Australia’s government has published its first formal assessment of how artificial intelligence could reshape the workforce, with the findings suggesting some office-based jobs are far more exposed than hands-on roles.
The report, AI and employment in Australia, used labour market data to map which occupations are most vulnerable to automation and which are, at least for now, relatively protected.
Published by the Department of Employment and Workplace Relations on 8 July 2026, the report is based on labour market data through February 2026 and focuses on whether employment has shifted away from occupations more exposed to generative AI since the public release of ChatGPT in late 2022. Its authors say the purpose is monitoring rather than prediction, and that the evidence to date does not show broad AI-driven upheaval in Australia’s jobs market.
Its findings point to a clear divide: jobs built around routine cognitive work are far more likely to be disrupted by AI than roles that depend on physical skill, practical judgement or experience gained on the job.
Clerical workers, programmers and administrative staff were named among the occupations facing the highest exposure, while caregivers, tradespeople and drivers were identified as being at the lower end of the risk scale.
The report said telemarketing, advertising and accountancy were especially exposed, largely because they involve repeatable, codified tasks that AI systems and large language models are increasingly capable of handling.
On the other hand, roles that require direct manual ability, such as plumbing or operating a forklift, were described as broadly insulated from AI disruption for the time being.
The government’s analysis also found that the most-exposed occupations have grown more slowly than less-exposed ones since November 2022. Employment in the most-exposed fifth of occupations rose by 5.6% over that period, compared with 9.5% for the least-exposed fifth, while job advertisements and hours worked in high-exposure occupations also lagged. The report says these patterns are consistent with modest softening in exposed roles, but not proof of a large AI-driven decline.
The document also said that this split has a noticeable gender component.
It said ‘nearly 70% of workers in the least-exposed group’ were male, since male-dominated manual trades tend to sit outside AI’s current reach.
Women, meanwhile, were ‘thought to be susceptible to automation because they are disproportionately employed in exposed occupations’, including many of the clerical and administrative roles flagged as high-risk.
Younger workers were also identified as being more at risk in some areas.
The report explained that AI models are ‘generally best at performing tasks that require codified knowledge’, or ‘book-learning’, but are ‘less capable of replacing tacit knowledge, the idiosyncratic tips and tricks that accumulate with experience’, which younger employees ‘disproportionately lack’ compared with seasoned workers.
The broader warning extends across a range of white-collar occupations. Translation work, for example, could shift toward human review of machine-generated drafts as large language models become more capable of processing technical and corporate documents.
Retail could also see further change. Beyond standard self-checkouts, AI may play a larger role in checkout systems, including stores where shoppers place items in a basket and are charged automatically.
Research-focused roles may also come under pressure. According to AI, it will be able to carry out tasks like researching competitor pricing, customer data, sentiment analysis, and putting all of these together into a final report.
This, the generated answer said, would then be used by humans in the final strategic decisions.
Content production is another area where automation could grow. That includes product descriptions for online retail as well as captions and ads for social media.
The need for a large volume of content makes this something that an automatic text generator could do, the AI said.
Legal support work was also highlighted. These are the professionals who review documentation including from previous cases, which can then be passed on to lawyers working on a case.
AI’s generated answer said that it could scan the documents very quickly, getting rid of this need.
Proofreading and copy-checking may also be vulnerable. The AI’s answer said that language models can be programmed specifically for things like grammar, meaning that it can catch errors.
As it can do this quickly, that means that it could also replace human proofreaders.
Bookkeeping and related finance administration were listed as another possible target for automation. These concern tasks like managing and monitoring expenses, as well as making sure that business ledgers are properly balanced.
In its answer, the AI said that future models could carry out these tasks.
Customer service roles may be affected as well, especially in situations where people call a company for support.
Large Language Models can be used in triaging customer tickets when they are calling up with questions.
Telemarketing was another role singled out as highly exposed. Also known as cold-calling, the AI-generated answer said that these roles could be performed by an AI instead of a human.
This is because they can often follow scripts, and AI can now be used to generate voices to make these calls.
At the top of the list was the data entry clerk, a role centred on working across databases and shifting values between systems.
The low cost of using AI software in this sort of task could mean that it’s a job which is very likely to be replaced.