Two Australias: Who Gains and Who Gets Left Behind in the Age of AI

There is a word that comes up in almost every report about AI and work. The word is augment. Workers will not be replaced, the reports say they will be augmented. Made more powerful. Given tools that amplify everything they can already do.

That is probably true, for some workers.

The Stack Overflow Developer Survey 2024 asked 65,437 developers across 24 countries whether they were already using AI tools in their work. Among the 1,260 respondents in Australia, 54.5% said yes. And when Australian developers who use AI tools are compared to those who do not, there is already a measurable salary gap, AU$14,000 in median annual compensation separating the two groups.

That gap is the story. Not because it is large today, but because of what it represents: two cohorts of workers, in the same industry, in the same country, beginning to pull apart. One group is learning to use tools that amplify their output. The other is not yet.

Outside of tech, in the clerical offices and administrative floors that employ far more Australians, the question of who gets access to AI and who gets displaced by it is considerably more urgent. A landmark 2025 study by Jobs and Skills Australia assessed the generative AI exposure of 998 individual occupations. Its finding: the jobs with the highest automation potential are not the ones most Australians expected. They are not factories or warehouses. They are offices specifically, the administrative and clerical roles that have employed generations of Australian workers, and that previous waves of automation left almost entirely untouched.

The data reorders a common assumption: it is 45 to 54 year olds who have the highest AI adoption rate among Australian developers at 60.7%, while 35 to 44 year olds have the lowest at 51.4%. The differences between age groups are small. This is a workforce wide shift, not a generational divide. What is most striking is the second set of bars. Across every age group, fewer than one in thirteen Australian developers considers AI a threat. The workers most likely to be early adopters, those in technology have largely recognised that the tools are going to extend, not end, their careers.

The concern is whether that equanimity extends to the 600,000 Australians in clerical and administrative roles who face a fundamentally different kind of exposure.

Australian developers are measurably more cautious about AI than the global average. Globally, 71.9% of developers hold a positive view of AI tools in their workflow, in Australia that figure is 65.8%, with a meaningfully larger share sitting in the neutral or unsure category. This is not resistance fewer than one in thirteen Australians holds a negative view. It is uncertainty, and a slightly more measured national temperament toward technology that will reshape work.

That caution may matter more than it looks. Australia’s path through the AI transition will depend partly on how quickly AI skills spread through the workforce. A population that adopts AI tools more slowly also develops AI skills more slowly and falls further behind on the wage premium those skills are already commanding.

The shaded zone in the bottom right corner is where the problem concentrates. Secretaries and administrative assistants (94.7% female), receptionists (90.4%), bookkeeping and accounting clerks (87.2%), billing and posting clerks (88.4%), and payroll clerks (87.3%) all cluster there high female concentration, below pay parity, and classified as high automation risk by both Frey and Osborne’s landmark research and Australia’s own Jobs and Skills Australia analysis.

These are not marginal occupations. They are the bedrock of Australia’s service economy, employing hundreds of thousands of people who have built stable working lives in roles that every previous wave of technology left largely untouched. They now face their first serious displacement pressure and they are doing so with a smaller financial buffer than comparable male dominated sectors.

Among Australian developers in 2024, those currently using AI tools earn a median of AU$151.6k per year versus AU$137.5k for those not using and not planning to, a gap of AU$14.1k. This is a single industry, a single survey year, and likely reflects some selection effects: workers at better-resourced companies have more access to AI tools. But if access drives adoption, and adoption drives earnings, the implication is that the divide will grow fastest in companies already paying the most.

Sales and Administration are major category that captures secretaries, receptionists, bookkeeping clerks, and administrative assistants sits in the most exposed position on this chart. It is majority female (58.7%), well below pay parity, and the gap barely moved between 2013 and 2016 despite years of equal pay advocacy. It is also the category most directly targeted by generative AI.

This is the crux of the concern. If AI displaces workers in clerical and administrative roles and the evidence from Australia’s own Jobs and Skills Australia, the IMF, and Deloitte all points in that direction those workers face the transition with limited financial cushion, in a labour market where the available alternatives all require significant time and resources to enter.

The data in these five charts is drawn from real surveys of real workers, 65,437 developers globally, 1,260 in Australia, four years of occupational earnings data from the Bureau of Labor Statistics. What it shows, taken together, is that the AI divide is already visible in the industry that has moved fastest. Developers who use AI earn more than those who do not. Australians are adopting more cautiously than the global average. And the occupations facing the highest automation potential are those already employing the most women at the lowest relative wages.

Australia has a window to shape that transition. The evidence says it is not yet acting at the scale the moment requires.