AI is going to make human labor more valuable because human labor and productive capital are complementary. Give a worker control over vastly more capital and you increase the value of that worker’s judgment, decisions, and effort. Make that capital widely available and competing businesses will bid aggressively for the people who can deploy it. My prediction is labor shortages and skyrocketing real wages, not a population waiting for government checks. Particular jobs will disappear, but eliminating tasks is not the same thing as eliminating demand for workers. The traditional individual contributor will increasingly disappear because everyone gets promoted to managing agents. The unemployment forecast notices that a company can shrink from twenty people to two. It misses how many more companies become viable when two people can accomplish what previously required twenty. An LLM connected to tools can research, write software, examine results, revise its approach, and delegate. Autonomous robots extend those capabilities into physical production. Humanoids will allow machine intelligence to operate inside buildings and use equipment designed for human bodies. The human’s role becomes choosing objectives, supplying context, allocating resources, setting standards, handling exceptions, and accepting responsibility. That is management. As agents improve, even mediocre managers of agents will outperform today’s best unaided individual contributors across a widening range of activities. The worst managers will produce worthless slop, but bad managers of people already do that with much larger payrolls. Cheap agents make controlled experiments cheaper; reckless permissions can make mistakes catastrophically expensive. The best managers will know the difference. They will construct large hierarchies of agents with clear responsibilities, independent verification, and sensible escalation rules. Their advantage will not be their ability to generate a million pages of output. It will be their ability to turn an enormous amount of inexpensive machine activity into something worth buying. The production function makes this clearer: Output = F(Capital, Labor). It is not capital plus labor. The complete function is known only to God; economists approximate it. For this purpose, capital includes land, buildings, factories, machines, vehicles, computers, software, infrastructure, and accumulated intangible assets such as organizational knowledge and reputation. Capital is not synonymous with money. Stocks and bonds are financial claims; the certificate does not manufacture anything. Labor is also more complicated than a headcount. Its productivity depends on human knowledge, relationships, contracts, and coordination within and between firms. AI changes how all of this fits together. An agent operates software, a robot operates machinery, and a human directs the organization operating both. The crucial question is not how much the machines produce in total, but how much better that productive system performs because a particular human participates. When one person’s decisions improve the operation of a much larger capital stock, that person’s contribution becomes more valuable. Complementarity is the mechanism connecting machine abundance to higher wages. The other half of the mechanism is expansion. We do not have a fixed inventory of worthwhile businesses, products, treatments, buildings, or scientific experiments. What is worth doing depends on what it costs. LLMs reduce the cost of investigating and planning projects; robots reduce the cost of executing them. A construction robot competes with a construction worker, but it also competes with the decision not to build the house. Lower costs mean more houses, factories, clinics, and power plants become economical.