AI rarely introduces new problems in DevOps. Instead, it amplifies the ones that were already there: unclear ownership, fragile automation, hidden costs, and teams stretched too thin. This talk looks at how AI acts as a stress test for DevOps practices. I’ll show how issues around reliability, accountability, and communication become impossible to ignore once AI-driven tools and workflows are introduced. Rather than focusing on models or tools, this session is about how teams can use AI as a signal to fix deeper problems — and avoid turning “AI adoption” into another layer of complexity on top of an already fragile system.