Overview
- Microsoft said Monday that Critique and Council are now live in the Researcher tool and that Copilot Cowork is available to customers in its Frontier early‑access program with a Microsoft 365 Copilot license.
- Critique splits the job into two steps where OpenAI’s GPT drafts and Anthropic’s Claude reviews for accuracy, completeness, and citation quality, and Microsoft reports a 13.8% gain on the DRACO research benchmark using this setup.
- Council runs models from Anthropic and OpenAI in parallel and then uses a judge model to summarize where the reports agree, where they differ, and what each model uniquely adds.
- Copilot Cowork handles long‑running, multi‑step tasks across apps like Outlook, Excel, Teams, and PowerPoint, draws on enterprise files, and runs inside tenant boundaries that customers manage.
- Microsoft frames the shift as a multi‑model strategy that reduces reliance on a single vendor and turns its orchestration and security layer into the core value for businesses competing with tools from Google, OpenAI, Anthropic, and Perplexity.