Microsoft is the latest company to equip its AI chatbot, Copilot, with deep research tools, allowing it to use reasoning and analysis to provide complex answers and complete tasks. It is also able to connect with third-party data connectors and derive insights across a company’s tools.
This adds to the growing wave of chatbots with advanced thinking capabilities. Most recently, Google introduced reasoning AI models to Gemini, and xAI’s Grok 3 model includes bots trained with “reinforcement learning” that helps it use a chain-of-thought reasoning process.
These changes come at a time where there are increasing concerns about the reliability of information provided by AI chatbots. AI blunders happen often, and where chatbots are now integrated into businesses, misinformation could pose a large risk.
What Do New Copilot Tools Consist Of?
The updated Microsoft 365 Copilot includes a ‘Researcher’ and ‘Analyst’ agent to provide a deeper reasoning ability and thorough research skills.
The ‘Researcher’ agent merges OpenAI’s deep research model with Microsoft’s own “advanced orchestration” and “deep search capabilities”. It also integrates with the other programs in a user’s toolbox, such as Salesforce, creating in-depth reports and strategies with a broad range of knowledge.
This just in! View
the top business tech deals for 2025 👨💻
The ‘Analyst’ model is built on OpenAI’s o3-mini reasoning model, and focuses on advanced data analysis to tackle complex queries and provide refined answers. The agent knows programming language Python, which the user is able to read and inspect for themselves as the agent is coding.
Both tools will start to be rolled out in April for customers on the new Frontier program, which will from then on be the place where experimental Copilot features first launch.
Copilot Keeping Up With The Competition
AI companies are in a constant race of improving and defining their chatbots within the crowd, particularly as AI continues to establish more and more integrated within the workplace.
Copilot is the latest chatbot to receive deep research and analysis abilities. Not so long ago, Google unveiled a new ‘reasoning’ update to its chatbot Gemini, allowing the bot to pause and ‘think’ before giving answers. ChatGPT also has a Deep Research feature, in the form of an agent OpenAI claims reports “at the level of a research analyst”.
The ability to ‘think deeply’ and refine this thinking as new information pops up, could be crucial in the development of AI. It is most certainly an important step in creating AI bots that are able to problem solve and combine knowledge as humans do.
Risks Remain When Using AI for Research
However, as exciting as deeper reasoning agents are, there are still concerns about how accurate the information AI bots provide, particularly if they are being used in professional settings.
The fear is that AI bots can “hallucinate” or make information up. There have been cases of lawyers being sactioned for submitting fake case citations given by ChatGPT, or Microsoft chatbot MyCity telling business owners it is legal to take a cut of their workers’ tips.
Not only do chatbots need the ability to think deeply about problems, but also take the sources they are given and decide whether it is reliable or accurate enough to mention in its research. At the moment at least, this seems like a level of intuition reserved only for humans.
