A new "jobs apocalypse" narrative threatens that AI will displace the nine-to-five, with Dario Amodei, CEO of Anthropic, warning that AI could eliminate up to half of entry-level white-collar jobs within five years, and JPMorgan, Ford and Amazon forecasting leaner payrolls. However, the doomsday case rests on flawed arithmetic: if AI can do 40 per cent of a job’s tasks, then 40 per cent of the work must go, but a job is not a to-do list. AI is a prediction engine that cannot tell you how biased its sources are, cannot show its work in a way you can audit, and can sound just as confident when it invents a citation as when it reports a fact. Because AI cannot flag its own margins of error, an organisation that fires the worker who can figure out whether to trust the output is automating away its quality control. The error in the predictions is treating AI as a product that arrives and subtracts, but because AI is a general-purpose technology, like electricity, the economy will be rebuilt around it. The answer is to become the person who knows when the machine has made an error, when it can be made better and when its output is worth putting your name on.