Microsoft Corp. is enhancing the capabilities of its popular artificial intelligence-powered Copilot tool with the launch late today of its first “deep reasoning” agents, which can solve complex problems in the way a highly skilled professional might do.

The new Researcher and Analyst agents in Microsoft 365 Copilot represent the latest evolution of AI agents, with their reasoning skills allowing them to handle complicated tasks that require deep analysis, methodical thinking and a nuanced understanding of the end goal.

In a blog post, the company explained that the new agents are powered by advanced reasoning models such as OpenAI’s o1. By connecting them securely to enterprise data such as emails, meetings, files and chats, they can take on the role of a trusted professional and manage ambiguous business processes just as a human employee would.

The new Researcher agent is a great showcase of this, automating and accelerating complex, multistep research to deliver detailed reports on almost any kind of business topic. It combines the reasoning skills of OpenAI’s o1 model with Microsoft 365 Copilot’s advanced orchestration and search capabilities to perform all manner of research-based tasks.

As an example, Microsoft said it can be used to create a detailed go-to-market strategy that’s based on the context of an organization’s internal data and also the broader market context, which can be sourced online. It can also perform tasks such as identifying whitespace opportunities based on new trends and internal performance, create comprehensive quarterly reports for clients detailing the services provided during that period.

Because Researcher is into research, it’s important to give it access to as much data as possible. To that end, it can also use third-party connectors to bring in data from external business platforms such as Salesforce, ServiceNow, Confluence, Trello, Asana and many others directly into Microsoft 365 Copilot.

Researcher’s capabilities are pretty impressive. In the following example, it’s asked to outline a strategy that explores how a fictional shoemaker that embeds electronics into its footwear can expand into the outdoor adventure market. The result is a slick, nine-page report that summarizes the company’s existing offerings and their strengths, and how these can be leveraged to differentiate itself in the new market.

As for the Analyst agent, it’s designed to be more like a skilled data scientist, and its mission is to transform raw data into actionable insights. It leverages OpenAI’s o3-mini reasoning model, which has been optimized for advanced data analysis, and uses “chain-of-thought” reasoning to work through problems iteratively, taking as many steps as necessary to refine its thought processes to provide answers that mirror human analytical thinking.

Microsoft said Analyst can be used to transform raw data scattered across multiple spreadsheets into a comprehensive demand forecast for a new product line, or else create a visualization of customer purchasing patterns, detailed revenue projections and so on.

The Researcher and Analyst agents will become available to Microsoft 365 Copilot subscribers in April via a new Frontier program that provides early access to the latest Copilot innovations.

Smarter AI agents everywhere

While customers will have to wait a while to get their hands on Research and Analyst, they’ll be able to apply the latest advances in AI reasoning to their own Copilot agents sooner.

Later this month, Microsoft plans to debut deep reasoning and agent flows in Microsoft Copilot Studio, which is a platform that allows customers to design, create, deploy and manage their own AI agents. They’ll soon be able to integrate the most advanced reasoning models into AI agents customized for numerous different kinds of business processes.

When they’re powered by reasoning models, AI agents will spend more time carefully thinking about the task in hand. They’ll be able to parse large volumes of data and recognize intricate patterns within them, generating more thoughtful responses and outputs, Microsoft says. It paves the way for AI agents that can be applied to entirely new business processes, such as forecasting demand across global markets or optimizing supply chains.

The company said deep reasoning is being made available in both conversational and autonomous agents, and it’s being integrated in the background, so there’s no need for customers to adjust the underlying model that powers each Copilot agent.

In addition, Copilot Studio is getting a new Agent Flows capability that allows agents to follow predefined sequences of actions to ensure more consistent results in structured tasks. By following rule-based workflows based on deterministic paths, AI agents can perform predictable and repetitive duties like document processing and financial approvals much more efficiently and with fewer mistakes.

Finally, Microsoft is making its first autonomous agents available in Copilot Studio. They’re designed to respond proactively to more than 50 different pre-built triggers so they can plan tasks and manage ambiguous scenarios by themselves, without any supervision, the company said. They’re tailored for numerous business tasks, including device procurement, supplier discovery, fraud prevention, due diligence, engagement management and many more.

Images: Microsoft