AI agents expose legacy testing gaps and require new coordination rituals

Engineering organizations are experiencing bottlenecks in testing and code review as AI agents increase code generation speed. In some teams, code piles up untested because organizations lack automated testing infrastructure, relying instead on manual QA or retrofitting tests after development. This issue is not new to AI but is being exacerbated by higher throughput. For example, one team reported that releasing new features broke core functionality because they had no automated testing suite. The discussion highlights that AI is exposing pre-existing process gaps and technical debt rather than creating new problems. Teams that previously had strong engineering practices are finding that AI accelerates the volume of code that needs review and testing, making existing inefficiencies more painful. As engineering teams adopt AI agents, traditional software development rituals like planning, reviewing, and demoing work remain necessary but are evolving. Some argue that frameworks like Scrum are obsolete, but the underlying need for coordination persists. With multiple agents building software simultaneously, teams are using planning documents, loops, and graphs to communicate requirements and coordinate agent work, preventing conflicts. These agentic workflows mirror human team rituals, ensuring that agents know when they are done and do not step on each other's toes. The number of agents involved is increasing, which heightens the need for clear work agreements and time horizons to maintain visibility into progress and outcomes.