For those of us who’ve had the privilege of using the "premium" Microsoft 365 Copilot license for a while, having a powerful AI assistant at our side has become second nature -summarizing documents, drafting emails, and getting quick answers on demand. Copilot has evolved into a trusted generalist, our go-to for just about everything. Now, with the introduction of two specialized AI agents, Researcher and Analyst, Microsoft is raising the bar and redefining what it means to have “expertise on demand.” In this blog post, we’ll explore what these new agents bring to the table - and how they can become invaluable teammates in your daily work.
Some advanced users have criticized Microsoft 365 Copilot for being “too basic,” citing limitations like the inability to select specific data models or perform deep analytics. But those concerns are quickly being put to rest. At Storyals, the feedback we’ve received from these agents from executives during webinars, 1:1 sessions, and group coaching sessions has been overwhelmingly positive. With the introduction of Researcher and Analyst, Microsoft 365 Copilot licensees now have access to AI agents that don’t just assist - they think, reason, and act like domain experts.
The Researcher and Analyst agents were officially announced by Microsoft in March 2025 and rolled out to users as part of their new “Frontier” program, designed to give early adopters access to the very latest innovations by Microsoft. The agents were made publicly available to all Microsoft 365 Copilot licensees June 2nd, 2025. Let’s have a deeper look at what these agents are.
The Researcher agent is your go-to for deep-dive research. It mimics how a human researcher would approach a complex question - breaking it down, gathering data from internal and external sources, and synthesizing it into a structured, well-cited report. Researcher uses a structured reasoning loop—Reason → Retrieve → Review—to iteratively gather and evaluate information. It pulls from your organization’s internal data (SharePoint, Teams, Outlook, etc.) and the web, using Microsoft Graph connectors and Bing search. It then compiles findings into a report with full citations. It’s powered by a research-optimized version of OpenAI’s O3 model (a successor to GPT-4), fine-tuned for multi-step reasoning. Microsoft’s orchestration layer, the Copilot Control System, ensures secure access to enterprise data.
The Analyst agent is your virtual data scientist. It takes raw data - Excel files, CSVs (Comma-Separated Values), databases - and turns it into actionable insights using statistical models, visualizations, and Python code (Python is a powerful programming language often used for data analysis). Analyst is built on OpenAI’s O3-mini model, post-trained by Microsoft for analytical tasks. It uses a chain-of-thought approach to break down data problems, generate Python code, execute it in a secure sandbox, and refine its analysis based on results. It can clean messy data, join datasets, detect outliers, forecast trends, and generate charts - all while exposing the code it used, so you can verify or reuse it.
To prepare for this blog post, we used the Researcher agent to handle the background work - gathering insights on technical architecture, capabilities, limitations, and customer feedback. You can view the full prompt and output from Researcher here, “Microsoft 365 Copilot “Researcher” and “Analyst” Agents: In-Depth Analysis”. Research that would typically take us days was completed in under 30 minutes (albeit with some initial hiccups due to time-outs).
The use cases for these agents are vast. Whether you're in HR, IT, or business operations, Researcher and Analyst can dramatically cut down the time spent on research and analysis - freeing your team to focus on strategy and innovation.
Imagine you're preparing for a board meeting. Instead of manually compiling updates, you ask Researcher to generate a quarterly business review. It pulls from project files, meeting notes, and industry news to deliver a polished report.
Or say you're analyzing employee engagement trends. Analyst takes your survey data, identifies patterns, and highlights areas for improvement - no formulas or coding required. According to tests conducted by Microsoft, product managers and account managers that tested the Researcher agent estimated that it saved them on the order of 6–8 hours per week each – essentially reclaiming almost a full workday that was previously spent on gathering and synthesizing information.
Unlike the general-purpose Copilot assistant that delivers instant responses, the Researcher and Analyst agents take their time, typically 10 to 15 minutes, sometimes longer. When we used Researcher to support this blog post, it initially timed out. But the final result? A comprehensive, 30-page report packed with insights.
Watching these agents work is fascinating. You can see them break down the task, plan their approach, gather information, analyze results, and synthesize everything into a professional-grade output. It truly feels like handing the task to a highly educated domain expert.
There are a few practical considerations to keep in mind:
Patience Pays Off: These agents “think” more deeply, so while they may take longer, the quality of the output is well worth the wait.
Curious to see these agents in action? Watch our on-demand webinar, “Exploring Copilot Spring Wave 2,” for a firsthand look at how Researcher and Analyst work in real life. If you're ready to explore how these agents can elevate your organization’s productivity, we’re here to support you. Through our Copilot Kick-start Program, we offer hands-on, instructor-led coaching sessions that introduce new ways of working with Microsoft 365 Copilot - alongside your new AI teammates, Researcher and Analyst.