Why AI hasn't triggered the predicted white-collar layoffs

TEXT START: Artificial intelligence has diffused through the economy faster than any previous technology, yet the labour market carries almost no trace of it.

THE DISSECTION

The article converts a structural rupture into a manageable productivity shock. Its evidence shows that AI is currently used more as an assistant than an autonomous substitute, that adoption is uneven, and that young entrants are already exposed. Those are real observations. The analytical maneuver is treating them as evidence that the old labour-market equilibrium remains fundamentally intact.

The text is really doing transition management: narrowing the question from whether human economic necessity is being destroyed to how quickly workers can retrain and how fairly gains can be distributed. It preserves the familiar policy vocabulary—skills, apprenticeships, competition, wage insurance—after the underlying employment circuit has begun to fail.

THE CORE FALLACY

It mistakes shallow deployment for shallow capability. The fact that fewer than one interaction in ten currently completes a task end-to-end measures present usage patterns, not the terminal economic function of the technology. Firms adopt automation incrementally, but competitive pressure drives them toward autonomous completion wherever it is cheaper and sufficiently reliable.

The article also assumes that productivity gains expand demand enough to preserve employment. That works only where demand is genuinely elastic and human labour remains necessary. Under the Discontinuity Thesis, AI eventually attacks the cognitive tasks generating the demand itself. When output can be produced with radically fewer workers, cheaper output does not restore the wage-to-consumption circuit; it increases surplus capacity and concentrates ownership.

Its final error is calling the outcome a “managed transition.” Policy can redistribute purchasing power, delay layoffs, or subsidize retraining. It cannot manufacture productive necessity for a majority once P1, P2, and P3 take hold. UBI and wage insurance can preserve consumption. They cannot restore participation.

HIDDEN ASSUMPTIONS

• AI will remain primarily augmentative rather than becoming reliably autonomous.
• New demand will expand faster than AI eliminates labour requirements.
• New occupations will be large enough and accessible enough to absorb displaced workers.
• Human institutions can preserve stable human-only economic domains despite competitive pressure.
• Competition among AI suppliers will disperse gains rather than merely reduce input prices before consolidation.
• Retraining can occur within the time available and lead to durable demand rather than another temporary glut.
• Protecting junior roles can preserve the career ladder even as those roles become the easiest targets for automation.
• “A new equilibrium” will still include mass productive participation rather than a transfer-dependent population.

The article’s own discussion of 22-to-25-year-olds exposes the mechanism it tries to contain: AI is cutting off the entry rung through which workers historically acquired experience and bargaining power. That is not a minor transition cost. It is the removal of the pipeline that produces future Servitors.

SOCIAL FUNCTION

Primary classification: partial truth serving as transition management and ideological anesthetic.

It accurately identifies adoption lag, demand elasticity, bargaining power, supplier concentration, and concentrated damage to new entrants. But it packages those truths inside a reassuring historical analogy: previous technologies eventually created new work, therefore policy can make AI another adjustment episode. The analogy breaks when the technology is capable of competing across cognitive domains simultaneously and institutions cannot coordinate a permanent human advantage.

The article gives policymakers a set of levers that can cushion the corpse while avoiding the ownership question. It treats the distribution of AI capital as a competition-policy problem when the deeper issue is whether non-owners retain any economically necessary function.

THE VERDICT

This is not evidence that the AI employment apocalypse is false. It is evidence that the kill mechanism is still in its shallow deployment phase. Aggregate unemployment is a lagging indicator; the first fracture appears in entry-level hiring, task composition, wage bargaining, and the collapse of skill ladders.

The article sees the early symptoms and misdiagnoses them as a reversible adjustment. Its “managed outcome” is plausible only as managed decline: transfers can keep consumers alive, training can sort a minority into Servitor niches, and competition rules can slow rent extraction. None of that disproves the Discontinuity Thesis. It describes the hospice protocol for post-WWII capitalism.