RBA analysis: young workers in AI-exposed jobs fare worse in Australia

An analysis by Mike Major for Australia's central bank (the RBA) built a monitoring system that looks where AI effects are likely to show up first: entry-level workers and highly AI-exposed occupations such as programming, clerical work and customer service. The headline finding is that AI has barely displaced any jobs in Australia so far, but young workers entering highly exposed fields like software development are faring worse.
According to the analysis, job postings have fallen across all groups since ChatGPT launched in late 2022, with the steepest drop in the highest-exposure group. In that group, workers aged 15 to 24 and young college graduates have weaker employment rates and higher unemployment than older workers.
Major cautions against early conclusions. The study does not establish direct causation, and it measures only theoretical exposure, not actual use by companies. Post-pandemic shifts in the business cycle or in remote work could also be driving the numbers. According to SEEK data, AI is complementing existing work while also creating new demand for AI skills.
The Decoder sets the RBA work beside other studies. An August 2025 Stanford study based on millions of US payroll records found employment among 22- to 25-year-olds in AI-exposed occupations down 13 percent in relative terms since late 2022, and nearly 20 percent among young software developers. More experienced workers in the same occupations held steady or gained ground.
A March 2026 Anthropic study also found signs of declining hiring among 22- to 25-year-olds in exposed occupations, with job-finding rates down about 14 percent. It found no systematic rise in overall unemployment, however, and said theoretical automation potential and actual AI use are far apart. Anthropic also offers a company-level example: according to an internal review, the share of junior hires in technical roles has fallen from 27 percent to 9 percent since early 2025, and in August 2026 its AI agents were logging six times as many hours as human employees. Anthropic concedes that, as an extremely AI-intensive software company, it is a special case. Even so, researchers warn of a "tragedy of the cognitive commons", in which disappearing junior roles lead to a future shortage of experienced professionals, assuming those professionals are still needed.
A large study of 25,000 workers in Denmark confirms the lack of economy-wide impact. Despite a 64 percent AI adoption rate, researchers found no measurable changes in wages or hours worked, and workers saved an average of just 2.8 percent of their working time.
A January 2026 analysis of 10.6 million LinkedIn profiles and US unemployment data cautions against blaming the downturn on AI alone. The team led by Morgan Frank at the University of Pittsburgh found that the problems in exposed occupations started in early 2022, months before ChatGPT launched. Macroeconomic factors, such as the Fed's shift in interest rate policy or a correction after pandemic-era hiring surges, may play a bigger role than previously assumed.
AI is also creating demand. According to the World Economic Forum's Future of Jobs Report 2025, 77 percent of surveyed companies are prioritizing upskilling over job cuts, and WEF researchers project a net gain of 78 million jobs by 2030. According to LinkedIn, AI-related job postings grew sixfold in one year.
Key facts
- An analysis by Mike Major for the RBA found job postings fell across all groups since ChatGPT launched in late 2022, with the steepest drop in the highest-exposure occupations.
- In that highest-exposure group, workers aged 15 to 24 and young college graduates have weaker employment rates and higher unemployment than older workers.
- The study does not establish causation and measures theoretical exposure, not actual company use; post-pandemic business-cycle or remote-work shifts could also explain the numbers.
- A Stanford study (August 2025) found employment down 13 percent in relative terms for 22- to 25-year-olds in exposed jobs, and nearly 20 percent for young software developers.
- A Danish study of 25,000 workers found no measurable change in wages or hours despite 64 percent AI adoption, and a University of Pittsburgh analysis says problems began in early 2022, before ChatGPT.
Why it matters
A central bank is now tracking AI's labour-market effects with a purpose-built monitoring system, focused on the places where effects should appear first: entry-level workers and highly exposed occupations. The RBA finding fits a pattern in other studies. The Stanford and Anthropic work also point to weaker outcomes for 22- to 25-year-olds in exposed jobs, while broad measures of unemployment and wages have barely moved. The concern raised by researchers is a "tragedy of the cognitive commons": if junior roles disappear, there may be too few experienced professionals later, assuming they are still needed.
Who it affects
Mainly young people starting careers in highly exposed fields. The RBA analysis names programming, clerical work and customer service, and says workers aged 15 to 24 and young college graduates in the highest-exposure group have weaker employment rates and higher unemployment than older workers. The Stanford study found more experienced workers in the same occupations held steady or gained ground. Employers hiring junior staff are affected too, as Anthropic's own fall in junior technical hires from 27 percent to 9 percent shows.
How to use it
The source offers no practical guidance for individuals. For readers following the debate, it works as a map of the evidence: the RBA monitoring approach, the Stanford and Anthropic findings on young workers, the Danish and University of Pittsburgh results that push back, and the WEF figure that 77 percent of surveyed companies prioritize upskilling over job cuts.
How solid is it
Moderate, and the author says so. Major cautions against early conclusions: the study does not establish direct causation and measures only theoretical exposure, not actual use by companies. The RBA findings are qualitative; the source gives no Australian figures from the analysis. Other evidence is mixed. The Danish study of 25,000 workers found no economy-wide effect, and the January 2026 Pittsburgh analysis of 10.6 million LinkedIn profiles found problems in exposed occupations began in early 2022, before ChatGPT launched. The Anthropic study itself says theoretical automation potential and actual AI use are far apart.
Risks and caveats
Alternative explanations are on the table: post-pandemic shifts in the business cycle or in remote work (Major), and the Fed's interest rate shift or a correction after pandemic-era hiring surges (Frank's team). SEEK data suggests AI is complementing existing work while creating demand for AI skills. Anthropic concedes it is a special case as an extremely AI-intensive software company, so its junior-hiring drop may not generalise. The Decoder's account is a secondary summary of these studies, not the studies themselves.
“tragedy of the cognitive commons”
— Researchers, as cited by The Decoder