Economists at the Federal Reserve Bank of Richmond have published new research suggesting the AI-driven squeeze on hiring is landing on specific groups of workers — including some the headlines rarely mention. According to an economic brief published in August 2026 by Katarína Borovičková and Claudia Macaluso, the recent decline in job-finding rates — how quickly unemployed people find work — is concentrated in particular worker types rather than spread evenly across the labour force.
🔍 THE BOTTOM LINE
The Richmond Fed findings complicate two competing narratives at once. Job seekers in AI-exposed occupations face measurably harder conditions than they did three years ago. But the aggregate picture is softer than the “AI jobs apocalypse” framing suggests, and one of the groups most affected isn’t inexperienced newcomers — it’s stable, well-attached workers in exposed occupations. Both things can be true, and the distinction matters for where you direct a career.
Who Is Finding It Harder to Get Hired?
The researchers classified unemployed workers along three dimensions: why they were searching, how attached they were to the labour market, and how exposed their occupation is to AI, using task overlap with capabilities described in AI patents. Highly exposed occupations in this framework include computer programmers, financial analysts and engineers. Less-exposed work includes construction, food service and personal care.
Three findings stand out from the brief:
- New entrants are not the main story. Job-finding rates for labour-force newcomers have declined, but modestly. “Entrants do not appear to be the margin where the action is,” the authors write.
- Stably employed workers — the “primary” type — have seen the largest declines. From a November 2022 peak to a September 2025 trough, the job-finding rate fell 13 percentage points for this group, versus 2 points for frequently churning workers. In every recession on record, including the Great Recession, this group moved the most.
- Since 2023, workers in highly AI-exposed occupations have seen the largest declines in job-finding — a divergence from history, when outflow rates across exposure quartiles moved tightly together with the business cycle.
The timing is notable. Historically, AI exposure said little about how fast someone found work. Since 2023 — the period when AI use became genuinely widespread — the quartiles have separated.
Context: The Wider Data Is Less Dramatic
A July 2026 policy brief from Stanford’s Institute for Economic Policy Research synthesising the research found AI’s impact on aggregate employment is “likely small right now.” Unemployment among the most AI-exposed quintile of workers has risen since 2022, but by slightly less than among the least-exposed quintile (0.77 versus 0.85 percentage points). Stanford also found that among firms that adopted enterprise AI, employment grew 10 percent in the two years after adoption.
Meanwhile, Challenger, Gray & Christmas data published in early August showed US layoff announcements hitting a two-year low in July, with announced hiring plans up 25 percent year on year — “while AI is shifting the labor market, it is not dismantling it,” as the firm’s Andy Challenger put it. And Indeed’s August 2026 US snapshot recorded AI-related job postings climbing to 6.3 percent of all postings, well past their 2022 peak of 3.3 percent.
Taken together, the picture is: fewer mass layoffs than feared, more hiring than expected, but a genuine slow-down in the rate at which exposed workers — particularly established ones — land new jobs. It’s a market that rewards staying put, or carefully choosing a pivot.
What It Means for Careers
For workers in exposed occupations, the practical read is blunt. The Richmond Fed data suggests the difficulty isn’t only a first-job problem — experienced, stable workers in those occupations are also finding re-entry slower. That strengthens the case for building adjacent, less-exposed skills (the judgment-heavy, client-facing, physical-world work the PwC Barometer flagged as growing) before a forced search begins, rather than after.
For everyone else, the aggregate numbers counsel calm. Most industries are cutting less than last year, and demand is concentrated in physical-world sectors — aerospace, energy, manufacturing — “work that happens on a floor rather than a screen,” in Challenger’s words.
The New Zealand Angle
New Zealand’s labour market is not the US market, but the mechanism the Richmond Fed describes — slower re-entry into exposed occupations once a job is lost — travels. Stats NZ recorded an unemployment rate of 5.6 percent in the June 2026 quarter, an 11-year high, meaning the baseline for any search is already harder than it’s been in a decade. In a market where unemployment is elevated for cyclical reasons, an AI-exposure disadvantage compounds: fewer vacancies overall, and the tightest competition for roles in occupations where AI is simultaneously reshaping task content.
The AI Forum’s survey work found AI has created new career opportunities in 55 percent of adopting NZ organisations, even as nearly half of adopters reported reduced hiring needs. For Kiwi workers in programming, analysis and engineering roles, the Richmond Fed research is less a warning siren than a planning document: the skills that get you into your next role may need to be different from the ones that got you into your last.
FAQ
What is a job-finding rate? The rate at which unemployed people transition into employment. A falling job-finding rate means searches take longer, even when layoffs aren’t rising.
Which occupations count as AI-exposed in the Richmond Fed study? Based on Michael Webb’s patent-based framework: computer programming, financial analysis, engineering and similar desk-based analytical work. Construction, food service and personal care count as less exposed.
Are entry-level workers the hardest hit? Not in this study. The Richmond Fed researchers found new labour-force entrants saw only modest declines. The biggest declines hit stably-employed workers in AI-exposed occupations.
Is this true in New Zealand? The research is US-based and its authors caution it’s early evidence. NZ-specific effects are harder to measure, but Stats NZ’s June 2026 unemployment rate of 5.6 percent means Kiwi job seekers face a tight market regardless of AI exposure.
— CJ Murden, editor of Singularity.Kiwi. Former digital technologies teacher, author of AI-focused books. Writing with a New Zealand focus.
