Microsoft Launches Copilot Cowork Globally, Bringing Agentic AI to the Fortune 500

Furthermore, Microsoft revealed that its proprietary "Cowork 1" model is scheduled for release in the coming weeks. This secure, fine-tuned model has been post-trained specifically for agentic tasks. Microsoft designed Cowork 1 to deliver optimal cost and quality while focusing on enterprise-grade requirements, such as the removal of model bias. The company expects this model to be a primary choice for cost-sensitive workloads, as it is engineered to handle everyday tasks at a substantially lower price point than current frontier models.

The Economics of Agentic AI: Pricing and Credits

With the general availability of Cowork, Microsoft is introducing a usage-based pricing model designed to reflect the significant compute resources required for long-running, autonomous tasks. To access Cowork, organizations must first have a Microsoft 365 Copilot User Subscription License (USL). This base subscription provides the standard suite of AI tools, including Copilot Chat, integration with Office applications like Excel and PowerPoint, and the Work IQ context engine.

Beyond the base subscription, Copilot Cowork is billed based on consumption, denominated in "Copilot Credits." The cost of each task is calculated using four distinct inputs: the specific AI model used, the depth of context retrieval required from organizational data, the number of tool calls executed, and the total runtime of the task.

To assist organizations in budgeting for this variable cost, Microsoft has identified three common task patterns based on observations from the Frontier program:

Light tasks are characterized by a small number of knowledge sources and limited reasoning requirements, typically producing a single output. Medium tasks draw on multiple data sources, require more structured reasoning, and often generate two or more outputs. Heavy tasks are the most resource-intensive, aggregating data broadly across the organization, applying deep reasoning, and producing a high volume of outputs.

By mapping these task types against four distinct user personas—corporate knowledge workers, management and senior leaders, customer-facing workers, and technical staff—businesses can create more accurate cost estimates. Microsoft has released a specialized estimator tool to help IT leaders forecast their monthly credit consumption based on these variables.

Competitive Performance and Cost Management

In a move to position Copilot Cowork as the most cost-effective solution in the market, Microsoft released internal benchmarking data comparing its system to Claude Cowork utilizing a Microsoft 365 connector. According to the company’s analysis, Copilot Cowork was, on average, 30% to 40% cheaper per prompt than the competitor’s offering. This analysis was based on 125 test runs across various task complexities using the Opus 4.8 model.