- Most AI implementations fail by repeating old mistakes, such as monolithic, unmanageable prompts that are costly, fragile, and hard to debug.
- This session makes the case for technical leadership to stay close to the code and understand the real challenges AI introduces to systems.
- Learn how CircleCI’s team used agent decomposition to transform unreliable workflows into composable, maintainable systems.
- You’ll walk away with a practical framework for modular AI agent design, and hard-won lessons from scaling AI beyond proof-of-concept.
Rob Zuber, CTO, CircleCI