Microsoft has unveiled significant enhancements to its Researcher agent within Microsoft 365 Copilot, introducing two new multi-model capabilities, Critique and Council, aimed at improving the accuracy, depth, and reliability of AI-driven research workflows.

The updated Researcher tool is designed to support complex research tasks directly within everyday work environments. With the introduction of Critique, Microsoft is redefining how AI systems approach research by separating content generation from evaluation. The system leverages multiple advanced models from leading AI developers, including OpenAI and Anthropic.

In this architecture, one model leads the research process planning tasks, retrieving information, and producing an initial draft, while a second model acts as an expert reviewer. This reviewer evaluates the content, refines the structure, and ensures the final output meets high standards of accuracy and completeness. According to Microsoft, this dual-model approach outperforms traditional single-model systems, delivering higher-quality and more reliable research outputs.

Complementing Critique is Council, a feature that presents multiple model-generated responses side by side. It also includes a summary “cover letter” that highlights areas of agreement, divergence, and unique insights across models, enabling users to better understand different perspectives on a topic.

The Critique system mirrors academic and professional review processes through a rubric-based evaluation framework. It focuses on key dimensions such as source reliability, report completeness, and strict evidence grounding. By enforcing the use of authoritative sources and requiring precise citations for key claims, the system aims to enhance trust and transparency in AI-generated research.

Critique will serve as the default research experience within Researcher when the “Auto” mode is selected, signalling Microsoft’s push toward more rigorous, multi-layered AI workflows.

With these advancements, Microsoft continues to strengthen its position in enterprise AI, emphasising not just speed and automation, but also accuracy, accountability, and depth in knowledge work.