Australia’s AI Jobs Evidence Shows Slowing—not Mass Displacement

Australia’s official labour-market monitoring does not show AI eliminating jobs at scale. It finds continued employment growth even in the occupations most exposed to automation, although that growth has been slower than in less-exposed work.

The evidence therefore supports concern about uneven transition, not a claim of mass displacement. Occupational exposure measures what AI could automate; workplace adoption measures whether people or firms use it; displacement requires evidence that adoption caused positions to disappear. Australia has evidence for the first two, a tentative signal of relative slowing, and no reliable national count of AI-caused job losses.

The observed slowdown is real—but limited

The Australian Government’s official AI and employment report found that employment in the most-exposed fifth of occupations grew by 5.6% between November 2022 and February 2026, compared with 9.5% in the least-exposed fifth. Employment rose in both groups, so the comparison shows a growth gap rather than an absolute collapse.

The report’s statistical model produced a modest negative signal: an occupation with exposure one standard deviation above average had employment about 2% below the level implied by its pre-ChatGPT trend. That is a modelled difference from a counterfactual trend, not a finding that employment fell by 2% or that AI eliminated those positions.

The result also weakened under alternative exposure measures and some modelling choices. Routine cognitive and clerical work had been losing employment share before generative AI became widely available, while post-pandemic normalisation, sectoral demand and broader economic conditions may have affected hiring. The model measures potential exposure by occupation; it does not observe which firms adopted AI or identify particular jobs created or removed because of it.

An evidence ladder separates four different claims

Putting the available measures in order shows how far each conclusion can safely go:

• Observed employment — high confidence: employment in both the most- and least-exposed occupational groups was higher in February 2026 than in November 2022. Broad labour-market conditions, youth outcomes and occupational reshuffling did not display the pattern expected from economy-wide disruption.
• Relative slowing — moderate confidence: employment, hours worked and job advertisements generally performed less strongly in highly exposed occupations. The direction appears across several indicators, but its size and statistical significance depend partly on methodology.
• Workplace adoption — moderate confidence: surveys show substantial AI use among their respondents. They establish that tools are entering work, but not how much of each role is automated or whether headcount changed.
• AI-caused displacement — low confidence: the available national analysis cannot determine how many positions have disappeared specifically because an employer implemented AI.

The missing links matter. A job can contain automatable tasks without its employer deploying AI. A firm can adopt AI to assist employees while retaining or expanding headcount. Employment can also slow because of economic or occupational trends that have nothing to do with the technology.

Adoption evidence does not equal a job-loss count

Firm-level hiring research provides a useful counterexample to a simple replacement story. A CSIRO summary of Australian research says a study of job advertisements from more than 4,000 firms found identified AI adopters posted 36% more non-AI advertisements over time than non-adopters after adjustments for firm size, industry and location.

That result measures advertised labour demand, not completed hires, and the underlying period was 2020–2023. The firms were classified through signals in job postings, and the data largely predates widespread use of newer generative systems. It cannot establish what later tools will do, but it demonstrates why adoption should not automatically be recorded as displacement.

Employee use is a different measure again. The Hays FY26/27 workforce survey reported that 60% of respondents regularly used AI at work, while 78% had received no formal employer training. Its methodology covered more than 6,500 professionals across Australia and New Zealand, including 5,223 Australian respondents, between 6 February and 1 March 2026.

Those percentages describe a recruited and weighted professional survey, not every Australian worker. They nevertheless identify a preparation gap inside a large workforce sample: regular tool use is substantially more visible than organised employer training.

Who faces the greatest transition risk

The evidence does not produce a definitive ranking of workers likely to lose their jobs. It does support a more cautious risk screen: transition pressure is most plausible where high task exposure, active workplace implementation and limited training occur together.

Routine cognitive occupations—including clerical and administrative work—deserve particular attention because many tasks can be standardised and completed digitally. Exposure still applies to potential task automation, not the full value of a role. Highly exposed occupations also have not moved uniformly: some have weakened, some have remained broadly stable and some software-related work has grown strongly.

Workers may face disruption even without redundancy if routine tasks shrink, responsibilities are redesigned or employers demand a broader skill set. Entry-level employees could be vulnerable when simpler tasks that previously built occupational knowledge are automated, but the official monitoring found no significant national deterioration among young graduates. That remains a plausible transition risk rather than an observed economy-wide outcome.

The clearest reading of Australia’s evidence is therefore narrower than either optimism or alarm. Slower growth in exposed occupations merits continued attention, especially where employers are changing workflows without structured preparation. It does not yet establish mass displacement, and adoption or technical exposure alone cannot fill that evidentiary gap.

Also read:

• AI Adoption Leads to Job Growth, Not Mass Layoffs: New Study Finds Companies Hiring More After Implementing AI
• Tufts Report: 9.3 Million US Jobs at Risk from AI in the Next 5 Years — And the Hits Are Coming for High-Tech Hubs, Not Rust Belt Towns
• Part 7. 95% of “Blockchains” Are Not Blockchains At All — They’re Centralized Databases in Disguise