Six findings from the federal government's flagship Australian workforce report — and our take on what each one means for the people trying to navigate AI disruption. Sourced from Jobs and Skills Australia's 2025 Jobs and Skills Report, its Gen AI Capacity Study, the Indeed Hiring Lab AU, and Deloitte Access Economics. Read time: ~12 minutes.
Jobs and Skills Australia's 2025 Gen AI Capacity Study scores every Australian occupation on two independent measures. Automation exposure is the share of tasks AI can do instead of the worker. Augmentation exposure is the share AI can do alongside the worker. The two scores aren't opposites — most occupations are higher on augmentation than automation. JSA describes Gen AI's labour-market impact through three mechanisms: automation, augmentation, and adaptation (the structural reshaping of jobs over time).
Sources: JSA Occupation data on AI exposure, 2025.
This is the most important framing in the report — and it's also the one that misleads people. The temptation is to read "augmentation, not automation" as "you're probably fine". You aren't. If your job is being augmented, that means AI is now doing significant chunks of what you used to do. The colleague next to you who learns to direct the AI doing those chunks will be three to five times more productive than you within a year. That's how people lose jobs to AI — not because the AI took the job, but because someone with the AI took it.
Two scores for every occupation in Australia, from JSA's 2025 Gen AI Capacity Study. Below is a curated 16-occupation slice illustrating the patterns. The full dataset covers 714 occupations.
These roles are mostly structured information processing. AI can both help with the work and do significant chunks of it without the worker. Pivoting is the realistic response.
Knowledge work that AI helps with a lot but rarely does end-to-end. The role isn't disappearing — it's concentrating. The person who masters AI in the role keeps it (and does the work of three). The person who doesn't, doesn't.
Care, trades, and clinical work. AI augments these roles (often usefully) but rarely does the task itself — the work is physical, situated, or requires judgement and presence machines don't have. These are credible places to land if you're pivoting out of the first cluster.
Source: Jobs and Skills Australia, Our Gen AI Transition — Occupation data on AI exposure, 2025. Scores are 0–1: the share of an occupation's tasks classed as highly augmentable or highly automatable by current Gen AI. View the JSA source.
JSA's CGE (computable general equilibrium) modelling identifies the occupations projected to experience the steepest declines by 2050. Named: General Clerks, Receptionists, Accounting Clerks and Bookkeepers, Sales/Marketing/Public Relations Professionals, and Business and Systems Analysts and Programmers. The corresponding industry exposure lands in Retail Trade, Public Administration, Financial and Insurance Services, Professional/Scientific/Technical Services, and Rental/Real Estate.
Sources: JSA Our Gen AI Transition, 2025; JSA 2025 Jobs and Skills Report Chapter 4.
The most striking thing on this list isn't the clerical roles — those are widely expected. It's that "Business and Systems Analysts and Programmers" is on the decline list. Programming as a profession is in JSA's modelled-decline column. Not at the top, but on it. At the same time, ICT Managers are projected to grow 25.5% by 2035. The lesson is that AI doesn't take whole occupations — it eats from the bottom of the skills ladder and grows the top. If you're a junior coder, that's a problem. If you're a senior architect or engineering manager, it's an opportunity. The pivot from one to the other is the most credible career shape for a lot of mid-career tech workers right now.
JSA's employment projections to 2035 (built with Victoria University's VUEF model, calibrated to Australian Treasury macro data) show Health Care and Social Assistance adding +541,900 jobs (+22.9%) — the largest growth in any industry. Professional, Scientific and Technical Services adds +250,100 (+18.5%) but the growth concentrates in management and senior specialist roles. Construction adds +160,900 (+11.9%). Top growth occupations by 10-year rate: Physiotherapists +35.1%, Health and Welfare Services Managers +27.1%, Dental Assistants +26.8%, ICT Managers +25.5%, Nursing Support and Personal Care Workers +24.7%.
Victoria leads on absolute persons added: +579,500 (+15.3%) over the decade — bigger than NSW.
Sources: JSA Employment Projections 2025–2035, with Victoria University.
Two things in this data that are usually missed. First, the growth is not where most "future of work" content points. It's not in AI itself — it's in allied health, dental assisting, aged and disability care, education aides, and skilled trades. Less glamorous, much bigger, much more accessible to people without technical backgrounds. Second, the Melbourne-first GTM bet is well-founded: Victoria's projected ten-year increase in absolute persons is the largest of any state.
If you're advising someone in their forties on where to pivot, "into trades" or "into the care economy" beats "into AI" almost every time. The latter sounds smarter; the former is correct.
Deloitte Access Economics 2024 (cited in JSA's 2025 report): approximately 27% of Australian workers use Gen AI at work without formal employer awareness or sanction — the so-called "shadow use" pattern. Indeed Hiring Lab Australia (2026) finds AI-mentioned job postings on the platform nearly doubled in a year (3.3% → 6.2%, Feb 2025 to Feb 2026), and that ~30% of Australian job postings sit in high-AI-exposure occupations. Separately, 57% of ASX200 companies mentioned AI investment in their 2024 annual reports.
Sources: Deloitte Access Economics 2024; Indeed Hiring Lab AU, 2026.
The "shadow use" stat is the most underrated number in the whole field. It means more than one in four of your colleagues are already using AI to be faster than you — and they aren't telling you, their manager, or HR. The competitive pressure isn't theoretical or coming; it's here, distributed unevenly across desks. The asymmetry is what hurts: the worker who used Gen AI for two hours yesterday and saved four hours on a task isn't going to walk into your standup and explain why.
This is the single best argument for structured upskilling. If a third of the workforce is teaching itself AI in the shadows, the bottom half of the workforce is going to be on the wrong side of a productivity gap within the year.
JSA's Recruitment Experiences and Outlook Survey (REOS) — Youth module 2024, n=2,500 employers — asked employers why they hired the young person they hired. Top reasons: "had the right attitude" — 60%. "Had relevant experience" — 19%. "Good communication skills" — 18%. "Had sufficient qualifications or skills" — 8%. Qualifications come in below "interviewed well". JSA's broader shortage analysis identifies "suitability gap" as the primary shortage driver for many engineering and management occupations — employers have qualified applicants but consider them not job-ready.
Source: JSA REOS Youth module, 2024; JSA 2025 Jobs and Skills Report Chapter 3.
Most "AI upskilling" content treats the credential as the product. The credential is the ticket to the interview. What actually decides the hire — and the keep — is employability skills: attitude, communication, judgement, work-readiness. A course that teaches AI tools in isolation and certifies completion is selling a ticket. A course that teaches AI tools through projects that build the judgement and comms muscles around them is teaching the thing that matters. We aim for the second.
This finding is also why we don't believe in "AI literacy" courses as a category. AI literacy without the surrounding employability skills is a half product. JSA agrees — Recommendation 3 of the 2025 report calls for embedding digital and AI skills into the design and delivery of existing qualifications, not replacing them.
JSA's 2025 report: roughly one in four Australians lack the digital connectivity or skills needed to engage with Gen AI tools. Digital exclusion is highest among First Nations people, those with disability, older adults, and people with lower education or income. JSA writes: "Without targeted support, Gen AI risks deepening existing inequalities and leaving vulnerable cohorts further behind." The National Foundation Skills Strategy 2025–2035 now classifies digital literacy as a foundation skill alongside reading, writing and numeracy — but Australia is still operating on PIAAC adult-skills data from 2011–12 (next survey 2025–2029).
Sources: JSA 2025 Report Chapter 2.5 and Chapter 4.3.1; National Foundation Skills Strategy 2025–2035 (Commonwealth of Australia, 2024).
This is the part of the AI workforce conversation almost nobody is having. The AI augmentation story assumes you can use AI in the first place. For one in four working-age Australians, that assumption is wrong. Without specific, targeted support — affordable connectivity, accessible training, employer-funded uplift for older or low-foundation-skill workers — AI doesn't accelerate productivity for everyone; it just widens the gap between the people who already had digital fluency and the people who didn't. That's exactly what JSA warns about, and it's the reason Pathway C exists in our model.
These six findings describe the landscape. They don't answer the only question that actually matters: given your specific occupation, your location, your existing skills and your financial runway — what's your credible next move?
That's what the assessment does.