The latest update to Microsoft 365 Copilot brings AI-powered search, so-called reasoning agents, and a new Agent Store. Some users already have access to certain features, while others may have to wait through May.
One of the additions is AI-powered enterprise search, which once challenged enterprise search vendors because of the complexity of applying policies across diverse data repositories to make sure only appropriate documents were called forth.
“With Copilot Search, we’re bringing the power of search and AI together to give you more relevant, context aware search results,” said Jon Friedman, corporate VP of design and research, in a video explainer. “It connects your work data and dozens of apps so you can quickly find what you need. And Copilot can summarize the files so you know what the content is without ever having to open the document.”
For anyone concerned about letting a large language model rummage through corporate files, Microsoft points to its Copilot Control System, intended to let IT admins and security folk manage Copilot interactions and access.
That includes an Apps and Agents section in Data Security Posture Management for AI with Purview for just this sort of oversight. There’s also a Copilot Analytics service and a Copilot Studio Agents Report available in Microsoft Viva Insights.
As for the AI itself, Microsoft 365 Copilot has taken on two new passengers: Researcher and Analyst. They’re “reasoning agents,” meaning they’re based on OpenAI’s reasoning models, although Microsoft hasn’t specified exactly which ones. Reasoning here refers to a process of breaking down tasks into steps and showing the execution of those steps.
The Researcher agent is intended to assist with multi-step research tasks, similar to the OpenAI deep research agent the AI giant introduced in February. Microsoft’s demonstration video of the process shows an agent being asked to come up with a marketing plan using the following prompt:
Imagine you are an ambitious but not very clever intern and your employer’s crack marketing team comes down with food poisoning after a recommendation engine’s suggestion to try a trendy found shellfish pop-up. You’d be thrilled to have an agent like Researcher craft a passable marketing plan for which you could take credit.
The Analyst agent promises similar info-wrangling feats in the context of data analysis.
“Analyst is built on OpenAI’s o3-mini reasoning model, and optimized to do advanced data analysis at work, including knowing when and how to use Python,” an unidentified Microsoft employee explains in a video. “So even if you’ve never written a single line of code, you can still get the rich data insights only Python can provide.”
You may not be able to understand the code, but Microsoft’s AI can explain it to you. Progress, apparently.
These new agents are available in the shiny new Agent Store, which also features offerings from various partners including Jira, Monday.com, and Miro, among others. Custom-built agents are in the mix too, giving companies the option to automate specific tasks or provide Copilot with access to specific data.
Model providers are investing billions annually to enhance GenAI models’ size, performance, and reliability. This paradox will persist through 2025 and 2026
Over in the Create experience, Copilot has been bolstered with the addition of the image generation capabilities of OpenAI’s GPT-4o. Would-be design professionals, or even talented visual artists disinterested in applying their skills, can now demand a broader set of banners, display graphics, and marketing copy from Microsoft’s compliant AI.
The Windows giant also introduced Notebooks, a place to dump assorted files that Copilot can then mull on demand.
“Copilot Notebooks organize all of my files on a particular project into one place, and lets me have grounded conversations about that project with Copilot,” explained Friedman.
Some features in Wave 2 (like Researcher and Analyst agents) are available immediately to Frontier Program customers, with broader availability expected through May.
The announcement comes as IT consultancy Gartner predicts an AI paradox in the coming year: Dissatisfaction with generative AI, accompanied by rising spending.
“Expectations for GenAI’s capabilities are declining due to high failure rates in initial proof-of-concept (POC) work and dissatisfaction with current GenAI results,” said Gartner Research VP John-David Lovelock in a statement. “Despite this, foundational model providers are investing billions annually to enhance GenAI models’ size, performance, and reliability. This paradox will persist through 2025 and 2026.”
Gartner projects that worldwide generative AI spending will surpass $644 billion in 2025, up more than 76 percent from 2024. ®
Discussion about this post