Running an AI-native engineering org – Claude (claude.com)

At Code w/ Claude SF 2026, Fiona Fung, Director of Engineering for Claude Code, detailed the transformative shift to "agentic coding" within engineering teams, significantly altering processes and structures for application development. This approach, which enables rapid code generation with minimal manual input, has redefined the landscape by reducing the time traditionally spent on coding while creating new bottlenecks in areas like verification, code review, and security. In response, the Claude Code team adopted a just-in-time planning method, favoring dynamic discussions and prototyping over rigid design documents, thus streamlining their workflow and enhancing responsiveness to user feedback. The implications of these changes are far-reaching for the AI/ML community. As human roles evolve, engineers now need to prioritize what requires expert judgment over tasks that can be automated, shifting their focus towards leveraging creative problem-solving and systems expertise. This reallocation of responsibilities allows for more collaborative environments, where non-traditional coders can contribute significantly while maintaining necessary human oversight in critical areas. Fung emphasized the importance of continuously reassessing engineering workflows to ensure they remain effective and relevant, underlining a crucial shift from mere throughput metrics to metrics that align with solving real problems in software development.