Early forecasts of AI's labor-market impact ranged from apocalyptic to dismissive. Enough empirical evidence has now accumulated to see which trajectories were closer to right.
What the data shows
Employment in categories heavily exposed to AI — mid-tier customer support, some writing and translation work, entry-level graphic design, some paralegal work — has softened measurably. Total employment in these categories is down, though not collapsed. New job postings in the affected categories have declined faster than existing employment, suggesting the primary channel is hiring freezes rather than layoffs.
In other categories the picture is opposite. Skilled trades, healthcare, and non-routine manual work continue to face labor shortages, and AI has produced no material impact on employment. The dispersion in outcomes is striking.
The wage effect
Where AI has increased productivity in cognitive work, wages have generally risen for workers who use the tools effectively and stagnated for those who do not. The result is widening within-occupation dispersion — top performers pull further ahead, median performers hold flat, weaker performers see relative decline. Between-occupation dispersion is a different story and depends on the specific field.
- Entry-level cognitive work has been the most affected on both hiring and wages.
- Skilled trades and non-routine physical work remain largely unaffected.
- Within-occupation wage dispersion is widening in fields where AI is materially productive.
The policy response
Government responses have varied. Several jurisdictions have expanded worker retraining funding. A handful have introduced transitional support for workers displaced from specific occupations. Universal frameworks — basic income, robot taxes — remain politically fringe even where their advocates are prominent. The dominant policy pattern is targeted intervention in specific sectors.
“The debate about whether AI would affect the labor market is finished. The debate about which workers deserve what kind of support has barely started.”
What to watch
The most important indicator over the next two years is what happens to entry-level pipelines. Fields that historically trained senior workers through junior tasks now automate many of those junior tasks. Whether replacement career paths emerge in time will shape the long-run distributional picture more than any single policy.
Key Topics
Extended Knowledge
- Empirical labor-market research on AI has matured substantially in 2024–26 as datasets have accumulated.
- Wage effects in AI-exposed occupations have been highly bimodal, favoring workers who use tools effectively.
- The career-ladder question — whether junior tasks will be replaced faster than replacement paths emerge — is the most consequential open question.
Frequently Asked
No. It is causing measurable employment declines in specific categories while leaving many others unaffected.
Entry-level cognitive work has been the most affected so far. Non-routine manual and skilled trade work remains largely unaffected.
Targeted retraining and transitional support have shown promise. Universal frameworks remain politically difficult.
