Following closely on the heels of AI announcements from Google and OpenAI, Microsoft has unveiled its own advancements in artificial intelligence, detailing two innovative “deep reasoning” agents for its Microsoft 365 Copilot. These agents, named Researcher and Analyst, are described as first-of-their-kind AI tools designed to enhance productivity and data analysis within the Microsoft ecosystem. The company is also introducing new capabilities for custom AI agents, further expanding the functionality of Copilot.
Microsoft Introduces Researcher and Analyst AI Agents for Copilot
Researcher: Advanced AI-Powered Research Assistant
Researcher leverages OpenAI’s advanced research AI model to conduct “complex, multi-step research.” This agent is designed to access and synthesize information from various sources, including third-party data via connectors to platforms like Salesforce and ServiceNow. This connectivity empowers business users to glean valuable insights from across their existing suite of business tools and applications, streamlining the research process and improving data-driven decision-making.
Analyst: AI for Data Interpretation and Reporting
Analyst, powered by OpenAI’s o3-mini reasoning model, is engineered to transform raw data into actionable intelligence. Microsoft states that through chain-of-thought reasoning, this AI agent can convert unprocessed data into organized spreadsheets, execute Python code with real-time monitoring, and perform data analysis at the level of a skilled data scientist. Analyst is capable of compiling comprehensive reports, effectively automating complex data processing and analytical tasks.
Early Access and Autonomous Agent Capabilities
Both Researcher and Analyst are scheduled to become available in April to Microsoft 365 Copilot license holders through an early access program. Alongside these specialized agents, Microsoft is also rolling out new autonomous agent capabilities within Copilot Studio, starting immediately.
Automating Tasks with Copilot Agent Flows
Microsoft asserts that the new agent flows within Copilot possess sufficient power to “automate virtually any conceivable task.” These flows utilize rule-based workflows that incorporate AI actions, offering a wide range of automation possibilities. As illustrated in a LinkedIn announcement, an example application includes an agent flow designed to route feedback emails efficiently to the appropriate teams. The practical effectiveness of these automated workflows, their superiority over conventional solutions, and the user-friendliness of the “low code” experience will be key aspects to evaluate as these AI agent capabilities are deployed and tested in real-world scenarios.