Microsoft has announced two new features for its Copilot Researcher tool—Critique and Council. These features enable OpenAI's GPT and Anthropic's Claude to work simultaneously on a single task to enhance accuracy.
Introducing Critique, a new multi-model deep research system in M365 Copilot.
— Satya Nadella (@satyanadella) March 30, 2026
You can use multiple models together to generate optimal responses and reports. pic.twitter.com/m4RlQmCKzs
Critique is a new multimodal deep analysis system designed specifically to tackle complex research processes. The tool utilizes a combination of neural networks from leading labs.
“One model leads the generation phase, planning the task, iterating through information searches, and creating an initial draft. The other is responsible for reviewing and refining, acting as an expert reviewer,” Microsoft stated.
Moreover, the LLMs can swap their roles.
The company emphasized that all current AI research tools operate similarly, with no oversight on their outputs. Critique aims to address issues of hallucinations, citation errors, and the provision of false or inaccurate statements.
In the DRACO test, which covers 100 complex research tasks across 10 domains, Critique scored 57.4, while Claude Opus 4.6 scored 42.7.
The second feature, Council, takes a different approach. It runs GPT and Claude in parallel and displays their full reports side by side. A third model then reads both responses and summarizes them, explaining where the AIs agree and where their arguments diverge.
In Critique, the LLMs collaborate, while in Council, they compete.
Critique is the default mode in Researcher, while Council must be selected from the menu. Both features are available to users registered in the Microsoft Frontier program.
As a reminder, in March, the company introduced the AI agent Cowork for handling complex tasks in Microsoft 365. It can interact with ecosystem applications like Outlook, Teams, Excel, PowerPoint, and SharePoint.