Microsoft has unveiled Copilot Cowork, a new iteration of its AI assistant that diverges from its exclusive reliance on OpenAI models. The updated service now incorporates technology from Anthropic, specifically their Claude Sonnet models. This strategic shift allows users of Microsoft 365 Copilot to leverage either OpenAI's or Anthropic's large language models, signifying a move towards a more flexible, multi-model approach for enterprise AI agents.
Copilot Cowork integrates Anthropic's Claude models into the Microsoft 365 Copilot ecosystem, marking a departure from its previous sole reliance on OpenAI's GPT models. This new offering, part of what Microsoft terms "Wave 3" of its Copilot development, aims to embed more "agentic capabilities" into its productivity suite, moving beyond simple assistance to enabling tasks that unfold over time with less direct human intervention. The product also introduces a new licensing tier, Microsoft 365 E7, which bundles advanced security features alongside the enhanced Copilot functionality.
Read More: xPUBench Shows DPUs Use Less Power for 10 Gbps, GPUs Better for 100 Gbps
The Copilot Cowork tool is designed to operate within Microsoft 365's cloud infrastructure, gaining access to a user's broader enterprise data. This differs from Anthropic's standalone Claude Cowork, which lacks this direct integration. Microsoft claims this allows Copilot Cowork to manage complex, multi-step tasks—such as app creation, spreadsheet building, and data organization—with limited human oversight. The company highlighted its "Work IQ" intelligence layer, which connects Copilot to user patterns, relationships, and content across the Microsoft 365 environment, facilitating these background operations.
Internally, Microsoft reports the existence of over 500,000 AI agents within the company, primarily focused on areas like research, coding, sales intelligence, customer support, and HR functions. The company has indicated that customers are not currently pushing for a shift away from per-user pricing models toward consumption-based ones, despite speculation that AI agents might necessitate such a change.
Read More: UK Online Crime Centre starts in April 2025 to stop bank scams and block fraud text messages
This development arrives amid a period of intense competition in the AI agent space, with Microsoft’s move signaling a broader trend of tech companies seeking to integrate more sophisticated AI capabilities into their core business offerings. Copilot Cowork is currently in testing and is slated for wider release to early-access users soon.