Rethinking Work with Cloudflare OS

Cloudflare has built and unveiled Cloudflare OS, an internal platform for safely integrating AI agents into work. The company had cautiously adopted AI through 2025, focusing on informational chatbots and code assistance, but the emergence of improved models and powerful tools late last year enabled AI to perform actual tasks.

A turning point occurred when a sales team member requested API keys and deployment pipeline admin permissions for approximately 12 systems to deploy a 'SuperApp' created with AI into production. Cloudflare needed to support employees in leveraging new tools while protecting internal systems and enterprise and customer data.

Cloudflare OS was constructed by connecting existing components of the Developer and Zero Trust platforms, such as Workers and Access, and adding customized services tailored to new ways of working. This enables both technical and non-technical roles to productively utilize AI and agents.

During the adoption process, Cloudflare established the following principles:

• AI is used to spend more time with customers and solve customer problems.
• Employees unfamiliar with developer tools must also be able to use AI's 'superpowers'.
• AI is a tool, not a team member, and humans are responsible for the quality and verification of outputs.
• Users and teams who deploy agents are responsible for the results of those agents.

The company first defined the 'jobs to be done' and bottlenecks that employees sought to resolve, then selected appropriate tools. Subsequently, it built features for engineers and non-engineers, cultivated champions to spread use cases throughout the organization, and drove the AI-based transformation of work.