External signal·Financial Times·Apr 9, 2026·John Burn-Murdoch, Madhumita Murgia·9 min read
The AI job loss story is all about bundles
“the same technological capability shrinks one job while expanding another”
Summary
An FT "AI Shift" newsletter arguing that AI job displacement is best understood through the "bundles" of tasks that make up a job. A new Federal Reserve paper by Leland Crane and Paul Soto is the first time official labour-force survey data has corroborated earlier private-payroll findings (from Stanford's Erik Brynjolfsson and colleagues): both estimate roughly half a million fewer coders are working than pre-LLM trends implied — a marked slowdown in growth, not an absolute decline. The piece maps this onto LSE economist Luis Garicano's framework: "weak bundle" jobs (junior and contractor coders writing to spec) have a large chunk broken off by AI and are diminished or obviated, while "strong bundle" senior and domain-embedded roles stay intact, with AI acting as an assistant and multiplier. Madhumita Murgia notes the pattern looks profession-agnostic — engineers, journalists, musicians, finance and law — and that AI agents are climbing the task ladder from mundane to complex (citing the METR benchmark and OpenAI's Mark Chen on lengthening task horizons). Co-author John Burn-Murdoch remains relatively optimistic, seeing little displacement outside coding so far.
Predictions for the future of work
Predicts AI displacement concentrates first on "weak bundle," early-career roles while enhancing senior, domain-embedded ones — the same capability shrinking one job and expanding another. Murgia warns that as agents handle longer, higher-level tasks, even experienced workers may be pushed toward managing agents rather than creating; Burn-Murdoch expects displacement to stay largely confined to coding for a while, with AI more multiplier than replacement.
Originally published by Financial Times · Apr 9, 2026
Read the original at Financial Times