As teams move beyond prompts and prototypes, the real challenges begin to surface. Systems that behave differently under pressure, workflows that don’t quite hold up, and complexity that wasn’t obvious at the start. The work shifts from generating outputs to managing systems. This keynote sets the direction for the day, what’s actually changed as AI moves from models to systems, what teams underestimated, and where things tend to break when usage becomes real. Key takeaways: → Why prompt-driven approaches start to fall apart at scale → Where system design, not model quality, becomes the bottleneck → What separates something that works once from something you can rely on