Microsoft Copilot, the artificial intelligence platform that had already positioned itself as a key piece of the productivity ecosystem, has taken an evolutionary leap in 2025.
Gone is the stage in which AI was limited to writing emails or summarizing documents: today it is consolidated as a suite of digital specialists with the arrival of two advanced reasoning agents: Researcher and Analyst.
The vision of a hybrid workforce
Introduced in March 2025 under the Frontier Innovation Program and released into general availability in June, these agents do not simply represent new features, but an entirely different category of AI.
They are autonomous collaborators, capable of tackling complex, multi-step projects, at the level of a human researcher or data scientist.
With them, Microsoft reinforces its vision of a hybrid workforce, where AI stops being a passive co-pilot and becomes an expert that works alongside the user.
Powered by OpenAI’s next-generation models (o3-mini and o3) and supported by Microsoft infrastructure, these agents are designed to tackle large-scale challenges.
Researcher: the intelligent researcher
Researcher is the perfect ally for in-depth research projects. Its strength is in hybrid search: it securely combines internal information (emails, Teams chats, OneDrive documents, meeting minutes) with external sources from the web and data from business systems such as Salesforce, ServiceNow or Confluence. The result: a panoramic and verifiable vision.
What distinguishes this agent is its ability to reason in a chain of thought. Instead of providing superficial answers, it breaks down each query into logical sub-questions, detects ambiguities, and asks the user for clarification.
It works like a true researcher, with the goal of understanding the context before delivering a result. Its reports include not only conclusions and visualizations, but also direct quotes from sources, guaranteeing transparency and reliability.
How to use it:
- Open Microsoft 365 Copilot and select the Researcher agent.
- Enter your query, for example: “Analyze the impact of remote work on team productivity and suggest best practices.”
- Answer any clarifying questions the agent asks.
- Receive a detailed report, with graphs, key insights and references.
- Refine or export the result to Word or PowerPoint (available from July 2025).
Advantages and limitations
Analyst: the virtual data scientist
If Researcher is the researcher, Analyst is the data scientist. This agent converts large volumes of information into practical knowledge, without the user having to master programming or specialized software.
Based on the o3-mini model, it works as a Python expert that accurately analyzes, cleans and visualizes data.
Its great strength is the ability to execute Python code in real time. This means that it not only processes complex calculations, but exposes the code used, providing an additional layer of transparency and trust.
Simply attach a data file (Excel, CSV, etc.) from OneDrive or your device and ask a question in natural language. The agent will return complete charts, tables, forecasts, and statistical analysis, even if the data is in error or out of order.
How to use it
- Select the Analyst agent in Microsoft 365 Copilot.
- Attach the data file to be analyzed.
- Pose your query, for example: “Compare sales by region and project demand for the next quarter.”
- Get a report with results, visuals and executed Python code.
Advantages and limitations
##Researcher vs. Analyst: which one to choose?
The decision depends on the nature of the task:
- Researcher: when the objective is to investigate and synthesize information. Useful for market reports, thematic reviews or strategic presentations.
- Analyst: when the important thing is to analyze numerical data and visualize it. Ideal for detecting trends, segmenting customers, developing projections or validating hypotheses with hard data.
In many cases, both agents complement each other: first, Researcher contextualizes a problem with qualitative information; Analyst then validates it with quantitative data.
Availability and future prospects
These agents are available to Microsoft 365 Copilot licensed users, subject to IT configuration. Its integration with Office applications and Copilot Studio opens the door to personalized workflows, where standard agents can be combined with agents custom designed for each organization.
The first reactions have been mostly positive. Users highlight the time savings and the ability to tackle projects that were previously unfeasible without a complete team of analysts.
However, the need for human supervision is also noted, both due to the limits of consultations and the inherent risks of “hallucinations.”
In any case, Microsoft is betting on a future in which AI stops being a basic assistant and becomes an autonomous specialist. With these new super-powers, we take a step towards human-machine collaboration, freeing up time and energy for what matters most: creativity and strategy.
This post is also available in: