Artificial intelligence may transform labour demand before its aggregate effects become large. US nonfarm business productivity growth averages 2.5% in 2023 Q1–2026 Q2, compared with 2.2% over 1948 Q1–2022 Q4. The matching hours series is presented as growth over time, without a long-run average benchmark. These descriptive observations do not identify AI's causal contribution. Firm surveys suggest a larger expected productivity contribution over the next three years. Payroll evidence shows a marked divergence for young workers in AI-exposed occupations, operating primarily through reduced hiring rather than greater separations. This paper combines that evidence with research on recession-driven restructuring and insider-outsider mechanisms. It argues that experimentation and workflow redesign can precede changes in recruitment and, under pressure, more abrupt employment adjustment. Illustrative productivity paths translate three published 2030 GDP scenarios into output-per-hour requirements, making explicit the assumptions about hours and diffusion. The policy priority is a system combining unemployment insurance, employment services and lifelong learning, supported by credible fiscal planning. Higher productivity can finance greater consumption and leisure, but broad gains require income and opportunities to reach those whose paid work declines.