Microsoft has announced a significant update to its Microsoft 365 Copilot platform, integrating technology from Anthropic to introduce new agent-based capabilities and expanding model diversity for enterprise clients.

The third wave of Microsoft 365 Copilot introduces Copilot Cowork, a feature designed to execute multi-step, persistent tasks within documents, spreadsheets, presentations, and emails, and to manage these tasks over extended periods. This development forms part of Microsoft’s strategy to increase both the intelligence and operational reliability of AI in enterprise environments.

Copilot Cowork leverages Anthropic’s Claude agent technology, marking the first time this model has been embedded directly into Microsoft 365 Copilot.

According to Microsoft AI at Work chief marketing officer Jared Spataro, Cowork’s design enables the AI to break down complex tasks, coordinate actions across multiple tools and files, and maintain continuity as work progresses.

Spataro, in an official Microsoft blog, wrote: “Your work is not limited by one brand of models. Copilot hosts the best innovation from across the industry and chooses the right model for the job regardless of who built it.”

The Copilot Cowork feature is currently available as a research preview for select customers, with plans to open broader access through Microsoft’s Frontier programme in March. By using Work IQ, the feature can contextualise tasks based on a user’s work history, related files, meetings, and communications.

Rather than issuing isolated commands and generating single outputs, users can now delegate ongoing processes to Copilot Cowork, monitor progress, and intervene when necessary. Spataro emphasises that the system provides observability and transparency, and that all content is managed within existing enterprise security and governance frameworks.

Copilot model expansion and integration in core apps

The expansion of Copilot’s model portfolio allows the platform to select AI models best suited to the task at hand, including both Anthropic’s Claude and the latest OpenAI offerings. This approach is said to address the fragmentation created by single-vendor model lock-in, reducing friction for users and complexity for IT managers.

Spataro wrote: “Copilot automatically applies the right model for the task, all grounded in your enterprise context and protected by Microsoft’s security and governance controls.”

Copilot’s agentic capabilities are now integrated into core Microsoft 365 applications. In Word, Excel, PowerPoint, and Outlook, Copilot can create, edit, and refine content within the environment where users are already working. The AI can, for example, enhance drafts in Word, generate formulas in Excel, produce slides in PowerPoint conforming to organisational standards, and draft or revise emails in Outlook.

All edits are trackable, reversible, and aligned with the context of the user’s recent work, files, and communications. Microsoft is said to have ensured that these features respect existing permissions, sensitivity labels, and data protection policies, with files automatically saved to OneDrive and SharePoint to enable audit, compliance, and retention processes.

The new Copilot features in Excel and Word are now generally available, while support for PowerPoint and Outlook is being rolled out in phases over the coming months.

Microsoft has also enhanced Copilot’s chat-based interface, allowing users to initiate workflows such as document creation, meeting scheduling, and email drafting directly from chat, bypassing the need to switch between applications.

Third-party applications and partner integrations, such as Adobe, Figma, and Monday.com, can now be accessed directly through the chat interface. Agent Builder and Copilot Studio tools provide options for both end-users and IT teams to develop custom AI agents for specific business processes.

Recent updates to Copilot Studio include tools for evaluating agent quality and coordinating agent activities across systems.

Agent 365 launches for unified governance

To centralise the management and governance of AI agents, Microsoft has launched Agent 365. This platform provides IT and security leaders with a unified control plane to observe, secure, and govern all agents operating across the organisation.

Spataro wrote: “Agent 365 is the control plane for agents. In practical terms, it gives IT and security leaders one place to observe, secure, and govern every agent across the organisation, and it provides the confidence to move from agent experimentation to enterprise-scale operations.”

Set to be generally available from 1 May, Agent 365 is expected to be priced at $15 per user per month.

Microsoft is also introducing Microsoft 365 E7: The Frontier Suite, available for purchase from 1 May at $99 per user per month. This package bundles Microsoft 365 Copilot, Agent 365, Microsoft Entra Suite, and Microsoft 365 E5 to provide comprehensive security, observability, and governance for organisations deploying AI at scale.

Last November, Microsoft and Nvidia announced plans to invest up to $10bn and $5bn respectively in Anthropic as part of new strategic partnerships. Anthropic is scaling its Claude AI model on Microsoft Azure, using Nvidia technology, to expand access and provide Azure enterprise customers with more model options.

Anthropic has also committed to purchase $30bn in Azure compute capacity and may contract for up to one gigawatt (1GW) of additional capacity.