JetBrains Survey Reveals AI-Driven Code Accelerates Output but Bottlenecks Shift to Review and Delivery

A new report from JetBrains, "The State of CI/CD in 2026," reveals a critical paradox emerging in software development: AI-generated code is dramatically increasing output, yet it's simultaneously shifting and exacerbating bottlenecks in the code review and delivery phases. The survey, which gathered insights from nearly 600 CI/CD tool users, indicates that while 83.4% of DevOps practitioners leverage CI/CD pipelines, the integration of AI into coding workflows is creating unforeseen challenges. This development is significant because it highlights a growing tension between the promise of AI-driven acceleration and the practical realities of maintaining software quality and security. For practitioners, this means that simply adopting AI coding assistants isn't a silver bullet for faster delivery. Instead, it necessitates a deeper look into how AI-generated changes are validated, tested, and integrated into existing CI/CD pipelines. The report notes that AI's impact is most felt in code reviews, testing, failure analysis, pipeline configurations, quality assurance, security, deployments, and incident response. This trend fits within the broader industry movement towards "AI-augmented DevOps" and "agentic AI." Companies like GitLab are making their Model Context Protocol (MCP) server generally available, allowing AI-compatible agents to access an expanded catalog of tools and agents, aiming to integrate AI more deeply into the software development lifecycle. Similarly, GitHub has introduced programmatic access to Copilot Code Review via APIs, enabling AI-driven reviews to become a configurable part of the pipeline rather than a manual step. The JetBrains report corroborates this, indicating that over a quarter of organizations interact with CI/CD tools through agentic interfaces, suggesting a move towards more autonomous, AI-driven workflows. In practice, this means organizations must evolve their CI/CD strategies to address the unique characteristics of AI-generated code. This includes implementing more sophisticated automated testing, potentially leveraging AI for test generation and analysis, and refining code review processes to efficiently handle larger and more frequent AI-driven changes. Practitioners should also explore tools and platforms that offer integrated AI governance and observability within their CI/CD pipelines. The report also highlights that 72.8% of teams still self-host their primary CI/CD systems, often for greater control over security and data, which becomes even more critical when dealing with AI-generated intellectual property. The trade-offs between AI's productivity gains and the increased complexity in verification and delivery will require careful consideration and investment in new tooling and processes to avoid becoming a bottleneck themselves.