Preface

Microsoft has integrated 'deep research' AI capabilities into its Microsoft 365 Copilot, enhancing its AI chatbot platform. This innovation follows the trend of recent deep research agents within chatbots, such as OpenAI's ChatGPT, Google's Gemini, and xAI's Grok. These tools use reasoning AI models that improve problem-solving and self-verification — essential skills for comprehensive topic exploration. Within Microsoft's lineup, the tools dubbed 'Researcher' and 'Analyst' make strides in business and data analytics.

Lazy bag

The newly introduced AI tools called Researcher and Analyst enhance Copilot by providing unique deep research capabilities using reasoning models and access to extensive data sources.

Main Body

Microsoft is enhancing its Microsoft 365 Copilot with an exciting addition: AI-powered deep research tools known as Researcher and Analyst. These tools are part of a broader movement in the tech industry, where large technology firms develop deep research agents that leverage reasoning AI models. Such models exhibit the ability to address complex problems and self-verify, which is particularly valuable in executing extensive research tasks.

Researcher is a tool that amalgamates OpenAI’s sophisticated deep research model with advanced orchestration and deep search functionalities. This tool can execute a variety of analyses, from crafting go-to-market strategies to compiling quarterly client reports. On the other hand, Analyst is based on OpenAI’s o3-mini reasoning model, highly optimized for performing advanced data analysis. Analyst goes through iterative problem-solving steps, refining each step and using Python programming to address intricate data queries. Moreover, it can reveal its analytical process for further scrutiny.

What distinguishes Microsoft's tools is their dual access to both internal work data and the extensive resources available on the internet. Researcher, for example, can interact with third-party data platforms through connectors from various AI agents and applications, such as Confluence, ServiceNow, and Salesforce.

Despite these advances, challenges persist. Ensuring AI models like o3-mini and deep research avoid fabricating information is crucial. These models are not flawless and may occasionally misattribute sources, make inaccurate conclusions, or reference unreliable public websites.

Plans to introduce these tools are underway through Microsoft's new Frontier program, where Microsoft 365 Copilot users will gain early access to experimental features. Beginning in April, Frontier participants will have access to both Researcher and Analyst, marking the start of significant advancements in AI-powered research solutions.

Key Insights Table