Copilot Cowork expanded to more Frontier customers

Microsoft on Monday added a multi-model Critique feature to its Copilot Researcher agent and expanded access to the Copilot Cowork autonomous agent for more customers in its Frontier program.

The Critique workflow runs a draft generated by OpenAI’s GPT and routes that draft to Anthropic’s Claude for a check of accuracy and quality before the user sees the result. Microsoft announced plans to make the workflow bi-directional so GPT can review drafts produced by Claude in the future.

Microsoft introduced Model Council, a tool that displays responses from different AI models side-by-side so users can compare answers directly. The company says the multi-model approach is intended to speed workflows, reduce instances of hallucination where AI produces false information, and deliver more reliable outputs.

Separately, Microsoft widened access to Copilot Cowork, an agent designed to run multi-step or continuous tasks with greater autonomy. The feature entered testing earlier this month and is modeled on Anthropic’s Claude Cowork.

Frontier is Microsoft’s early-access program that lets organizations trial experimental AI features. The company is using the program to allow customers to test Copilot Cowork before a broader release.

Nicole Herskowitz, corporate vice president of Microsoft 365 and Copilot, described the setup as ‘a way for customers to get the benefits of the models working together.’ Herskowitz noted that having multiple models from different vendors inside Copilot can provide higher quality and more trustworthy outputs.

Microsoft described the updates as part of efforts to enhance Copilot for workplace settings. The changes come as Microsoft competes with other providers that offer large language models and agent-style automation.