The map now exists, and it points at the office
The automation debate has left the realm of theory and landed in ministers’ inboxes. A May 2026 MBIE briefing to the Labour Market Ministers Group, released to Stuff under the Official Information Act, gives the most current official picture of where AI hits New Zealand employment. The headline finding is deceptively calm: only 2% of current occupations are at high risk of full automation at today’s capability levels, chiefly software programmers, ICT professionals, accountants and call-centre operators.
But the more useful number is the 36% of occupations with high potential for AI augmentation, concentrated among managers and higher-skilled knowledge workers. In other words, the exposure sits squarely in the white-collar core of most businesses, not the factory or the paddock.
Officials were reassuring about the pace. “Scenarios of AI leading to rapid displacement and mass unemployment appear unlikely at this stage,” the briefing noted, pointing out that new and disappearing jobs already account for 13 to 15% of all roles in a typical year. Churn is normal. AI, they argue, gets absorbed into it.
What employers are actually doing
That calm sits awkwardly next to what firms say they are already planning. Data from the AI Forum, cited in the same briefing, found 45% of surveyed businesses were planning to lower headcount as they implement AI. That is not a modelled scenario. That is stated employer intent in 2026.
Both things can be true. Displacement may be gradual and largely absorbed through natural attrition and retirement. But the gap between official reassurance and boardroom behaviour is the part worth naming, and it means the risk for individual firms is not abstract.
The exposure sits in the $50k to $150k layer
The Reserve Bank’s February 2026 analytical note sharpens the picture. Using 10 large language models to score exposure across occupations, it found roughly 30% of NZ workers face high combined exposure to AI and robotics. Crucially, AI exposure is most strongly associated with professional, managerial and administrative roles earning between $50,000 and $150,000 a year. That is the engine room of the average medium-sized New Zealand company.
Machinery operators and labourers show far lower AI exposure. And in a counterintuitive twist, the RBNZ found workers with no qualifications are more exposed to AI than those with secondary or tertiary qualifications, upending the old assumption that automation only threatens manual work.
Roles get hollowed before they disappear
The better framing than redundancy comes from a June 2026 Newsroom analysis drawing on US labour economist David Autor’s work. “Technology rarely replaces whole jobs. It replaces tasks,” it argues, and it takes the legible tasks first, the parts of a role that can be written down and optimised.
Applied to an accounts manager, a legal secretary or a customer service lead, the point holds: “thousands of roles will be hollowed of their measurable tasks, leaving a remainder nobody has thought to value.” For an owner, the question is not whether a role vanishes, but what it looks like in three years and whether you are training for the remainder.
The youth problem is structural
MBIE flags a specific casualty: entry-level roles being automated before young people can climb onto the ladder. Youth unemployment already sits near 15%, triple the wider working-age rate. Overseas, firms using AI to automate foundational tasks are hiring fewer junior staff, a pattern now appearing here. Cut the junior intake and you starve your own pipeline of experienced staff a decade out.
Where NZ is insulated
The silver lining is structural. In 2025, RNZ analysis noted that roughly 70% of NZ exports come from agriculture, horticulture, seafood and forestry, sectors demanding physical dexterity and sensory judgement that AI cannot yet replicate. That commentary argued NZ’s advantage lies in “deepening its strengths in sectors AI cannot yet touch, food production, care and infrastructure.”
The data that matters lands in September
Here is the catch. The MBIE briefing was largely adapted from Australian research rather than NZ-specific data, which officials concede limits how conclusive it can be. To fix that, MBIE is surveying 20,000 businesses, with findings due in September 2026.
That means the most important piece of local workforce evidence is still roughly two months away. Firms making headcount and training calls right now are doing so without it. The prudent move is not to wait. The map already exists, and it says the exposure runs through the office, not the shop floor. Employers who start redesigning roles and rethinking junior hiring now will be planning; those who wait for September, and for a redundancy wave to force the question, will be reacting.
Sources
- Stuff: The Kiwi jobs at highest risk of AI automation (2026-07-17)
- Newsroom: How the Govt’s wager on AI could backfire (2026-06-24)
- Newsroom: AI regulation can wait until public sector is cut, says Goldsmith (2026-05-22)
- RNZ: AI is peeling back the layers of ‘low-value’ work – NZ may be well-placed to adapt (2025-08-13)