Australia has done something most governments have not yet done: it has formally mapped, occupation by occupation, which jobs are most at risk from AI, and published the data.
The findings from Jobs and Skills Australia are not dramatic in the short term. No wave of layoffs has arrived. But the trend lines are clear. Between late 2022 and early 2026, employment in the least-exposed jobs grew by 9.5%. Employment in the most-exposed roles grew by only 5.6%. That gap, invisible in any single month's data, adds up to a real structural shift over time.
The most-exposed roles all share a common trait: they are built around routine cognitive work. Data entry, processing, responding to queries, drafting standard documents, reconciling records. These are tasks AI tools handle well today. Clerks, receptionists, telemarketers, call centre workers, bookkeepers, software programmers, marketing professionals, and accountants all sit near the top of the exposure list.
The least-exposed roles have a different common trait: they involve physical presence and human judgment in unpredictable environments. Tradespeople, aged care workers, cleaners, truck drivers, and construction labourers are at the bottom of the risk table. Their work is harder to automate because it cannot be done remotely by a language model.
The demographic pattern in the data is significant. The high-exposure roles are disproportionately held by women and by university graduates. This means the people who invested most in formal education to enter white-collar careers face the most direct pressure from a technology that can replicate exactly the kind of structured, rules-based work that those qualifications were built around.
Global evidence reinforces this. A November 2025 study from Stanford found a 16% decline in early-career employment across the most AI-exposed occupations since late 2022. Research comparing firms that adopted AI early to those that did not found that AI-adopting firms reduced junior hiring by roughly 9% relative to non-adopters. The Australian government's own report notes that in the United States, firms are already replacing graduate intakes with AI, even if that pattern has not yet appeared in Australian hiring data.
For accounting specifically, the picture is more nuanced than simple replacement. A Stanford study found accountants using AI tools can support more clients per week and close monthly books 7.5 days faster. The profession as a whole is projected to grow, but bookkeeping and clerical accounting roles face a projected 6% decline. The shift is not accountants disappearing; it is the base of the profession narrowing while the advisory top grows. Junior staff are under the most pressure, because AI now does much of the routine work that used to be how new hires learned the craft.
On regulation, Australia has taken a lighter path than Europe. The government has officially stepped back from earlier proposals for mandatory guardrails on high-risk AI systems, arguing that existing laws are adequate. A new AU$30 million AI Safety Institute will be established, but it sits inside the Department of Industry. The government has, however, held firm on one significant point: it ruled out any law that would allow AI developers to use Australian-created content, journalism, music, or books to train their systems without paying for it.
The gap between what the data shows and what the policy response looks like is the real tension here. The jobs most at risk belong to the most educated segment of the workforce. The growth in those roles is already slowing. The government's plan leans heavily on retraining and upskilling as the answer. Whether that is enough, and fast enough, is the question none of the reports can answer yet.