Australia’s RBA Finds Early AI Pressure on Young Workers in Exposed Jobs

Australia’s Reserve Bank now tracks artificial intelligence the way it tracks inflation: with a dedicated monitoring lens, not a headline panic. On 9 November 2026, RBA economist Mike Major published Insight 2026-09: Monitoring Early Effects of AI on the Labour Market. The bottom line is cautious. Aggregate disruption is small. The warning lights are younger workers in high-exposure occupations such as software development, clerical work, and customer service.

Governor Michele Bullock echoed the same framing in her 9 November 2026 speech What Happens to the Economy as “Intelligence” Gets Cheaper?: little economy-wide displacement so far, tentative weaker hiring among youth in exposed roles, and heavy uncertainty about NAIRU if AI shifts how firms hire.

For developers and hiring managers, the insight matters because a G20 central bank is naming the same cohort our US enrollment data check and Anthropic’s internal hiring report already flagged: people trying to enter software and desk-work pipelines, not tenured specialists.

TL;DR

Job cards beside a hiring chart, illustrating how central banks compare AI-exposed occupations to labour demand signals

What the RBA built

Major’s note is a screening tool, not a verdict. It links two channels where AI adjustment might appear before national unemployment moves:

1. Occupation exposure — Jobs and Skills Australia (JSA) automation mapping by ANZSCO code, derived from Gmyrek et al (2023) task analysis.
2. Youth and graduates — cohorts firms might trim first if AI substitutes for junior tasks.

The RBA compares outcomes before and after November 2022. That month matters because ChatGPT put generative AI in mainstream use, the same breakpoint Major uses in every chart.

Critically, high automation exposure often correlates with high augmentation exposure in the same occupation. Software roles can be automated on some tasks and augmented on others. The RBA warns that exposure tables alone cannot tell you which force dominates for a given firm.

Job ads: the leading indicator

Job advertisements usually move before employment stocks. Major maps JSA and SEEK postings into exposure quintiles.

What the RBA reports:

• Ad flows in the most-exposed quintile slowed a little more than less-exposed groups since November 2022.
• Occupations contributing most to the JSA ad decrease include general clerks, software and applications programmers, ICT business and systems analysts, call or contact centre workers, and receptionists.

SEEK and Indeed also track the share of ads mentioning AI keywords, which rose quickly (RBA Graph 4 in the insight). That can signal demand for AI skills rather than displacement. Some rise may be wording inflation — recruiters adding “AI” to legacy role descriptions.

A headset beside a hand receiving a parcel, echoing customer-service and contact-centre roles in the most AI-exposed quintile

Flows versus stocks

Job-finding rates and employment-to-unemployment transitions show little difference across exposure groups so far (Graphs 5–6). Volatile flow metrics can hide slow burns, but the RBA’s point stands: Australia is not seeing mass layoff spikes tied to exposure scores.

Employment growth and total hours worked look somewhat weaker in more exposed occupations (Graphs 7–8). Restricting to market-sector jobs does not erase the gap, so non-market hiring strength is not the whole story.

Unemployment-rate comparisons are more tentative because participation choices move the needle.

The youth gap inside exposed jobs

Youth outcomes are noisy. Australia’s youth labour market was tight in 2022, then eased. Major still highlights a pattern when you cross age with exposure:

Within the most AI-exposed occupations, younger workers show weaker employment and larger unemployment increases than older workers in the same jobs. That age split does not appear in less-exposed occupations (Graphs 13–16 in the insight).

That is the same shape — not the same magnitude — as US work the note cites, including Stanford Digital Economy Lab’s “Canaries in the Coal Mine” payroll study and Bank of Canada “Early Signs” research. Our global displacement audit graded similar US claims as real but easy to overstate.

A magnifying lens over a performance profile, echoing how the RBA screens youth outcomes inside exposed occupations

Australia-specific caveats Major repeats:

• Remote work may sideline younger workers in some sectors (RBA cites Emanuel et al 2026 US work on the mechanism).
• Post-pandemic cyclical easing hit youth harder after an exceptional 2022 peak.
• February 2026 is the latest complete occupation slice; ANZSCO-to-OSCA reclassification will force mapping updates later in 2026.

What the RBA does not claim

Major is explicit that the note is partial and descriptive:

• Exposure is potential, not measured AI use on the job.
• Liaison interviews still show most firms not cutting headcount because of AI over the next year (Governor speech Graph 8).
• Wage effects remain limited in literature the RBA cites.

So when a headline says “the central bank says AI is killing graduate jobs,” the accurate short version is: the RBA sees early divergence in youth outcomes inside exposed occupations, and will keep watching. It is not a certified causal attribution to ChatGPT.

How this fits global surveys

The insight stitches Australian charts to international context:

explainx.ai covered the Anthropic numbers in depth. The RBA uses them as one company-level example, not as proof for every Australian employer.

What builders and hiring managers should do

If you hire entry-level software or operations roles, assume central banks and regulators will read your posting volume, not just your press release. A stable aggregate unemployment rate will not shield you from scrutiny if junior pipelines thin while senior headcount holds.

Practical responses that match the evidence:

• Hire for AI-augmented workflows, not AI-free nostalgia. SEEK data in the insight shows complement occupations still posting.
• Document apprenticeship paths — Major warns about a future experience shortage if junior rungs disappear, echoing Anthropic’s “cognitive commons” concern.
• Separate cyclical cuts from tool substitution in workforce planning. The RBA’s comparator is November 2022, but some US researchers argue exposed-occupation softness predates ChatGPT (Frank et al LinkedIn study, cited in Decoder summaries Major’s footnotes mirror).

For students, pair core engineering with domains where exposure scores underweight judgment: regulated industries, on-site operations, and roles requiring accountable sign-off. Our UK entry-level survey shows employers self-report similar cuts; the RBA adds official macro monitoring.

Limits of the dataset

Major lists constraints worth copying into your own workforce models:

• Occupation codes lag real job titles (prompt engineer, agent ops, eval analyst).
• Exposure maps miss task bundling — automating one task does not delete a job.
• May 2026 labour force data are partial during a survey methodology transition.
• AI keywords in ads overstate skill demand when recruiters chase buzzwords.

Until firm-level adoption surveys catch up, central banks will lean on exposure × cohort screens exactly like this one.

Major closes by stressing that wage pass-through from AI productivity remains uncertain, and that liaison interviews still show most firms expect flat headcount over the next year even when they adopt copilots. That gap between task exposure and stated hiring plans is why the RBA will pair these charts with firmer adoption data before treating any youth divergence as structural.

What people are asking about RBA AI jobs data

Does Insight 2026-09 change Australian monetary policy tomorrow? No. Major frames the work as monitoring input. Governor Bullock’s speech ties AI to productivity, inflation pass-through, and possible NAIRU shifts, but both documents stress uncertainty. Markets should not treat a youth gap chart as a rate-cut trigger.

Is software development “dead” for graduates in Sydney and Melbourne? The RBA names programmers among high-exposure occupations with softer ads and hours, not mass unemployment. SEEK’s own economist work, cited in the insight, still shows AI-complement roles posting strongly. The nuance is entry difficulty, not zero hiring.

How is this different from vendor layoff headlines? Company announcements mix AI, rates, and post-COVID over-hiring. A central bank note compares cohorts across exposure quintiles on national statistics. It is slower, duller, and harder to spin — which is why it is useful for long-term planning.

Should I pause a computer science degree? The RBA does not advise individuals. It does imply that first-rung tasks in exposed roles face more competition from tools. Pair fundamentals with evaluated delivery: testing, security review, data handling, and customer-facing communication. Those layers appear in both automation and augmentation exposure scores.

Related reading

• Did AI actually take these jobs? A 2026 data check — how explainx.ai grades displacement claims
• Anthropic’s own employment shift — junior hiring and agent hours inside an AI lab
• US CS enrollment decline — pipeline fear versus jobs data
• Good night, good luck: software jobs and Indeed data — posting trends developers debate online
• India IT entry-level hiring claims — another large market’s early-career debate
• OpenAI education guidance gap — how vendors respond to classroom and exam stress

RBA insight figures, occupation mappings, and JSA exposure definitions reflect publications accessed 15 November 2026. Labour market series revise as new survey waves publish.