Australia’s central bank has produced the most careful early read yet on what AI is actually doing to hiring — and its central finding is a quiet one: firms are changing how they hire long before they change how many people they employ. On 9 November the Reserve Bank of Australia published a staff note monitoring AI’s early labour-market effects, alongside a speech from Assistant Governor Sarah Hunter at the UBS Australasia Conference, as summarised by labour-market tracker Metaintro.
🔍 THE BOTTOM LINE: In the RBA’s business liaison, 75 percent of firms report no AI impact on headcount at all. The adjustment is happening at the hiring gate — about 12 percent of firms are slowing hiring or leaving roles unfilled, against 8 percent cutting jobs and 6 percent adding them — and the first visible pressure is on young workers in the most AI-exposed occupations. For New Zealand, whose labour market moves in Australia’s slipstream, this is the leading indicator worth watching before anything shows up in our own numbers.
The liaison numbers: quiet, uneven, hiring-led
In Hunter’s speech, “What Happens to the Economy as Intelligence Gets Cheaper”, the distribution of firm responses over the past year broke down like this:
- No change in headcount from AI: around 75 percent of liaison firms
- Slower hiring or roles left unfilled: around 12 percent
- Reduced headcount: around 8 percent
- Increased headcount because of AI: around 6 percent
That spread matters. A firm that stops backfilling a role changes its hiring without making anyone redundant — for job seekers it surfaces as fewer openings, not layoffs. It is also a two-sided picture: the note is explicit that AI may also stimulate demand by lowering costs and prices or creating new types of work, which is why the RBA is building monitoring rather than declaring an outcome.
Adoption itself is moving faster than any job impact: the share of liaison firms with no AI adoption has fallen from about a third in 2025 to close to zero in 2026, while “limited adoption” rose from roughly 40 percent to about 60 percent. Firms are using AI broadly and still not cutting headcount — the gap between adoption and employment effects is the story.
Where the early pressure shows: young workers
The RBA’s evidence for actual labour-market movement is deliberately hedged. Comparing outcomes against a November 2022 benchmark — the month generative AI went public — economist Mike Major’s note finds tentative signs of weaker job ad flows and employment outcomes in the most AI-exposed occupations, especially for people aged 15 to 24. In the most exposed group — software developers, clerical and administrative workers — average hours for young workers have fallen most clearly, while older workers’ hours stayed near their benchmark. The most-exposed occupations contributing most to the fall in new job ads since November 2022 are general clerks, software and applications programmers, ICT business analysts, call centre workers and receptionists.
The note’s own caution is as important as the finding: exposure measures capture potential, not realised effects, the same age groups are affected by ordinary cyclical conditions, and job-finding rates across exposure quintiles show no sustained divergence since ChatGPT’s release. The RBA frames this as something to monitor closely, not as displacement confirmed — consistent with other national evidence cited in the note, including Bank of England analysis of the UK’s hiring slowdown and New Zealand’s own RBNZ analytical note finding AI exposure concentrated in professional, managerial and administrative roles on $50,000 to $150,000 incomes. It also matches the earlier evidence check on NZ youth jobs data: soft entry-level conditions visible in the data long before anyone can causally blame AI.
What it means for New Zealand
New Zealand’s labour market tracks Australia’s closely enough that the RBA’s numbers are the nearest available preview of what Stats NZ will eventually print here. The RBNZ’s own exposure mapping found roughly 30 percent of NZ workers face high combined AI-and-robotics exposure with few fully insulated occupations, and our jobless rate — about 5.7 percent against Australia’s — sits higher in the cycle, which makes the hiring-gate pattern doubly relevant: when demand is soft anyway, “we just didn’t backfill” is the easiest way for AI exposure to show up. Recent local data already showed filled jobs rising only slowly through 2026.
The practical read for NZ graduates and young workers cuts both ways, and the data on each side comes from the same labour market. The RBA and Bank of England analyses flag the squeeze: early pressure at entry level in AI-exposed white-collar occupations. But Seek’s own research, built on millions of Australian job ads from 2016 to 2025, finds the opposite force operating too — AI augmentation may be increasing employer demand for early-career workers in knowledge-intensive occupations, with a smaller negative effect confined to the most automatable roles. Both can be true at once: automatable entry tasks thin out while augmented junior roles open up, and which side you land on depends increasingly on demonstrable AI skills — which is exactly where the signalling value sits. The aggregate headline will say “no jobs apocalypse”. The composition underneath is what careers are built on — and on the demand side, freelance platforms show the same re-bundling at client level, with AI-linked projects carrying 52 percent larger budgets.
FAQ
Are Australian firms cutting jobs because of AI? Not mostly. In the RBA’s liaison program, 75 percent of firms reported no headcount change from AI, 8 percent cut jobs and 6 percent added them. The most common change was at the point of hiring: about 12 percent reported slower hiring or unfilled roles.
What did the RBA find about young workers and AI? Tentative signs of weaker job ad flows and employment outcomes in the most AI-exposed occupations — clerical, administrative and software roles — particularly for 15-24 year olds, whose average hours have fallen most clearly since November 2022. The RBA cautions these patterns are not yet proof of AI displacement.
How fast is AI adoption in Australian firms? Fast. Firms with no AI adoption fell from about one-third in 2025 to almost zero in 2026, while limited adoption rose to roughly 60 percent — broad adoption with little headcount effect so far.
Does this apply to New Zealand? Directly as a leading indicator: NZ’s labour market tracks Australia’s closely, and the RBNZ’s own exposure work found about 30 percent of NZ workers face high combined AI and robotics exposure. The hiring-gate pattern RBA describes would show in NZ as slower new openings before any job losses.
— CJ Murden, editor of Singularity.Kiwi. Former digital technologies teacher, author of AI-focused books. Writing with a New Zealand focus.
