As AI moves from conversation to action, the human role shifts from asking questions to overseeing autonomous agents. But keeping people in the process is only valuable if the interaction is designed to make them effective - not just compliant. In this hands-on workshop, teams will prototype agentic AI experiences grounded in real banking scenarios, designing interfaces where people don't just rubber-stamp AI outputs but genuinely sharpen them. Through interactive exercises, participants will explore patterns that surface the right context at the right moment, structure agent actions for real human judgment, and stress-test their designs against the traps that turn meaningful oversight into approval fatigue and illusory control. You'll leave with practical frameworks for human-agent collaboration where both the AI and the person perform better together.