The federal government's first attempt to systematically track artificial intelligence's effect on the labour market has found that university graduates and women are disproportionately concentrated in the occupations most exposed to automation, while tradespeople and care workers are among the least exposed The Guardian.

The report, AI and Employment in Australia, was written by the Office of the Chief Economist within the Department of Employment and Workplace Relations and published on 8 July DEWR. It is billed as a first-of-its-kind national exercise — the first time Canberra has formally tracked this data — with a commitment to keep monitoring and reporting the trends on a regular basis.

The headline finding will be the one that gets repeated at Senate estimates: AI has not yet produced widespread job losses in Australia. That's a carefully hedged claim, and worth reading as exactly that — a snapshot, not a guarantee. The report draws its occupation-exposure modelling from Jobs and Skills Australia, which ranks roles by how much of their work is made up of "routine cognitive" tasks — the kind generative AI tools are best at replicating.

On that measure, the occupations flagged as most exposed include telemarketers, call centre workers, clerks, retail managers, software programmers, accountants, receptionists, and advertising and marketing professionals. The least exposed list reads almost like a rebuttal to anyone who assumed university credentials were a hedge against automation: tradespeople, aged care workers, carers, truck and forklift drivers, cleaners and gardeners. The report notes that workers in the high-exposure occupations skew female and university-qualified, while those in the low-exposure group have the least university education and the most vocational training.

That inversion — degrees correlating with exposure, trade certificates correlating with insulation — is the part of this report that should unsettle the usual policy assumptions in Canberra about upskilling being synonymous with a university pathway. If the routine cognitive tasks sitting inside clerical, accounting and marketing work are the ones large language models eat first, then a decade of "learn to code" and "go to uni" advice needs at least a footnote.

Employment Minister Amanda Rishworth framed the findings as evidence for continued government intervention rather than cause for alarm, saying the government is "determined to ensure AI is harnessed to create good jobs, not threaten them" and that Australians would be given "the skills, training and pathways needed to adapt and benefit" The Guardian. It's the standard ministerial register for this kind of report — reassurance paired with a promise of support — and it will need to be tested against whatever actually turns up in next year's edition of the tracker.

The report doesn't shy away from citing the bear case. It quotes Anthropic chief executive Dario Amodei's prediction that AI could eliminate half of all entry-level white-collar jobs and push unemployment to between 10 and 20 per cent within one to five years. Including that forecast in an official government document is itself notable — it's a genuinely alarming number to put in the mouth of an industry insider rather than a critic, and it gives the opposition and crossbench something concrete to quote back at the minister in question time.

The timing isn't incidental to the broader policy cycle. The Albanese government was expected to reveal updated plans on AI regulation — covering industry settings, economic policy and safety guardrails — in the week following the report's release The Guardian. Releasing the employment data first gives the government an evidentiary base to point to when it unveils whatever guardrails follow, and lets it frame regulation as a response to measured labour-market risk rather than a reaction to industry lobbying or public anxiety.

For anyone working the DEWR or Jobs and Skills Australia brief, the practical question is what the "regularly reported" commitment actually means in cadence and methodology. A one-off report is a press release; a genuine tracking series is a policy instrument, and one that will eventually be used to justify decisions on retraining funding, visa settings for AI-adjacent occupations and industrial relations disputes over automation clauses. Whether this becomes the former or the latter is the thing to watch over the next few reporting cycles, not this one.