【AI前沿】Cost Reduction of Nearly 40%! Microsoft Launches Copilot Cowork Intelligent Agent, Directly Competing with Claude

AI NEWSLatest AI NewsArticleCost Reduction of Nearly 40%! Microsoft Launches Copilot Cowork Intelligent Agent, Directly Competing with ClaudePublished in Latest AI NewsTime :Jun 17, 2026Read :3minuteMicrosoft officially announced yesterday that its enterprise-oriented intelligent agent AI product, Copilot Cowork, has been launched globally. According to Microsoft’s official blog post, the product demonstrated extremely high cost-effectiveness during internal testing, with an average cost per prompt 30% to 40% lower than its competitor Claude Cowork.As an AI assistant designed specifically for enterprises, Copilot Cowork can perform complex and long-term multi-tool tasks end-to-end. Users only need to define the final goal in the system, and the intelligent agent will automatically call various tools to complete the entire process and return the results.Preferred by Half of the Giants, Focused on Cloud-Based Long-Term ExecutionBefore the official launch, Microsoft conducted a three-month preview test. During this period, more than half of the Fortune 500 companies, including Accenture, Zurich Insurance, and Capital Group, have already adopted and used the product.Compared to other similar products in the market, Copilot Cowork has unique technological advantages empowered by the Microsoft ecosystem. It not only supports cloud-hosted execution, ensuring tasks continue running even after users close their laptops, but also natively integrates a powerful context engine.Flexible Billing Models, Comprehensive Protection for Enterprise SecurityIn terms of security and compliance, the product fully inherits the trust boundary of Microsoft 365, meeting enterprise-level security standards. At the same time, it offers multiple large model options, which can match the optimal operational efficiency based on the actual needs of the task.Regarding billing, Copilot Cowork uses a more flexible pay-as-you-go model, settled through dedicated credits. The overall operating cost is composed of four dimensions: model usage, context retrieval, tool calls, and runtime, providing cost estimates for light, medium, and heavy tasks for different types of enterprise users.