Amid a steady drumbeat of “AI jobs apocalypse” headlines — some amplified by AI company leaders themselves — researchers at the Stanford Institute for Economic Policy Research (SIEPR) have pulled together the actual data. Their policy brief, “What is really happening to jobs?”, reaches a calmer conclusion: AI’s impact on total employment is likely small right now, though the picture is genuinely mixed for new graduates.
🔍 THE BOTTOM LINE
The data so far shows a softening labour market, not an AI-driven one. Unemployment has risen about as much in occupations least exposed to AI as in those most exposed. But under that quiet aggregate, two real stories are unfolding: entry-level hiring is difficult for graduates, and firms that use AI well are growing headcount and wages faster than firms that don’t.
The Finding That Cuts Through the Noise
The SIEPR brief’s central chart, built from US Current Population Survey data, tracks unemployment by AI-exposure quintile. Since 2022, the unemployment rate for the most AI-exposed workers rose 0.77 percentage points. For the least-exposed workers, it rose slightly more: 0.85 points. Whatever is driving the softer labour market, AI exposure does not appear to be the dominant factor in aggregate joblessness.
That does not mean AI is effect-free. The brief acknowledges that a tough job market for recent graduates may be partly attributable to AI — consistent with the Stanford Digital Economy Lab’s separate “Canaries in the Coal Mine” research, which found employment among workers aged 22–25 in highly AI-exposed occupations running about 19 per cent below where it would otherwise be expected, a gap the researchers themselves describe as suggestive rather than proven causation.
The Positive Side of the Ledger
Other findings in the synthesis lean encouraging:
- Productivity. AI’s impact on worker productivity is described as mixed but generally positive — one of the clearer positive signals in the research to date.
- Adoption is accelerating unevenly. Firms are adopting AI faster, but unevenly across the economy, which means gains are concentrated rather than universal.
- Retraining over firing. New York Fed survey work, cited in the broader literature the brief reviews, finds firms overwhelmingly plan to retrain workers for AI rather than replace them, and some firms are increasing hiring for AI-proficient staff at the same time as others reduce it.
The researchers are careful about the limits of the evidence. “Early evidence is hardly the last word,” the brief notes, pointing out that adoption is still spreading, that measures of AI exposure are imperfect, and that a fast-moving technology could produce very different outcomes in five years than it has in three.
A NZ Reading
New Zealand doesn’t feature in the US-focused data, but the pattern travels: a small open economy with a service-heavy workforce is exactly the kind of market where uneven adoption matters more than a dramatic headline. For NZ job-seekers, the practical takeaway from the research synthesis is less about avoiding “exposed” occupations and more about the graduate market being difficult — experience, tacit skills and customer-facing abilities are what employers are currently paying premiums for, as our earlier coverage of the New York Fed’s retraining findings and PwC’s wage-premium data also showed.
FAQ
Is AI actually causing mass job losses? Major research syntheses to date, including the Stanford SIEPR brief, find little evidence of AI-driven job losses in aggregate, with unemployment rising similarly in AI-exposed and non-exposed occupations.
Which workers are most affected by AI so far? Entry-level workers in AI-exposed occupations show the clearest strain, with research attributing a widening employment gap for recent graduates — driven more by reduced hiring than layoffs.
What should job-seekers focus on? The research points to AI-complementary skills: judgement, customer-facing work, and experience-based knowledge, which employers currently reward with faster wage growth.
— CJ Murden, editor of Singularity.Kiwi. Former digital technologies teacher, author of AI-focused books. Writing with a New Zealand focus.
