JetBrains Report: AI-Generated Changes Create Review Bottlenecks in CI/CD Pipelines

A new report from JetBrains, 'The State of CI/CD in 2026,' reveals a growing tension in software delivery: while AI is rapidly being adopted across CI/CD pipelines, the volume of AI-generated changes is creating significant bottlenecks in the code review process. The report indicates that 48.8% of surveyed organizations utilize AI for at least one CI/CD-related task, with common applications including code review, test writing/fixing, and diagnosing build failures. However, this increased automation is not without its drawbacks. The primary concern highlighted is that the sheer quantity of changes produced by AI tools is slowing down human review cycles and making it difficult to assess the overall reliability of the codebase. This directly impacts the velocity that CI/CD pipelines are designed to achieve. Furthermore, teams are grappling with unpredictable AI costs, challenges in tracking AI-made changes, and debugging issues introduced by AI. This trend fits within the broader context of AI's accelerating integration into software development and operations. As AI agents become more sophisticated and capable of generating substantial portions of code and infrastructure configurations, the traditional human-centric gatekeeping mechanisms are struggling to keep pace. The industry has been moving towards greater automation in IaC and CI/CD for years, but the advent of generative AI introduces a new magnitude of change that existing processes were not designed to handle. This also intersects with the growing need for robust governance around AI agent access to production infrastructure, as highlighted by other industry discussions. In practice, this means DevOps and platform engineering teams need to urgently re-evaluate their code review strategies. Simple human-in-the-loop approvals for every AI-generated change will become unsustainable. Practitioners should explore advanced tooling for AI-assisted review, focusing on anomaly detection, automated policy enforcement, and more sophisticated change validation mechanisms. The emphasis will shift from reviewing every line to validating the intent and impact of AI-driven changes, potentially leveraging AI itself to help triage and prioritize reviews. Organizations must also establish clear guidelines and guardrails for AI agent autonomy, especially when these agents interact with production environments, to mitigate risks associated with rapid, high-volume automated changes.