Microsoft Launches Copilot Cowork to Automate Tasks Across Microsoft 365

Microsoft has introduced a new capability in Microsoft 365 Copilot that allows its AI assistant to execute tasks across workplace applications instead of only responding to prompts. The feature, called Copilot Cowork, integrates technology from Anthropic’s Claude models and enables the system to convert user requests into automated workflows that run across applications such as Outlook, Teams, and Excel.

The feature is currently available through early access in Microsoft’s Frontier program, with broader availability expected later. The update reflects Microsoft’s ongoing effort to expand Copilot from a conversational assistant into a system capable of completing tasks within enterprise productivity environments.

“Working closely with Anthropic, we have integrated the technology behind Claude Cowork into Microsoft 365 Copilot,” Microsoft said in a blog post announcing the update.

Moving From AI Assistance to Task Execution

Over the past year, Microsoft has positioned Copilot as an AI assistant embedded across its productivity suite. With Copilot Cowork, the company is extending that role by allowing the system to perform multi-step actions based on a user’s request.

Instead of manually coordinating tasks across several applications, users can describe the outcome they want, and the system attempts to generate a structured plan and execute it automatically.

“Over the last year, we have been pushing Copilot toward taking action,” Microsoft said. “That means completing tasks, running workflows, and doing work on your behalf.”

The system translates a request into a workflow that runs in the background. Users can review the suggested steps before execution, approve actions, pause the process, or provide additional instructions. This structure keeps users involved while allowing automation to handle routine tasks.

Context-Aware Workflows Across Microsoft 365

A key challenge for AI systems that operate across enterprise software is understanding the context of a user’s work. Microsoft said Copilot Cowork addresses this through Work IQ, a system that gathers signals from Microsoft 365 applications.

By analysing emails, meetings, chat messages, files, and calendar data, the system determines which information is relevant to the task requested by the user. The AI assistant then uses that context to assemble workflows across multiple applications. For instance, if a user asks the system to prepare for a customer meeting, Copilot Cowork can gather information from email conversations, retrieve relevant documents, generate summaries, and compile presentation material within Microsoft 365.

Microsoft also outlined additional scenarios. The tool can review a user’s Outlook calendar to identify scheduling conflicts and suggest changes such as rescheduling meetings or adding focus time. It can also assist with research tasks by collecting financial reports, analyst commentary, and news sources, then organising them into summaries or spreadsheets.

In another example, the system can support product launch planning by compiling competitive comparisons, drafting value propositions, preparing presentation material, and outlining milestones with assigned responsibilities. These examples reflect Microsoft’s attempt to position Copilot Cowork not just as a conversational interface but as a task execution layer across productivity tools.

Security, Governance, and Enterprise Controls

Because Copilot Cowork interacts with enterprise data across multiple applications, Microsoft said governance and security controls are built into the system’s design.

The feature operates within existing Microsoft 365 security frameworks, including identity permissions, compliance policies, and auditing mechanisms. The AI assistant can only access information that the user already has permission to view.

Microsoft also said tasks run within a sandboxed cloud environment, allowing workflows to continue in the background while maintaining separation from other processes. This structure is intended to ensure that automated tasks remain compliant with enterprise data policies while maintaining traceability for actions carried out by the AI system.