The Australian Jobs Most Exposed to AI

Quick answer: Jobs and Skills Australia's 2025 Gen AI Transition study found that around 4% of the Australian workforce is in occupations with high exposure to AI automation, while 79% sits in low-exposure occupations. The jobs projected to lose the most employment by 2050 are concentrated in routine clerical and administrative work: general clerks, receptionists, and accounting clerks and bookkeepers top the list, alongside some sales, marketing, business analysis and programming roles. The study's central finding is that generative AI is, so far, more likely to change how these jobs get done than to remove them outright. If your work is script-based, data-entry-heavy or built around structured processes, call centre, admin or claims work in particular, you're in the group this data flags as most exposed. That's not a countdown clock. It's a signal that most people in your position have somewhere between 6 and 18 months to build the skills that sit on the other side of the automation line, before the choice gets made for them.

Find out where you actually sit. Take the free Career Change Check to see roles people from your kind of work moved into.

What "AI exposure" actually means, versus what people fear

Most of what circulates about AI and jobs is a headline number stripped of its method. "AI will replace 40% of jobs" sounds precise. It usually isn't sourced to anything you can check. Jobs and Skills Australia (JSA), the federal agency that produces Australia's official labour market data, took a different approach in 2025, running the first whole-of-economy study of how generative AI maps onto actual Australian jobs, task by task, using the same occupation classification (ANZSCO) that underpins the rest of the country's employment statistics.

That distinction matters. JSA isn't asking "could a robot do this job." It's asking, for every task inside every occupation, two narrower questions: could generative AI assist a person doing the task, and could it do the task without a person at all. Those two questions produce very different answers, and conflating them is where most of the scaremongering comes from.

Both concepts matter for reading the ranking below correctly. Exposure tells you where the pressure sits. Dynamism tells you how much give there is around it.

The ranking: which Australian jobs are most exposed

JSA's headline number is this: about 4% of the current Australian workforce is in occupations with high automation exposure to generative AI, 17% is medium, and 79% is low, based on the Overarching Report published 14 August 2025. Read plainly, that's a reassuring top-line number. Four in a hundred workers, not forty.

But averages hide where the exposure actually concentrates. The report is explicit about the pattern: "the higher potential for automation is concentrated in routine clerical and administrative roles." Many clerical tasks that survived earlier waves of computerisation, because they still needed a person to read, judge and type, are now squarely within what generative AI can attempt.

At the occupation level, JSA modelled which roles would lose the most employment by 2050 relative to a baseline without generative AI. The five occupation groups (at the 3-digit ANZSCO level) projected to lose the most jobs are:

• General clerks
• Receptionists
• Accounting clerks and bookkeepers
• Sales, marketing and public relations professionals
• Business and systems analysts, and programmers

The industries carrying the biggest losses track the same pattern: retail trade, public administration and safety, financial and insurance services, professional, scientific and technical services, and rental, hiring and real estate services. These are industries built on high volumes of structured, repeatable, document-heavy work, exactly the profile the study's task-level scoring flags as most automatable.

It's worth sitting with the flip side too, because it shows the pattern isn't random. The five occupation groups JSA projects to gain the most employment by 2050 are cleaners and laundry workers, midwifery and nursing professionals, business administration managers, construction and mining labourers, and hospitality workers. Physical presence, direct care, and complex people-management sit outside what current Gen AI tools can do. Data entry, documentation and routine correspondence sit right inside it.

Exposure is not a death sentence: what adaptation and dynamism add

Here's where the calibrated read differs from the scare piece. JSA's own framing is blunt: "Augmentation generally outweighs automation." As the study puts it, current Gen AI technologies are "more likely to enhance workers' efforts in completing tasks, rather than replace them, especially in high-skilled occupations." Even within the most exposed occupations, most individual tasks score higher on augmentability than automatability. That's the difference between a job changing shape and a job disappearing.

The study also found no evidence, yet, of widespread displacement of entry-level jobs in Australia. JSA Commissioner Professor Barney Glover was careful on this point when the report landed: "This has the potential to displace people in some jobs, particularly administrative and clerical roles" (Information Age, 20 August 2025). Minister for Industry, Innovation and Science Tim Ayres said generative AI may lead to "some displacement in the labour market", similar to earlier waves of technological change, and dismissed doomsday predictions of mass job losses as narratives pushed by so-called "tech boosters" (Information Age, 20 August 2025).

That's the calibration worth holding: the government's own data body isn't saying nothing will happen. It's saying something will happen, unevenly, over years rather than months, concentrated in specific task types, and shaped by how fast employers adopt the technology and how well workers adapt around it. Adaptation, in JSA's language, is what happens after adoption: employers reorganising how work is divided up, and workers building the skills and mobility to move with that reorganisation. Dynamism is what makes that possible in the first place, whether there are enough adjacent roles and enough employer willingness to redeploy people rather than simply cut headcount.

The named wedge roles, read against the study

If you work in call centre, admin or claims processing, here's how the JSA data actually maps onto your job, rather than the generic "AI is coming for admin" version you've probably already seen.

Admin and back-office roles are the closest match to JSA's own top-5 list. General clerks, receptionists, and accounting clerks and bookkeepers are named explicitly as the occupations projected to lose the most employment. If your day is built around data entry, correspondence, scheduling or record-keeping, this is the part of the study that applies to you most directly. See our breakdown of admin and back-office AI exposure for which tasks are exposed and which aren't.

Claims processing doesn't appear as its own named occupation in JSA's headline lists, but it sits inside financial and insurance services, one of the five industries flagged with the greatest projected losses, and it shares the task profile the study calls out: structured documentation, rules-based decisions, and high-volume correspondence. That's a strong directional match even without a claims-specific score. We've gone deeper on which parts of the claims workflow are genuinely at risk in what survives in claims processing.

Call centre and customer service work is the least precisely quantified of the three in JSA's public data. There's no single "call centre operator" line in the headline tables. But the report notes notable ASX200 company investment in AI-driven customer service, and finds augmentation effects are strongest in data entry, documentation and research tasks, a large share of what call centre work involves once you strip away the phone call itself. Treat this as directionally consistent with the clerical pattern above rather than an exact published figure. We unpack the mechanics in why AI is absorbing call centre work.

Across all three, the pattern holds: these jobs don't vanish overnight. The routine, structured slice of the task list shrinks first, and what's left over time is the part of the job that needed judgement, empathy or an actual human decision all along. Whether your role still has enough of that left to be viable in three years is worth checking rather than guessing at. See the full picture of at-risk jobs in Australia for how this plays out across the wedge occupations we track.

What to do with your 6 to 18 months

Nobody can tell you the exact date your tasks get automated, and JSA doesn't try to. What the study gives you instead is a way to read your own situation honestly: which of your daily tasks score high on automatability, which score high on augmentability instead (meaning you're more likely to keep the job but do it differently), and how much dynamism exists around your occupation if things shift.

Three concrete moves, in order:

1. Get your own exposure read, not the occupation-average one. JSA's numbers are aggregated to occupation level. Your actual mix of tasks sits somewhere on that spectrum, and it's worth knowing where. The free Career Change Check (linked above) shows roles people from your kind of work moved into, and the skills you already have for each. The exposure figure JSA published for your occupation is on the role page for it.
2. Look at where the dynamism actually is. Growth is concentrated in care, health, construction and hospitality-adjacent roles, and in management and coordination work generally. If your role sits in a low-dynamism corner of a high-exposure occupation group, that's the signal to start looking sideways now, not after a redundancy letter. Our career change pathways map realistic moves that build on skills you already have.
3. Build the augmentation skills, not the AI-literacy buzzwords. The tasks Gen AI struggles with are judgement calls, exceptions, relationship management and accountability. Roles and workplaces that lean into those skills are the ones most likely to adapt rather than shrink.

None of this requires panic. It requires using the window you've got.

FAQ

Will AI take my job in Australia? Most people worried about this are in low or medium exposure roles, based on JSA's 2025 data (79% low, 17% medium, 4% high). Generative AI is currently more likely to change parts of a job than remove it entirely, especially where judgement, exceptions or relationship management are involved. Routine, script-based or data-entry-heavy roles carry the highest risk of task loss over time.

What percentage of Australian jobs are at risk from AI? Jobs and Skills Australia's Gen AI Transition study (14 August 2025) found about 4% of the workforce sits in occupations with high automation exposure, 17% medium, and 79% low. High exposure concentrates in routine clerical and administrative roles rather than being spread evenly across the economy. Retail, financial services and public administration carry the biggest share of that high-exposure work.

Which Australian jobs are safest from AI? JSA's projections show cleaners and laundry workers, midwifery and nursing professionals, business administration managers, construction and mining labourers, and hospitality workers gaining the most employment by 2050. These roles rely on physical presence, direct care or complex people-management, which current generative AI tools can't replicate. If your current role sits far from routine data entry or scripted communication, that's a genuine buffer, not just a coincidence.

Is call centre work at risk from AI in Australia? JSA doesn't publish a specific exposure score for call centre roles, but the study notes notable company investment in AI-driven customer service and finds augmentation effects are strongest in data entry, documentation and research tasks, a large share of call centre work. That makes it directionally similar to the clerical roles the study flags as most exposed.

How is AI job exposure measured in Australia? JSA scores every task within an occupation on two measures, both 0 to 1: augmentability (can Gen AI assist a human with it) and automatability (could Gen AI do it without one). These task scores are rolled up to occupation-level exposure using the ANZSCO classification, giving a whole-of-economy view rather than a single guessed percentage.

Get your own read. The national numbers are a starting point, not your answer. Take the free Career Change Check to see roles people from your kind of work moved into.

Related reading

• the jobs AI can't replace
• where exposed workers are actually landing
• safer shortage destinations

Sources

• Jobs and Skills Australia, Our Gen AI Transition: Implications for Work and Skills, Overarching Report, 14 August 2025. jobsandskills.gov.au/publications/generative-ai-capacity-study-report
• Jobs and Skills Australia, Our Gen AI Transition Analysis Papers, 2 September 2025. jobsandskills.gov.au/publications/our-gen-ai-transition-analysis-papers
• Jobs and Skills Australia, Our Gen AI Transition, Final release, 30 September 2025. jobsandskills.gov.au/publications/our-gen-ai-transition-final-release
• Jobs and Skills Australia, Our Gen AI Transition: Exposures, Adaptation, Dynamism (occupation data, ANZSCO v1.3 unit level). jobsandskills.gov.au/studies/generative-artificial-intelligence-capacity-study/occupation-data-on-ai-exposure
• Tom Williams, "Aussie jobs most vulnerable to AI outlined in govt study," Information Age (Australian Computer Society), 20 August 2025.

For the full list of official data sources we cite across this site, see our sources page.

*This article draws on Our Gen AI Transition, Jobs and Skills Australia, © Commonwealth of Australia, licensed under Creative Commons Attr