Microsoft Copilot Cowork AI agent launches globally to handle entire workplace projects

Microsoft Copilot Cowork Expands to Global Availability

Microsoft has launched Microsoft Copilot Cowork to all Microsoft 365 Copilot customers worldwide, marking a shift in how AI agents tackle workplace productivity 1

1

. After spending three months in the Microsoft Frontier preview program, the service has already attracted more than half of the Fortune 500, alongside organizations including Accenture, Zurich Insurance, Capital Group, Koch, Ooredoo Qatar, and Avanade 2

2

3

3

. Charles Lamanna, EVP of Copilot, Agents and Platform at Microsoft, described Cowork as "the fastest growing feature in the history of our Frontier program" with among the highest user satisfaction of any Copilot or agent experience the company has shipped 1

1

.

Enterprise-Grade Task Automation That Executes Complete Workflows

Unlike traditional AI assistants that merely generate drafts or suggestions, this AI agent is designed to autonomously execute multi-step workflows from start to finish 2

2

. The service automates complex workplace tasks including batch-editing spreadsheets, comparing thousands of files across product versions, generating dependency charts, and evaluating at-risk opportunities for sales teams 1

1

. Microsoft attributes this capability to cloud-based processing, enterprise security controls, and Work IQ—a context engine that pulls information from the tools and systems businesses already use within the Microsoft 365 ecosystem 2

2

.

Multiple AI Models and Third-Party Integrations Enhance Flexibility

Microsoft emphasizes flexibility by allowing customers to tap into different AI models depending on task requirements. At launch, the service runs on Anthropic's Opus 4.8 and Sonnet 4.6 models, while Microsoft Frontier customers can access GPT 5.5 1

1

3

3

. A new in-house model called Cowork 1 is expected to arrive in the coming weeks, designed to deliver enterprise-grade performance at substantially lower cost for everyday tasks 1

1

. The platform now supports nine partner plugins immediately available, including Monday.com, Miro, and Moodys, with eight more coming soon from workplace heavyweights such as Adobe, Atlassian, Box, and Canva 1

1

.

Consumption-Based Pricing Model Introduces Cost Controls

Microsoft has introduced a different pricing strategy for AI-driven project completion. While Copilot Cowork requires a Microsoft 365 Copilot subscription, usage is billed separately through a consumption-based model at $0.01 per Copilot Credit 3

3

. Organizations are charged based on resources required for each task, including model usage, context retrieval, tool calls, and runtime 2

2

. Microsoft has identified three common categories: light, medium, and heavy tasks, ranging from simple requests to large-scale jobs requiring deeper analysis 2

2

. The company claims internal testing showed the service to be roughly 30% to 40% cheaper per prompt than competing enterprise AI offerings using Microsoft 365 connectors 2

2

. Administrators can now manage access, create budgets, set spending limits, configure alerts, and monitor usage across organizations 3

3

.

Enhanced Security and Browser Capabilities Expand Use Cases

The general availability release introduces expanded compliance features including audit logs, Data Security Posture Management, eDiscovery, Insider Risk Management, Data Lifecycle Management, and Communication Compliance 3

3

. Data Loss Prevention capabilities will be added in a future update 3

3

. Microsoft Frontier users can now leverage Edge browser integration, allowing Copilot Cowork to get online and expanding the range of tasks it can complete on behalf of users 1

1

. Organizations that participated in the preview program between March 30 and June 16 will not be charged for usage until July 1, though billing begins immediately for new customers 3

3

. Microsoft is betting that the next phase of AI-powered workplace productivity tools isn't about generating content faster—it's about handing entire projects to an AI agent and letting it bring back finished work 2

2

.