This talk argues that organisations are asking the wrong question about AI. Instead of "what should we do with AI?", which anchors thinking to current capabilities and tends to optimise for producing existing outputs faster, it proposes "what would you do with a thousand more employees?", a framing that shifts attention to outcomes and strategy. Drawing on productivity research (AI makes knowledge workers roughly 1.5–2x more productive) and token costs falling around tenfold per year, it estimates that the effective labour of a thousand workers will become affordable for the price of a single subscription sometime in the early-to-mid 2030s. But cost, the talk argues, is not the binding constraint: the real limits are human attention, AI autonomy, and above all human intention, because a thousandfold increase in cognitive capacity is worthless without a thousand things worth doing. Updating earlier work on cognitive forcing functions for an agentic world in which AI now plans and executes long-running tasks, the talk argues that critical thinking must move up from individual sentences or lines of code to the level of plans, and presents experimental evidence that prompting people to surface the assumptions behind an AI-generated plan reduces overreliance more effectively than asking them to imagine failure modes, even though users prefer the latter. It also notes that articulating goals upstream improves every downstream AI output, and highlights studies showing that human teams with AI clearly outperform individuals with AI: AI raises the floor, but only humans raise the ceiling. The talk closes by contrasting studies of organisations where workers hide their AI use with those where visibility and peer influence drive adoption, concluding that in a world of abundant output, the scarce resources are intention, culture, goal quality and teams, and that thriving organisations will invest in their people as seriously as in their technology.