Generative AI can act as a powerful force multiplier for developers—but it has limitations, especially when deployed at scale with minimal human oversight. How can we design tools meant for large-scale use that anticipate these challenges and prevent autonomous agents from going off-track, without falling into a repetitive cycle that constantly requires human intervention?

In this talk, we’ll walk through a real-world example of a tool built at Moderne based upon OpenRewrite for mass-scale code transformations. We’ll use it to highlight practical strategies for designing scalable AI workflows: reusing previously built functionality, validating inputs before execution, and processing the results efficiently. We’ll also explore how agents can be enhanced by providing well-structured, context-aware tools—whether via MCP, LangChain, or other orchestration frameworks—to help them operate reliably and safely at scale.