We talked to 100+ dev teams about how they use AI. Here are some things we learned.
We spent four months talking to more than 100 tech leads, architects, and senior engineers about how their teams actually use AI. Review time is a real constraint. But the one they kept circling back to was harder to name: control. The dominant story about AI and code review is a throughput story: agents write faster than any human can read, the PR queue backs up, and the fix is to review faster…
For four months, researchers interviewed over 100 tech leads, architects, and senior engineers to understand how their teams utilize AI. A recurring theme emerged: control. While efficiency was often discussed, it was apparent that humans were still essential for maintaining quality and system integrity. The review process, where humans made critical decisions, remained a bottleneck.
Control was the deeper concern, not just time constraints. Teams that structured themselves around agents backed off; code files grew too large for agents to manage effectively. Writing specifications, while LLMs could generate boilerplate code, required human expertise for business-specific details. Reviewing and validating code remained crucial, with about 80% of teams still involving humans in this process.
AI can generate code quickly, but without human oversight, the consequences can be disastrous. Early-stage startups, where speed is crucial, faced unmanageable codebases later on, leading to delays and increased costs. The mental model of the codebase allowed senior developers to review PRs efficiently, but this model was lost with AI agents.
Review bots flagged individual lines but failed to address the broader system implications. The review bottleneck was not solely about time; it was about control - the ability to understand and confidently make decisions about code changes. The shift in constraint from writing code to understanding it marked a significant change in software processes.
Despite new tools, the core principles of software development remain unchanged, with understanding and maintaining code being the true limiting factor.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
