After a decade leading data and AI across very different industries — eBay, Schiphol Group, ING, and now as Chief Data Officer at Eneco — Dr. Tor Bøe-Lillegraven offers a clear-eyed take on AI agents: real rewards, real risks, and the realities of managing swarms of agents at scale. His message is direct: today's AI is evolutionary, not a revolution — yet. The organisations creating real value aren't chasing a big-bang transformation. They adopt AI steadily and safely, especially where mistakes are costly and can't be undone. At the heart of the talk is a tension every AI system has to balance, drawn from Tor's own research on ambidextrous AI agents: explore versus exploit. Every AI agent constantly weighs making the most of what it already knows, and testing new approaches. The most effective AI systems, like the most successful organisations, manage to do both at once — and getting that balance right between exploration and exploitation.