Microsoft adds ‘deep reasoning’ Copilot AI for research and data analysis

The Verge - Mar 26th, 2025
Open on The Verge

Microsoft has announced new AI developments for its Microsoft 365 Copilot, introducing two pioneering 'deep reasoning' agents named Researcher and Analyst. The Researcher agent utilizes OpenAI’s advanced research model to conduct complex, multi-step research, integrating third-party data from services like Salesforce and ServiceNow. Meanwhile, the Analyst agent employs OpenAI’s o3-mini reasoning model to transform raw data into organized spreadsheets and run Python code, functioning at a level comparable to skilled data scientists. These tools are set to launch in April as part of an early access program for Microsoft 365 Copilot users, alongside new features in Copilot Studio that include advanced autonomous agent capabilities.

This development marks Microsoft's latest foray into AI, following similar announcements from Google and OpenAI, highlighting an intense competitive landscape in AI-driven productivity tools. The introduction of these agents suggests a significant step toward automating complex tasks across business ecosystems, potentially reshaping workflows and enhancing efficiency. By integrating AI-driven insights into everyday business tools, Microsoft aims to offer a seamless low-code experience that promises to revolutionize how tasks are automated and managed, though its practical effectiveness remains to be seen upon rollout.

Story submitted by Fairstory

RATING

7.0
Fair Story
Consider it well-founded

The article provides a clear and timely overview of Microsoft's new AI agents, effectively communicating their potential capabilities and applications. While the factual accuracy is strong, with well-supported claims about the technology, the article could benefit from more balanced reporting by including diverse perspectives and expert opinions. The reliance on Microsoft's announcements limits source diversity, and the lack of detailed transparency about methodologies and potential conflicts of interest slightly undermines trust. Despite these limitations, the article is well-written and accessible, making it a valuable resource for those interested in recent developments in AI technology. Enhancing engagement through exploration of potential challenges and controversies could further improve the article's impact and public interest.

RATING DETAILS

8
Accuracy

The story accurately presents the introduction of two new 'deep reasoning' AI agents, Researcher and Analyst, for Microsoft 365 Copilot. It correctly attributes these developments to Microsoft's collaboration with OpenAI, highlighting the use of OpenAI's models. The claims about the capabilities of these agents, such as Researcher's ability to pull off complex, multi-step research and Analyst's proficiency in turning raw data into spreadsheets, are consistent with available information. However, the story could benefit from more detailed verification of specific technical claims, such as the precise nature of the 'chain-of-thought reasoning' and the practical implications of integrating third-party data via connectors.

7
Balance

The article predominantly presents Microsoft's perspective on their new AI agents, focusing on the capabilities and potential benefits without exploring potential limitations or challenges. While it briefly mentions the need to see these tools in action to assess their effectiveness, it lacks viewpoints from industry experts or competitors that could provide a more balanced view of the technology's impact and potential drawbacks. Including such perspectives would enhance the article's balance by offering a more comprehensive understanding of the AI agents' place in the broader tech landscape.

8
Clarity

The article is generally clear and well-structured, effectively communicating the main points about Microsoft's new AI agents. The language is accessible, and the progression from introducing the agents to detailing their capabilities is logical. However, some technical jargon, such as 'chain-of-thought reasoning' and 'rule-based workflows,' might benefit from further explanation to ensure comprehension by a broader audience. Overall, the article maintains a neutral tone and presents information in a straightforward manner.

6
Source quality

The article relies on Microsoft's announcements and claims, which are credible given the company's authority in the tech industry. However, the lack of external sources or expert opinions limits the depth of analysis and verification. Incorporating insights from independent tech analysts or referencing third-party evaluations of the AI models could improve the reliability and depth of the reporting, providing a more nuanced view of the technology's capabilities and potential market impact.

6
Transparency

The article provides a clear overview of Microsoft's new AI tools but lacks detailed transparency regarding the methodologies and data sources used to support the claims. While it mentions the use of OpenAI's models and third-party data integration, it does not delve into the specifics of how these integrations are achieved or the potential limitations. Greater transparency about the technical underpinnings and potential conflicts of interest, such as Microsoft's business motivations, would enhance the reader's understanding and trust in the information presented.

Sources

  1. https://siliconangle.com/2025/03/25/microsoft-365-copilot-gets-ai-reasoning-skills-advanced-research-analysis/
  2. https://www.thebridgechronicle.com/tech/microsoft-deep-reasoning-copilot-ai
  3. https://mezha.media/en/news/microsoft-adds-deep-reasoning-ai-agents-to-copilot-300700/
  4. https://www.zdnet.com/article/microsoft-365-copilot-gets-two-deep-reasoning-ai-agents/
  5. https://techcrunch.com/2025/03/25/microsoft-adds-ai-powered-deep-research-tools-to-copilot/