How to rethink software releases in the age of AI

The rapid increase in AI-driven code production has outpaced existing software release processes, creating significant bottlenecks in development lifecycles. While 57% of organizations rely on human-in-the-loop reviews for every line of AI-generated code, this manual approach is becoming unsustainable. Furthermore, only 49% of companies have implemented specific guardrails for AI-generated code, leaving them vulnerable to unintended consequences like those seen in recent high-profile outages. To navigate this shift, engineering teams must adopt progressive delivery principles, such as feature flags and automated testing, to decouple releases from launches and limit the blast radius of new code. As AI agents take on more development tasks, the industry must transition from manual oversight to automated, policy-driven guardrails to ensure quality and security without sacrificing velocity.