Microsoft introduced two new artificial intelligence (AI) agents to its 365 Copilot suite of features on Tuesday. These new AI agents are aimed at its enterprise users and can perform complex research and data analysis-related tasks. Dubbed Researcher and Analyst, these are built on OpenAI’s models with fine-tuning done by the Redmond-based tech giant. Alongside, the company also announced new AI features in Copilot Studio that will enhance the agent creation process for users. New capabilities to let users build autonomous agents have also been added.
Microsoft Introduces New AI Agents
After Adobe, Alibaba, Google, and Nvidia, Microsoft has become the latest to jump on the AI agent announcement wagon this week. In a blog post, the company announced two new AI agents for enterprises that will assist them in tasks involving research and analysis. Both of these will start rolling out to its clients in April.
Researcher is based on OpenAI’s deep research model combined with Microsoft 365 Copilot’s advanced orchestration and deep search capabilities. The company says it can tackle “complex, multi-step research at work” and conduct research to develop product development and prepare quarterly client reports. The AI agent can access work data as well as information from the web to carry out research plans. Additionally, it can also be connected to third-party data sources such as SalesForce and ServiceNow to unify information siloed across various channels.
The second AI agent is Analyst. Built on OpenAI’s o3-mini model, it can process raw data to conduct advanced data analysis and provide digestible insights for businesses. The tech giant claims that the agent works like a skilled data scientist and its chain-of-thought (CoT) mimics human analytical thinking. The agent can also run Python code to process the most complicated data-based queries.
Separately, in a LinkedIn post, the company announced two new features — deep reasoning and agent flows — in Copilot Studio. Deep reasoning allows agents to handle analysis-related tasks that require methodical thinking and nuanced understanding, the tech giant said. The output is grounded by only letting the agents connect to the internal enterprise data sources. Microsoft said this capability will let businesses generate reports on forecast demand across global markets or optimise supply chains.
Agent flows in Copilot Studio let businesses add structured, rule-based workflows that incorporate multi-step actions. Microsoft said the capability is designed to let agents handle predictable and repetitive scenarios such as document processing, routine financial approvals or compliance tasks.
Finally, Microsoft is also enabling autonomous agents in Copilot Studio. These agents can proactively respond to specific triggers (such as incoming emails from specific clients) and plan tasks, and manage scenarios without any human supervision. The company has added more than 50 pre-built triggers across processes such as engagement management, device procurement, supplier discovery, fraud prevention, and more.
