As systems take on more autonomy, control doesn’t disappear, it just becomes harder to define. Decisions are no longer made in a single step. They emerge across prompts, tools, and workflows that evolve over time. What looks controlled in isolation can behave differently once systems are live, making ownership, accountability, and intervention less straightforward. This panel explores where control actually sits in modern AI systems, how teams are designing for it, where it breaks down, and what it takes to keep systems predictable without slowing them down. Key takeaways: → Where control shifts as systems become more autonomous → How teams balance flexibility with guardrails in practice → What it takes to maintain oversight without introducing friction