Recent News in Analytics and AI: July 2025 Edition

This month's summary of recent news in analytics and AI dives into updates on Power BI Enhanced Report Format (PBIR), info on the Fabric data agents and Microsoft Copilot Studio integration preview, the launch of Copilot Memory, a deep dive into why Azure Databricks is arguably the best flavour of Databricks, a use case run-through on the power of utilising Meta AI’s Llama in your business, and the release of DuckLake version 0.2 with updates you need to know about.

Read on and get up to speed.

Power BI

• We saw the Power BI Enhanced Report Format (PBIR) update come into play. This allowed Microsoft to address some of PBIR’s final limitations and get closer to general availability. Key things to note include compatibility with both the 'save a copy' and report usage metrics features, and full integration with the Power BI report REST APIs. Overall, this will give you better collaborative development in Power BI report development workflows. Learn more.

• Power BI Report Server (PBIRS) will become the default on-premises reporting solution for new versions of SQL Server. As Microsoft begins consolidating all current reporting services, no new SQL Server Reporting Services (SSRS) are going to be released. But don’t worry, you’ll be able to get info on specifics when SQL Server 2025 is in general availability. Learn more.

• From consolidation to retirement, it was announced last month that both Cognitive Services and Azure ML are set to retire by 15th September 2025. This means that as of 11th August, you won’t be able to create new models or apply transforms using cognitive services. Anything existing will refresh until the retirement date, but past that point, they won’t be supported on either Power BI Desktop or the Power BI service. Learn more.

• Organisational themes are now available across Power BI. This means that you’re able to create consistent branding and styling across your reports. This applies whether you’re creating reports manually or with Copilot. Read on to discover how you can incorporate a centrally managed look and feel into your report creation processes. Learn more.

Microsoft Fabric

• Autoscale Billing for Apache Spark in Microsoft Fabric is now generally available. This serverless billing model provides you with a lot more flexibility and transparency when you’re looking to run bursty Spark workloads at scale. Similarly to how Spark is billed in Azure Synapse and Azure Databricks, the Spark jobs can be run separately from your Fabric capacity and billable via usage-based model. This gives you a new option for greater cost efficiency. Learn more.

• The Mirroring for Azure Databricks Unity Catalog in Microsoft Fabric capability is now generally available. This secure integration now sutiable for production use, allowing you to access your Azure Databricks tables from Fabric with ease, and giving you a unified experience without having to duplicate your data. Read on to discover how Fabric and Azure Databricks customers are already harnessing this powerful integration. Learn more.

• This month, Microsoft also announced the new OneLake shortcut transformations capability in Fabric. This functionality creates a much more simplified process of converting your raw files (starting with .CSV) into Delta tables in OneLake. For several transformation scenarios, it gives you the option to skip complex ETL pipelines and instead transform your data in a few clicks with minimal setup. Learn more.

• One of the biggest recent announcement in Microsoft Fabric is a preview of the integration between Fabric data agents and Microsoft Copilot Studio. If you’re looking for your agents to be able to operate cross-platform seamlessly, this one’s for you. This integration will enable you to streamline the way you develop and scale your AI agents and weave them into empowering agentic AI workflows. Learn more.

Microsoft 365 Copilot and Copilot Studio

• An exciting new feature came to the Microsoft 365 Copilot in July with the introduction of Copilot Memory. With the aim of personalising the way you use AI to boost productivity, your Copilot now has memory to make it more intelligent, useful, and has the ability to adapt to how you like to work. Your Microsoft 365 Copilot will now remember your work preferences and topics you work on, as two examples. So if you want a more helpful AI assistant at work, Microsoft 365 Copilot is much better positioned to meet your needs.Learn more.

• Speaking of agents, Microsoft have detailed a case study of sorts, sharing how EY have harnessed the power of Microsoft 365 Copilot and agents in business. During the first quarter of the year, Microsoft customers created over a million custom agents across Copilot Studio and Microsoft SharePoint. It’s believed that over 1.3 billion agents will be a reliable part of the workforce across companies by 2028. It’s safe to say the way we work is evolving! Read on to see how your company can do the same. Learn more.

• If you’re constantly looking for information you know someone shared, but you can’t remember who or when, this next update is going to feel like a breath of fresh air for you. Microsoft have launched new powerful knowledge sources in the Microsoft Copilot Studio agent builder embedded in Microsoft 365 Copilot. This gives your agents the ability to refer to both Outlook emails and Teams messages in your search. So, never lose a key reference point in group chats, channels, or even file uploads again! Learn more.

• Microsoft published a new Copilot Studio article detailing everything you need to know about the Auto Triage AI Agent. If you’re interested in automating the bug reporting process across software development, give this one a read. The idea is a much more efficient, streamlined process with the help of generative AI, to take the burden off your customer support teams and provide a better customer experience all around. Learn more.

Azure Analytics and AI

• The Azure Databricks team has presented a compelling case explaining why Databricks in Azure is the best version of Databricks today. Databricks can be used across a range of different cloud service providers (CSPs), but Azure Databricks comes directly from Microsoft and is co-engineered by Microsoft and Databricks. And if you’re interested in a seamless integration across your Microsoft products across the board, Azure Databricks is a no-brainer if you have your mind set on adopting Databricks. Learn more.

• Databricks launched the Databricks AI Governance Framework (DAGF). Covering off 5 pillars and 43 key considerations, the framework aims to support you in managing regulations and compliance while scaling your AI programs. From best practices to reduce risk to ethics, the DAGF acts as a perfect guide for your business. Learn more.

• When implementing anything new, having a use case to reference is always a plus. Microsoft launched exactly that for scaling generative AI in the cloud. When combined with the cloud, generative AI can reach its full potential. In this blog post, discover how to utilise real-time data, embed generative AI into workflows, work with generative search, and more. If you’re looking to gain greater insight, improve customer experience, or even optimise or streamline across the board in your business, this will help. Learn more.

• It’s been announced that Native vector support in Azure SQL Database and Azure SQL Managed Instance will be generally available this summer. No specific dates have been given, but everything is underway, and some regions have received the feature (one of which could be yours). This means that Azure SQL continues to expand well beyond data, enabling you to work with semantic search, harness RAG pipelines, or bring in multi-agent workflows directly inside your database. Learn more.

Open-Source Analytics and AI

• Meta AI shared a case study documenting how the leading freelancer platform Upwork utilises Llama. Whether you’re a job marketplace platform or not, understanding how to leverage Llama to power your own AI developments could be a game-changer. If you’re looking to streamline processes for your team or customers and boost overall experience, have a read of this one. Learn more.

• Following the release of DuckLake back in May, DuckDB just introduced DuckLake version 0.2 with various updates. In the DuckDB ducklake Extension specifically, named and unnamed secrets can be used to manage DuckLake credentials. There’s also a new DuckLake list files function, which could come in handy when integrating DuckLake into other systems. Read the full update for a debrief on how DuckLake just got better. Learn more.

• Preset gave us a lot of food for thought on the topic of open source. Claiming that proprietary software is becoming irrelevant (with an undeniable argument) as open source takes over, well, everything! With the benefits that come with using open source, such as collaboration, transparency, endless use cases, and a plethora of contributors, to name a few, they make a fair point. Pour a coffee, a click through, and have a read. Learn more.

• Interested in GitHub’s hosted server and want to try it out? Then this piece is for you. Forget about the limitations of a local Model Context Protocol (MCP) server and move to GitHub’s managed MCP to focus purely on deploying code. Doing so makes rotating access tokens and pulling updates way easier. If you’re ready to get started or just interested to find out what it looks like, head on over to the tutorial. Learn more.

Stay in the Know

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