Executive Summary
September was a month of broadening adoption and sharper national choices.
The strongest new evidence came from business and health. AWS reported that half of New Zealand businesses now use AI consistently, up from 42% a year earlier, while local examples showed AI being used in agritech, insurance, media analytics and customer operations. The Government also extended its AI Advisory Pilot to another 120 small businesses. aboutamazon.com.au
Health produced two notable applied-AI developments: Tāmaki Health reported promising early results from an AI Health Coach in primary care, and the Royal New Zealand College of General Practitioners began testing AI-supported simulated consultations for GP registrars. These are bounded, supervised use cases rather than autonomous clinical systems, but they extend the health story beyond emergency-department documentation. tamakihealth.co.nz
The month also moved AI further into public debate. The National Cyber Security Centre warned that frontier AI is rapidly changing the cyber-threat environment, while Labour released an election policy proposing an Office of AI, stronger oversight, creator protections and new requirements for large data centres. Datagrid’s Southland project continued into its next construction stage, but consent reviews and concerns about noise, water and energy showed that AI infrastructure is now a local political and environmental issue as well as an economic one. ncsc.govt.nz
The overall picture is broader and more operational than in August, but still fragmented. New Zealand is accumulating practical AI deployments and capability-building activity, yet the country still lacks a clearly visible national research-platform outcome, a comprehensive AI regulatory framework, or evidence of widespread enterprise-scale transformation.
What Happened in the Past Month
Business adoption became easier to measure
AWS released its Unlocking New Zealand’s AI Potential 2026 research on 22 September 2026, alongside its Cloud and AI Day in Auckland. The survey of 1,000 businesses and 1,000 individuals found that 50% of New Zealand businesses were consistently using AI, compared with 42% in the previous year. Among AI-adopting businesses, 84% reported revenue growth, although the research does not establish how much of that growth was caused by AI. aboutamazon.com.au
The more important finding was the gap between adoption and organisational readiness. Reporting on the AWS research said only 15% of businesses had a formal AI strategy, with businesses that had a strategy more likely to report advanced use and confidence in measuring returns. This aligns with Datacom’s August research, which found that 91% of larger New Zealand organisations were using some form of AI, but 81% remained in exploratory or implementation stages. reseller.co.nz
AWS also highlighted several New Zealand examples of AI moving into ordinary business operations. Halter has used an AI agent to help triage system alerts; ZURU is analysing 20,000 social videos a day; Parrot Analytics is using AI to verify content data; and Tower Insurance deployed an AI-powered contact-centre system. These cases suggest that the most visible commercial activity is still concentrated in workflow improvement and sector-specific applications rather than locally developed foundation models. aboutamazon.com.au
Skills demand moved beyond technical roles
The September employment data pointed to AI becoming part of broader business practice. SEEK reported that AI-related skills appeared in 4% of New Zealand job advertisements in August, up 93.3% year on year. AI appeared most often in information and communications technology, marketing and communications, and consulting and strategy roles. nz.seek.com
The composition of demand is also changing. SEEK’s July data showed particularly strong annual growth in advertisements mentioning AI ethics and governance, ahead of agentic and generative AI. This does not prove that New Zealand has a large new AI-governance profession, but it does indicate that employers are increasingly treating judgement, oversight and responsible deployment as part of AI capability. nz.seek.com
That finding matters because it reinforces a theme from the previous snapshot: New Zealand’s bottleneck is shifting from access to AI tools towards the ability to integrate them safely into existing organisations. Training, data management, accountability and workflow redesign are becoming at least as important as model selection.
Government extended practical support for small businesses
On 22 September 2026, the Government extended its AI Advisory Pilot until 30 June 2027 and added places for a further 120 small businesses. Around 150 businesses had already joined the pilot since its launch in January. The programme provides expert advice and co-funds tailored AI roadmaps and implementation costs, with support capped at $15,000. beehive.govt.nz
The announcement also said the Regional Business Partner Network had supported approximately 650 businesses through AI-related courses between July 2025 and June 2026. This is modest compared with the scale of the economy, but it represents a shift from broad awareness campaigns towards supported implementation.
The policy is consistent with the pattern seen in May, June and August: government intervention is focused less on building a New Zealand-owned general-purpose model and more on helping existing businesses apply commercially available systems. That approach may produce wider near-term productivity gains, but it also leaves firms dependent on overseas model and cloud providers.
Health AI expanded through supervised pilots
Health delivered some of the month’s clearest new evidence of applied adoption.
Tāmaki Health reported on 14 September 2026 that an AI Health Coach had been deployed across six primary-care clinics in partnership with New Zealand company Groov. In an early six-week evaluation, AI-supported outreach engaged seven times more people than comparable previous human-only text campaigns, generated 105 additional patient touchpoints without additional workforce time, and increased collection of patient-reported outcomes by 8%. The programme targeted people with poorly controlled diabetes who had not attended a clinic for at least six months. tamakihealth.co.nz
The results are preliminary and come from the provider rather than an independent evaluation. They nevertheless show a different model of health AI from the emergency-department scribe: AI is being used to extend human outreach and support between appointments, while clinical staff retain responsibility for escalation and care.
On 15 September, the Royal New Zealand College of General Practitioners announced a separate pilot with Gestalt, a New Zealand-built clinical reasoning platform. GP registrars, including international medical graduates, will use simulated consultations and receive structured feedback. The College said it would assess engagement, user experience and educator feedback before deciding whether the tool should have a wider role in training. rnzcgp.org.nz
Together, these initiatives suggest that health AI is broadening into patient engagement, clinical education and workforce support. It remains a supervised and evidence-building phase rather than a new wave of autonomous diagnosis or treatment.
Cyber risk became a national AI priority
The National Cyber Security Centre’s Cyber Threat Report 2026, released on 24 September 2026, gave AI a more prominent place in New Zealand’s national risk picture.
The NCSC said frontier AI is rapidly reshaping the cyber landscape and warned that advanced models could help malicious actors automate attacks, identify vulnerabilities and conduct more personalised targeting. Its assessment was that, by early 2027, criminal actors may have access to advanced AI capabilities currently available only through leading frontier models. ncsc.govt.nz
The report recorded 369 incidents of potential national significance handled during 2025/26, including 162 linked to criminal or financially motivated actors. The NCSC’s message was directed at boards and senior leaders: AI security is no longer only a specialist technology concern, and conventional cyber-security fundamentals remain essential even as attack capabilities become more automated. ncsc.govt.nz
This develops a theme present in earlier snapshots, when “shadow AI” and frontier-AI preparedness were emerging concerns. The difference now is that AI is being treated as part of the immediate operating environment for New Zealand organisations, not simply as a future technology risk.
AI infrastructure became more politically contested
Datagrid’s Southland project moved further into construction administration during September. Southland District Council records show that stage-two construction, environmental, traffic, landscape and community-liaison plans were approved on 4 September, with a notice of commencement of stage-two work dated 14 September. southlanddc.govt.nz
At the same time, an independent hearing commissioner decided that the council would not review Datagrid’s resource-consent conditions after community concerns about low-frequency noise, cooling systems, generators, acoustic modelling and other potential effects. The council said the legal threshold for a review had not been met. southlanddc.govt.nz
The project therefore moved forward physically while becoming more contested politically. The central questions are expanding beyond whether New Zealand can attract AI infrastructure. They now include who pays for energy and network upgrades, how water and land are managed, how local communities share in the benefits, and whether large facilities create durable national capability or mainly host offshore demand.
That debate was sharpened by Labour’s AI policy, released in September as part of its election platform. Labour proposed an Office of AI within the Department of the Prime Minister and Cabinet, an online-safety regulator, a framework for creator payment when copyrighted work is used by AI, and rules requiring large data centres to secure their own renewable energy, pay connection costs, manage peak demand and use water efficiently. These are proposals, not current law, but they show that AI infrastructure has entered electoral politics. nzlabourparty.client.trfg.au
Applied AI continued to reach primary industries
Agriculture provided a smaller but useful signal of sectoral depth. AGMARDT announced on 17 September 2026 that its transitional Agribusiness Innovation Grant round had funded a project by neXtgen Agri International to develop an AI-enabled ultrasound and computer-vision system for measuring intramuscular fat in live sheep. The project received $109,608, with industry co-funding also committed. agmardt.org.nz
The project illustrates where New Zealand may have a defensible AI advantage: combining local domain expertise, physical-world data and sector-specific export problems. It is not a large investment, but it is more representative of New Zealand’s likely applied-AI pathway than a bid to compete directly with global model providers.
The national AI ecosystem gathered, but did not yet consolidate
The Aotearoa AI Summit, held in Wellington on 8–9 September 2026, brought together business leaders, researchers, policymakers, investors and founders. Its programme focused on infrastructure, sovereign capability, trust, workforce transformation, agents, robotics, quantum technologies and practical implementation. aotearoaai.nz
The summit’s significance was ecosystem-level rather than the announcement of a single major programme. It showed that New Zealand’s AI conversation has become more connected across government, business, research and civil society. The AI Forum also announced finalists for the 2026 Aotearoa AI Awards and began work on a disability and accessibility workstream, widening the ecosystem conversation beyond productivity and commercialisation. aiforum.org.nz
Trend Line Across Recent Snapshots
The direction across the year is now clearer:
- March: AI activity centred on health deployment, privacy, cyber risk and compute infrastructure.
- May and June: government shifted towards practical adoption support, health procurement, public-service tooling and workforce readiness.
- July: responsible-use guidance, public-sector pilots, biosecurity applications and early startup activity became more visible.
- August: NZIAT launched, Datagrid began physical works, public-service use cases doubled and local vertical-AI startups attracted capital.
- September: business adoption became more measurable, health activity broadened, cyber risk became more urgent and AI governance became part of election politics.
The momentum is broadening rather than uniformly accelerating.
There are now visible signals across business, health, education, agriculture, public services, infrastructure and cyber security. However, many initiatives remain pilots, surveys, proposals or enabling programmes. The country is building the conditions for adoption faster than it is demonstrating economy-wide transformation.
The most important shift is conceptual. The national question is moving from whether New Zealand will use AI to how AI will be governed, paid for and translated into local value.
What Looks Quiet, Unchanged, or Early
The biggest unresolved issue remains the NZIAT AI Research Platform.
MBIE’s public programme material continues to describe the phase-two contest, the five shortlisted concepts and the availability of up to $70 million over seven years. A privately maintained September tracker now labels the initiative “funded” and “building”, but also says that delivery, utilisation and commercial proof remain open questions. The public record reviewed for this snapshot still does not clearly identify the final consortium, contract terms, delivery milestones or user-access arrangements. mbie.govt.nz
That matters more as the rest of the ecosystem becomes more active. New Zealand now has an operating advanced-technology institute, growing AI infrastructure, measurable public-sector adoption and more commercial pilots. The absence of a transparent account of the country’s central AI research investment is increasingly difficult to reconcile with that broader momentum.
There was also no September evidence of a major local frontier-model initiative, a large new New Zealand AI venture round or a nationwide private-sector deployment comparable to the emergency-department scribe rollout.
Enterprise transformation remains early. Datacom’s research found that only 22% of surveyed organisations had a dedicated AI leadership role, while 83% were concerned about offshore data risks and only 35% had a formal policy governing where critical data was stored and managed. datacom.com
Regulation is likewise still fragmented. Labour’s proposals would represent a more coordinated policy approach, but they are campaign commitments rather than enacted rules. New Zealand continues to address AI through existing privacy, online-safety, cyber-security, health and employment frameworks, alongside voluntary or sector-specific guidance.
Overall Assessment
September was a stronger month than August in breadth, although not in the form of one decisive breakthrough.
AI adoption is now visible in more parts of the economy. Businesses are reporting consistent use, health providers are testing AI-assisted outreach and training, job advertisements increasingly mention AI governance, and government is extending support for small firms. At the same time, the NCSC is treating frontier AI as an immediate cyber-security issue, while the Datagrid debate is forcing questions about energy, water, infrastructure and national value.
The overall pattern is selective acceleration with rising political stakes.
New Zealand is not becoming a frontier-model power, and there is still no evidence of a coordinated national transformation programme operating at scale. Its more credible path remains applied AI: sector-specific systems embedded in health, agriculture, public services, finance, manufacturing and other areas where local data, trust and domain expertise matter.
The next test is whether these distributed activities can be connected. That means making the NZIAT research investment transparent, turning business experimentation into repeatable productivity gains, validating health pilots independently, and setting clearer expectations for infrastructure and governance.
September’s message was therefore not that New Zealand has solved AI. It was that AI is becoming impossible to treat as a narrow technology issue. It is now part of the country’s productivity strategy, health-system reform, cyber posture, infrastructure planning and election debate.