Deep dive · May 2026

JSA exposure dataset — what's at risk, how people protect themselves.

All 357 Australian occupations in the Jobs and Skills Australia Gen AI exposure dataset, mapped to the Three Pathways. Seven findings. For each one: what the data shows, who is at risk, and the concrete protection action — by pathway. Sources: JSA Gen AI Capacity Study, 2025 (CC-BY 4.0). Read time: ~10 minutes.

← Back to Insights

357unique Australian occupations scored by JSA
0.73correlation between augmentation & automation — they move together
154occupations face both high augmentation and high automation (43%)
91in the top automation-exposure quartile — should consider Change industry
2×high-augmentation roles relearn skills twice as fast as low-augmentation roles
139low-exposure manual occupations — the Pathway B destinations
Finding 01 · Filter and explore

Every occupation, mapped on both exposure dimensions.

Across all 357 occupations, augmentation and automation correlate at r = 0.73. They are not opposites. The roles AI helps with most are also the roles AI can do without you most.

Showing all 357 occupations. Hover any dot for the role, sector and scores. Filter to a sector to see how its workforce splits across the quadrants.
Upskill-rich
(high aug + low aut)
→ Upskill pathway
27
Double exposure
(high aug + high aut)
→ Upskill + exit option
154
Displacement risk
(low aug + high aut)
→ Change industry
37
Low exposure / manual
(low aug + low aut)
→ Pivot destination
139
0.10.10.20.20.30.30.40.40.50.50.60.60.70.70.80.8Automation exposure → AI does it INSTEAD of the workerAugmentation exposure → AI does it WITH the workerDOUBLE EXPOSURE · 154UPSKILL-RICH · 27DISPLACEMENT RISK · 37LOW EXPOSURE / MANUAL · 139
What's at risk
  • Self-misclassification. People in Double exposure roles read JSA's "augmentation not automation" framing as reassuring and assume they only need Pathway A.
  • The 154 occupations in the top-right quadrant (43% of the workforce in the dataset) — most knowledge work.
  • Specifically: Bookkeepers (aug 0.77, aut 0.69), Software Programmers (0.77 / 0.63), Accounting Clerks (0.74 / 0.71), Marketing Pros (0.79 / 0.54), Financial Dealers (0.77 / 0.69).
How people protect themselves
  • Take the quiz to classify accurately against your specific role and confirm which pathway combination applies.
  • If you're in Upskill-rich (top-left, 27 roles): commit to Pathway A. AI multiplies you and won't easily do the work alone.
  • If you're in Double exposure (top-right, 154 roles): run Pathway A and Pathway B in parallel. Build the AI fluency for your current role now, and validate one Pathway B target this year.
  • If you're in Displacement risk (bottom-right, 37 roles): Pathway B is the priority. There is no AI-augmented version of your role to upskill into.
Take the AI Readiness Quiz →
Finding 02 · Pathway sizing

How the workforce splits across the Three Pathways.

Splitting at the median of each score (aug = 0.65, aut = 0.33), the 357 occupations distribute as: 27 Upskill-rich, 154 Double exposure, 37 Displacement risk, 139 Low-exposure manual. The "clean" quadrants (pure upskill or pure displacement) cover only 18% of the workforce. The remaining 82% need a combined strategy.

Routing all 357 occupations through the Three Pathways278%15443%3710%13939%All 357 occupationsUpskill-rich (27)→ Upskill (Pathway A)Double exposure (154)→ Upskill primary, Change industry as exit optionDisplacement risk (37)→ Change industry (Pathway B)Low exposure / manual (139)→ Pivot destination for Pathway B
What's at risk
  • Wrong-pathway routing. Workers picking only Upskill when they're in Double exposure end up over-invested in a role that disappears; workers picking only Change industry when they're in Upskill-rich leave years of compensation on the table.
  • The 154 Double-exposure workers who treat one pathway as sufficient.
  • The 37 Displacement-risk workers who hope upskilling will save a role with no AI-augmented version.
How people protect themselves
  • Upskill-rich → Pathway A only. Direct your existing role with AI; don't waste cycles on a pivot you don't need.
  • Double exposure → Pathway A + Pathway B together. 80/20 split: A is the day-job investment, B is the contingency you validate quarterly.
  • Displacement risk → Pathway B primary. Identify a destination in the Low-exposure quadrant (Finding 03 names the safer sectors), build the bridge skill in 12–18 months.
  • Stranded (within Displacement risk) → Pathway C. If your role is in the Top-12 stranded list (Finding 05) and you can't realistically self-pivot, the Soft Landing support frame is built for you.
Finding 03 · The sector league table

Which Australian sectors are most exposed.

Top three sectors by mean automation exposure: Accounting, Banking and Financial Services (0.60), Administration and Human Resources (0.58), Information and Communication Technology (ICT) (0.52). Bottom three: Automotive (0.24), Personal Services (0.24), Construction, Architecture and Design (0.24).

Mean automation exposure by sector (sectors with ≥8 occupations)Bar length = mean automation score. Dot = mean augmentation score.0.10.20.30.40.50.60.7Accounting, Banking and Financial Services (n=14)0.60Administration and Human Resources (n=27)0.58Information and Communication Technology (ICT) (n=11)0.52Sales, Retail, Wholesale and Real Estate (n=18)0.45Advertising, Media and Public Relations (n=8)0.43Legal and Insurance (n=9)0.43Education and Training (n=16)0.40Electrical and Electronics (n=9)0.35Hospitality, Food Services and Tourism (n=25)0.34Arts and Entertainment (n=12)0.33Transport and Logistics (n=13)0.32Health and Community Services (n=48)0.30Government, Defence and Protective Services (n=8)0.30Engineers and Engineering Trades (n=16)0.29Agriculture, Animal and Horticulture (n=24)0.28Manufacturing (n=26)0.25Automotive (n=9)0.24Personal Services (n=10)0.24Construction, Architecture and Design (n=37)0.24
Bar = mean automation Dot = mean augmentation
What's at risk
  • The four high-automation sectors — Banking & Finance, Admin & HR, ICT, Sales — together cover 70 occupations and a significant share of Australian white-collar employment.
  • Sector-wide compression, not just role-specific. When a whole sector's mean is > 0.45, demand for new headcount slows even in roles that look safe individually.
  • Workers in Health, Education, and Government have high augmentation but low automation (aug 0.65–0.67, aut 0.30–0.40) — they still need Pathway A; the work won't disappear but the way it gets done will.
How people protect themselves
  • If you work in Banking, Admin, ICT or Sales: treat the next 24 months as a runway. Commit to Pathway A inside your sector AND identify a Pathway B destination in the Low-exposure floor (Construction, Personal Services, Automotive, Manufacturing).
  • If you work in Health, Education or Government: Pathway A is sufficient and urgent — the work isn't going, but tooling is changing under it. Skip Pathway B unless you have other reasons.
  • Targets for Change-industry pivots: the four lowest-exposure sectors all sit at aut ≤ 0.25 and offer hands-on, locally-bound work AI can't perform.
See Pathway B destinations →
Finding 04 · The skill-change treadmill

The more AI helps you, the faster your skills go stale.

JSA's "rate of skill change" rises with augmentation exposure: 2.87%/yr in the low-aug tertile, 4.00%/yr in the mid-aug tertile, 5.77%/yr in the high-aug tertile. ICT occupations average 10.1%/yr; Multimedia & Web Developers sit at 18.3%/yr — the highest in the dataset.

Rate of skill change vs augmentation exposure024681012Skill change rate (% of skills changing per year)2.9mean (median 2.6)Low augn = 814.0mean (median 3.7)Mid augn = 845.8mean (median 5.2)High augn = 79↑ Web developers off-chart (18.3%/yr)
What's at risk
  • One-and-done AI training. A 2024 prompt-engineering course was load-bearing then and is dead weight now.
  • Knowledge workers in high-aug roles whose employer paid for a single workshop and considers them "AI-ready".
  • ICT staff and adjacent roles — relearning at 10%+/yr means full skill-base rotation every decade, with the front of the wave (web/dev/data) rotating in 5 years.
How people protect themselves
  • Treat Pathway A as a subscription, not a course. Quarterly refresh minimum; monthly for ICT-adjacent.
  • Block calendar time — 2 hours/week for tool tracking, 1 day/quarter for substantive re-learning. Without a calendar commitment it does not happen.
  • Anchor learning to one production tool a quarter, not generic literacy content. Replace a real workflow; don't just watch demos.
See the Pathway A cadence →
Finding 05 · Stranded occupations

Twelve occupations where automation is high and the exit door is narrow.

Subtracting transition-fit rate from automation exposure ranks the most stranded roles in the Australian economy. The top five: Telemarketers (aut 0.81 / trans 0.07), Keyboard Operators (aut 0.81 / trans 0.08), Call or Contact Centre Workers (aut 0.75 / trans 0.09), Filing and Registry Clerks (aut 0.76 / trans 0.15), Debt Collectors (aut 0.69 / trans 0.10).

Top 12 displacement risk — high automation, low transition fitBar = automation exposure. Diamond = transition fit (how easily someone in this role can pivot).Credit & Loans Officersaut 0.68trans 0.14General Clerksaut 0.71trans 0.16Accounting Clerksaut 0.71trans 0.16Bookkeepersaut 0.69trans 0.14Tourism & Travel Advisersaut 0.73trans 0.18Receptionistsaut 0.66trans 0.11Insurance, Money Market & Statistical…aut 0.68trans 0.13Debt Collectorsaut 0.69trans 0.10Filing & Registry Clerksaut 0.76trans 0.15Call or Contact Centre Workersaut 0.75trans 0.09Keyboard Operatorsaut 0.81trans 0.08Telemarketersaut 0.81trans 0.07
What's at risk
  • Australian finance back-office staff — bookkeepers, accounting clerks, credit and loans officers, insurance & statistical clerks — are the largest concentrated stranded population in the dataset.
  • Contact-centre and inbound-comms roles — telemarketers, call centre workers, receptionists, debt collectors.
  • Common pattern: the in-role skills don't port cleanly to anything else, so "just upskill" doesn't apply and self-directed pivots fail.
How people protect themselves
  • Pivot now, not later. Wait-and-see in these roles is the worst strategy — the transition window narrows as more colleagues hit the market at once.
  • Choose a destination from the Low-exposure floor deliberately (Finding 03). Healthcare assistant, trades apprentice, logistics, hospitality team-lead are all common Pathway B destinations from clerical work.
  • Use Pathway C as scaffolding. The Soft Landing programme exists precisely so people in these roles don't have to engineer a pivot alone.
See Pathway C support →
Finding 06 · Strategy split

For most high-augmentation roles, specialising pays more than generalising.

For the 181 above-median-augmentation occupations with both scores reported: mean specialisation potential = +1.6, mean hybridisation potential = -10.2. The deeper-in-domain move out-scores the broader-across-domains move.

Hybridisation vs Specialisation — for the 181 above-median-augmentation occupations-30-25-20-15-10-5510-10-8-6-4-2246810← deepen specialisation → broaden specialisation← drop hybridisation → add hybrid skillsSoftware & Applications ProgrammersBookkeepersAccounting ClerksInsurance AgentsChemists & Food & Wine ScientistsSpecialist PhysiciansJournalists & Other WritersAmusement, Fitness & Sports Centre Ma…Nutrition ProfessionalsArchitectural, Building & Surveying T…Advertising & Marketing ProfessionalsPublic Relations Professionals
What's at risk
  • Generic "AI literacy" investments bolted onto an existing role, without deepening domain expertise.
  • The "T-shaped professional" rhetoric — for most high-aug occupations, the data says it underperforms a deeper-specialist play.
  • Specifically the high-spec roles: Chemists, Specialist Physicians, Journalists, Insurance Agents, Marketing Professionals, Bookkeepers, Accounting Clerks — all reward going deeper, not broader.
How people protect themselves
  • Default Pathway A move: deepen your domain expertise AND learn to direct AI inside that specialism. Not "AI literacy + your job"; "your job, done with AI you direct".
  • Exceptions (hybrid-rich roles): if you're a Centre Manager, Surveying Technician, Nutrition Professional, or Other Engineering Professional — these reward adding adjacent capability. Pathway A content should lean toward cross-domain tooling for you.
  • Test before committing: a quarter spent deepening tends to produce visible output gains; a quarter spent generalising often doesn't. Use that as a real-world signal.
Finding 07 · The hard-to-pivot sectors

Some sectors can't easily pivot — even when exposure is low.

Transition-fit rates vary by an order of magnitude between sectors. Lowest: Agriculture, Animal & Horticulture (0.05), Personal Services (0.06), Health & Community Services (0.09). Highest: Engineers & Engineering Trades (0.40), Electrical & Electronics (0.34), Construction (0.30).

Transition-fit rate by sector — who can actually pivot from where they areMean share of plausible high-fit pivot pathways from occupations in each sector.0.00.10.20.30.40.5Science (n=4)0.08Health and Community Services (n=26)0.09Education and Training (n=13)0.13Arts and Entertainment (n=6)0.13Hospitality, Food Services and Tourism (n=20)0.14Information and Communication Technology (ICT) (n=10)0.17Sales, Retail, Wholesale and Real Estate (n=12)0.18Transport and Logistics (n=9)0.18Legal and Insurance (n=4)0.19Government, Defence and Protective Services (n=4)0.21Manufacturing (n=5)0.23Administration and Human Resources (n=23)0.24Accounting, Banking and Financial Services (n=13)0.24Construction, Architecture and Design (n=21)0.30Electrical and Electronics (n=7)0.34Engineers and Engineering Trades (n=12)0.40Executive and General Management (n=4)0.49
What's at risk
  • Care workers when sector-level funding tightens — Health & Community is Australia's largest employer, but average transition-fit is 0.09. The work is low-automation but skills don't port.
  • Agriculture and Personal Services workers facing region-specific shocks (drought, tourism downturns, demographic shifts).
  • Arts & Entertainment (trans 0.14) — high creative skill, narrow alternative paths.
How people protect themselves
  • Build an adjacent qualification before you need it. Care workers: pick up a Cert IV in Mental Health, Disability, or Allied Health Assistance while still employed.
  • Use Pathway C as a community, not only as crisis support. The Soft Landing peer network is the network you'll need before displacement, not after.
  • Engineering / electrical / construction workers: your transition-fit is high, so a Pathway B move is genuinely available if your specific sub-sector slows down — you're better-positioned than most.
Strategic summary

If you take one thing from this page.

Find your row. The right pathway combination is what the data points to — not a personality test.

If you're in… The risk The protection Pathway
Upskill-rich roles (27 occupations)
Engineering managers, civil engineers, school principals…		 Over-pivoting. Don't waste cycles changing industry; your role is the win. Commit to Pathway A. AI fluency inside your existing specialism, quarterly cadence. Upskill
Double-exposure roles (154 occupations)
Bookkeepers, programmers, marketing pros, accountants…		 Treating one pathway as enough. AI helps you AND can do the work without you. 80/20 split — Pathway A as the day-job investment, Pathway B as a contingency you validate this year. Upskill + Change industry
Displacement-risk roles (37 occupations)
Telemarketers, contact centre, clerks, debt collectors…		 "Upskill into AI" doesn't apply — no AI-augmented version of the role to upskill into. Pathway B is primary. Target a Low-exposure destination (Construction, Personal Services, Health support, Trades) in 12–18 months. Change industry
Stranded subset (~10–12 occupations)
Top automation quartile AND bottom transition quartile		 Cannot realistically self-pivot — in-role skills don't port; transition fit < 0.1. Use Pathway C as scaffolding for the pivot. Community, structured support, dignity. Not a substitute for Pathway B — a way of getting through it. Soft Landing → Change industry
Low-exposure manual roles (139 occupations)
Trades, care work, cleaners, drivers, hospitality…		 Sector-specific shocks (funding cuts, downturns) where transition fit is low (Care, Agriculture). Pathway A optional, Pathway B not needed for AI reasons. Build an adjacent qualification as insurance. Pathway C community is useful pre-emptively. Optional Upskill

Take the AI Readiness Quiz →   The quiz uses these same rules to route you to the right pathway combination for your specific role.

Methodology

How we built this.

Source: Jobs and Skills Australia, Occupation data on AI exposure (2025, CC-BY 4.0). The raw file contains 714 rows: each ANZSCO 4-digit occupation appears twice — once in "All occupations" and once in its sector matrix group. We deduplicated by ANZSCO code, keeping the sector-tagged row where available, giving 357 unique occupations.

Quadrants are split at the dataset median (aug 0.65, aut 0.33). Top/bottom quartiles use standard quartile cutoffs. The "displacement risk" composite is automation exposure − transition fit; "stranded" means top-quartile automation and bottom-quartile transition fit simultaneously. All numbers, charts and tooltip values are generated directly from the CSV at research/market/jsa-ai-exposure-occupations.csv.

Companion reading: our reading of the JSA 2025 Jobs and Skills Report and Which Australian industries AI is actually hitting.