Where humans stay hired
The accountability principle
Legal systems run on the idea that responsibility has to attach to a person or entity that can be held liable. Executives are accountable. Doctors are accountable. Judges are accountable. Algorithms are harder to hold responsible — you cannot sue a neural network. Work that carries meaningful legal liability therefore resists automation even when the underlying capability is matched.
Embodied experience
Customers pay premiums for experiences that require a human to be actually present. Concerts, sports events, weddings, therapy, hospice care, teaching. Recordings and simulations exist and are cheaper. People still pay for the original. This is not irrational: neuroscience suggests humans are wired to register and value the presence of other humans in ways recordings do not reproduce.
O-ring and relational sectors
David Autor's work identifies O-ring sectors — work where every component has to be done well or the whole output fails — as persistently human. These overlap heavily with relational work: nursing, teaching, management, skilled services. The growth of these sectors, even as others have shrunk, is why aggregate employment has held up despite large sectoral shifts.
What this does and doesn't tell us
The human exception is real but limited. It is not a reason for complacency. It suggests some work will persist; it does not say how much, or that the persisting work will be well paid. In a post-labour world, the human exception is one of several survival channels — and most working lives will involve navigating the boundary between it and automation.
Name three situations in your own life where, given the choice between a human and a machine at equal quality, you would pay extra for the human. Why?