this month. This update is going to allow you to make changes to existing import models with Power Query, directly in your browser. You'll be able to add new import tables, edit queries, and refresh data directly on the web. Learn more.

• Microsoft is celebrating being named a Leader in the 2025 Gartner Magic Quadrant for Analytics and Business Intelligence Platforms. Why should you care? Ranking furthest yet again for Completeness of Vision and highest in Ability to Execute in the Magic Quadrant for the seventh year running, you know you're using the best of the best. Learn more.

Microsoft Fabric

• OneLake security capabilities now support mirrored data with Azure Mirrored Databricks Catalog. This allows you to access the full range of OneLake's enterprise-grade security features across the mirrored assets at the table, column, or row level. It gives you the added benefit of extra security in your downstream sources, like lakehouses or semantic models. Learn more.

• New updates have been released across AI functions. In response to user feedback following the public preview earlier in the year, Microsoft has now upgraded the default model that powers AI functions to a more intelligent model. They've also preinstalled AI functions on Fabric 1.3 runtime, meaning you no longer have to manually install it. All in all, this update gives you more powerful AI functions at a more cost-effective price point. Win-win! Learn more.

• Microsoft Fabric has released Model Context Protocol (MCP) support for Real-Time Intelligence (RTI). The MCP protocol, developed by Anthropic, enables AI models like Azure OpenAI models to interact with other tools. Now, with the RTI integration, you get to receive data and AI-driven insights in real time. This gives you an enhanced and more seamless data query and analysis functionality. Learn more.

• If you've been wanting to get the best out of your data agent, Microsoft has now released a best practices guide to help you do it. Forget trying to figure it out on your own; the guide shows you how to get a more reliable performance out of your agents. Read on to get started. Learn more.

Microsoft 365 Copilot and Copilot Studio

• It's no secret that generative AI is opening up new opportunities for the way we work. But this can also mean a lot of complexity where regulations are concerned. If you're looking to keep your security, privacy, and compliance in check, the Microsoft 365 Copilot team recently shared some guidance on their blog. It drills down on exactly how to build privacy, security, and governance into your AI operations from the start, helping you to stay compliant. Learn more.

• You can now publish agents from your Microsoft Copilot Studio into Microsoft 365 Copilot for full use within the app. In both web and desktop formats, along with Microsoft Teams, this opens up agent assistance in your everyday work. It gives you the opportunity to automate workflows and streamline processes for greater efficiency across all business areas. Learn more.

• The MCP integration is generally available across Copilot Studio. This means adding AI apps and agents in Copilot Studio is just a few clicks away. Thanks to this capability, your agents can be automatically updated with the latest actions and information for a faster way to build and scale your agentic workflows. Learn more.

• A new feature has been added to Copilot: 'Copilot Memory'. This is a function that remembers facts about its users. From their preferences to favoured work style to details picked up in conversations, Copilot will automatically apply them. A prime example is preferred formatting. This is a game-changer for Copilot, making it much more efficient and tailored to your personal needs. Learn more.

Azure Analytics and AI

• Apache Iceberg v3 has evolved with new features and data types. Huge improvements have been added, including deletion vectors and new types for semi-structured data. This is going to give you a much more efficient experience when processing and querying your data. Learn more.

• MLflow 3.0 has been officially announced. It combines traditional ML, deep learning, and GenAI development, meaning you no longer need to use separate tools. Available in open source MLflow and as a fully managed option on Databricks, MLflow 3.0 allows you to create more consistent workflows and offers improved reliability. Learn more.

• Databricks is donating Declarative Pipelines to Apache Spark. This declarative API is worth its weight in gold when it comes to building robust data pipelines. Reducing the work required and simplifying the process, giving you the ability to work on full pipelines in Apache Spark, instead of single queries. Learn more.

• Data virtualization for Azure SQL Database is now in public preview. This gives you the ability to leverage the power of Transact-SQL (T-SQL) to directly query data in Azure Data Lake Storage Gen2 (ADLS Gen2) or Azure Blob Storage. Easily one of the most highly request features in Azure SQL databases, it removes or reduces the need to duplicate data, giving you a much faster ability to analyse your data and gain insights. Learn more.

Open-Source Analytics and AI

• The final release of Delta Lake 4.0.0 is here. Along with it comes a bunch of new features across both Delta Spark and Delta Kernel Java. Some of the most notable include preview support for catalog-managed tables and the Delta Connect extension for Spark Connect, both of which allow for greater efficiency. Learn more.

• Polars has collaborated with NVIDIA to supercharge its power and push it's already high-performing nature with the integration of a GPU engine, with the aim of accelerating compute-bound queries. Using one parameter, you can trigger GPU execution with barely any code changes. If you're looking for more power in your GPU engine, this could be it. Learn more.

• Meta has introduced the V-JEPA 2 world model. Meta Video Joint Embedding Predictive Architecture 2 (V-JEPA 2) is the first of its kind, trained specifically on video. It enables understanding and prediction in the physical world and is their next step in achieving advanced machine intelligence (AMI). It gives your AI systems the ability to learn about the world in the same way humans do, planning and adapting as they develop. Learn more.

• DuckDB's high-performance SQL engine is the perfect tool for simplifying and streamlining your text analytics in Python. Their latest guide is walking you through exactly what to do to harness their keyword, full-text, and semantic similarity search with embeddings capabilities. This is perfect when you want a more streamlined, simplified process for text analytics use cases. Learn more.

Stay in the Know

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