Recent News in Analytics and AI: April 2025 Edition

This month's summary of recent news in analytics and AI uncovers some of the latest advancements in Power BI and Microsoft Fabric, specifically within both org apps and Dataflow Gen2, dive into a recap of FabCon if you couldn't attend, explore the release of Apache Airflow 3.0, discover the first Llama 4 herd models from Meta, and more. Each section presents three analytics and AI news highlights.

Read on and get up to speed.

Power BI

• Microsoft announced the ability to create Direct Lake semantic models in Power BI Desktop. This means you now get the latest OneLake data to help you delve quickly into insights inside Power BI without duplicating data, thanks to the new Direct Lake storage mode. This means you can now add in tables from a range of data sources, which gives you a lot more flexibility with OneLake data. Learn more.

• The Microsoft Fabric Community Conference (FabCon) took place last month and saw a record-breaking year for attendance. While over 6,000 people attended, if you missed it, don't worry. You can watch the FabCon Las Vegas keynote recording to see it all in action. There were lots of Power BI announcements. Learn more.

• In April, further updates to org apps (Preview) landed. New combined navigation options help you customize your apps, which gives you more control over the consumer experience. There's also Git integration for developers, which gives you access to backups and version history, and more. Learn more.

Microsoft Fabric

• There's been a key update to the query parameters in Dataflow Gen2. Public parameters in Dataflow Gen2 with CI/CD support mean that you get a more dynamic experience where different parameters can be set for each refresh, giving you more flexibility and control. Learn more.

• Microsoft's internal Azure Data team has used Microsoft Fabric to develop a data engineering solution over the past few years. They're now sharing their insights on how they refined their approach to Continuous Integration and Continuous Deployment (CI/CD). Their journey gives you strategies for Fabric workspaces, a Git branching strategy, and more to help you apply their findings to your own CI/CD experience. Learn more.

• Using Microsoft Fabric, you can now design and implement an end-to-end metadata-driven Lakehouse. With a well-architected lakehouse, you have the ability to manage and process your data for things like analytics and machine learning more efficiently. Learn more.

Microsoft 365 Copilot and Copilot Studio

• After last year's announcement of agents in Microsoft 365, new experiences have been announced. There is now the Agent Store, which allows you to find the right agent for your needs for a more personalized experience. There's also the new Skills Agent, based primarily around the People Skills data layer to help you harness workplace expertise, and more. Learn more.

• With the new Microsoft 365 Copilot Wave 2 Spring release, innovations in Copilot Control System (CCS) have been announced. You'll benefit from improved security and governance thanks to the new Microsoft Purview Data Security Posture Management for AI, which will preview in June 2025. Learn more.

• Microsoft Copilot Studio has announced the launch of computer use. This means that you'll now have the ability for your Copilot Studio agents to use both websites and desktop applications as tools. Computer use means your agents can work with any system with a graphical user interface. It gives your agents the power to work on anything from data entry and market research, without the need for an API to be available. Learn more.

Azure Analytics and AI

• Databricks has announced the launch of two new assets. Streaming Tables (STs) are great for real-time data pipeline usage thanks to their ability to continuously ingest streaming data. Then, by pre-computing and storing query results, Materialized Views (MVs) are perfect for enhancing performance in SQL analytics and BI dashboards. Why does this matter? This significantly enables you to improve performance and keep costs down while sharing fresh and relevant data. Learn more.

• Microsoft announced the latest additions of the OpenAI o3 and o4-mini models on Microsoft Azure OpenAI Service in both Azure AI Foundry and GitHub. They bring new developments in safety and quality, particularly within the reasoning feature. This is useful to know if you're looking for improved performance and the ability to process and analyze visual data. Learn more.

• The first Llama 4 herd models, developed by Meta, have been announced in Azure AI Foundry and Azure Databricks. This is going to give you the power to create completely personalized multimodal experiences. If you're a developer, you can now bring Llama 4 models into your apps that have heavy data sets to create summaries or reports. The benefit? Increased efficiency, personalization, and reasoning. Learn more.

Open-Source Analytics and AI

• Apache Airflow 3.0 has been released, four years after 2.0. But why is this a major milestone and relevant to you? The updates have been significant – even since the 2.10 release in December 2024. The biggest takeaway is that it's now easy to use, as requested by data practitioners. Enjoy fewer restrictions during development and a streamlined transition in navigating between workflows. Learn more.

• Polars has shared a case study with METRO.digital. It details their experience in creating Key Performance Indicators (KPIs) across their 19 countries while moving away from SQL and over to Python. If you want more efficiency, readability, and maintainability with your machine learning or business and product KPIs, give this one a read. Learn more.

• DuckDB conducted an interesting evaluation of reading CVS files with the Pollock Benchmark. No surprise, their CSV Reader came out top. Take a look at their findings if you want a fast and robust solution for your CSV files. Learn more.

Did You Find This Useful?

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