Microsoft Copilot: New AI agents for research and data analysis presented

The US software company Microsoft has added two AI agents to its Copilot chatbot to help employees with research and analysis. The AI agent called Researcher is designed to help employees gather information, while the one called Analyst can process large amounts of data. Both AI agents are given access to internal data such as documents, emails or chat messages in Microsoft 365. They also search the internet independently.

AI agent combines internal data with web research

Microsoft uses the OpenAI deep research model for Researcher, but adds its own research functions to its AI agent. According to the company, Researcher can create its own analyses from internal data and online research, for example for market launch strategies or quarterly reports. Researcher can also access data from third-party providers via interfaces. This means that external sources, including Salesforce, ServiceNow or Confluence, can be integrated into the chatbot. The AI agent can also use data from other agents such as Sales Chat.

The basis for Analyst, on the other hand, is the o3-mini reasoning model from OpenAI. The AI agent uses a chain of thought to iteratively solve problems and gain insights based on raw data within minutes. In addition, Analyst can handle Python code and , according to Microsoft, users can view the executed code in real time and check the work of the AI agent. Analyst can thus compile and visualize raw data from different tables, for example for sales and demand forecasts.

At the same time, Microsoft announced the addition of reasoning capabilities and agent flows to its Copilot Studio platform. Companies can create their own AI agents there in a low-code environment. Researcher and Analyst can be used from April 2025 as part of the new Frontier program, which gives users advance access to functions that are still under development. Most recently, the Copilot chatbot was given a new interface and additional functions.

(sfe)