AI Forum NZ — Generative AI Working Group A living document
Living Whitepaper
Latest

AI Adoption in Healthcare in New Zealand: A Living Whitepaper

Updated: 10 June 2026

Executive Summary

AI adoption in New Zealand healthcare is still best described as targeted deployment under tightening control. The clearest scaled public-sector use remains ambient documentation in emergency departments, where AI scribe technology is live nationwide for 1,250 frontline staff, with further licences being progressed largely for mental health teams. Health NZ’s updated privacy statement also confirms current use cases now span transcription, consultation summarisation, document generation, staff knowledge tools, and preliminary image review, while stressing that many uses remain in pilot phase and that AI is not used to make automated decisions about patient care. (beehive.govt.nz)

The biggest shift since the previous update on 2 April 2026 is that imaging has moved one step closer to formal deployment. On 27 May 2026, the Government said procurement was underway to select a preferred AI mammogram-reading tool for testing and validation in BreastScreen Aotearoa, ahead of a planned rollout from early 2027. That makes breast screening the most visible next-wave national clinical AI programme. (beehive.govt.nz)

At the same time, governance has become more operational. Since April, Health NZ has updated its enterprise privacy statement, published staff guidance on generative AI and LLMs, and reinforced that employees must not put patient or sensitive organisational information into unapproved tools or use LLMs for clinical decisions. Separately, a University of Auckland-led evaluation framework endorsed by Health NZ’s Executive Leadership Team is now being used to assess all proposed AI tools across Health NZ services. (info.health.nz)

The overall picture is therefore clearer than it was in early April: New Zealand is still early-stage at a system level, but no longer merely exploratory. The live model is clinician-supervised augmentation; the fastest diffusion is in documentation-heavy workflows; the strongest clinical scale-up lane is imaging and screening; and privacy, equity, Māori data sovereignty, and cybersecurity are increasingly being treated as prerequisites for scale rather than side issues. (health.govt.nz)

What Has Changed Since 2 April 2026

  • Breast screening has advanced from exploration to procurement. Health NZ is now selecting a preferred AI mammogram-reading tool for testing and validation, with a planned rollout from early 2027. The proposed model would have AI perform one of the two independent reads, while clinicians remain central to diagnosis and follow-up decisions. (beehive.govt.nz)
  • Health NZ’s AI privacy position is more explicit and more current. Its privacy statement was updated on 30 April 2026 and now lists current AI uses including transcription, consultation summaries, document generation, staff knowledge tools, Microsoft Copilot-style support, and preliminary review of images, x-rays, scans, or mammograms. It also says many of these uses are still being tested on a limited basis. (info.health.nz)
  • Health NZ has published staff-facing generative AI rules. Updated in May 2026, the guidance says staff must not enter patient or other sensitive information into unapproved LLMs and must not use such tools for clinical decisions or personalised advice to patients. It also says Health NZ does not currently have any private LLMs available. (healthnz.govt.nz)
  • AI assessment inside Health NZ appears to be more standardised. A nationally tailored AI evaluation framework, endorsed in March 2026, is now being used to assess all proposed AI tools across Health NZ services. (transform.auckland.ac.nz)
  • Aged care has become a more visible adoption setting. On 12 May 2026, Metlifecare said it was rolling out the HEIDI ambient scribe tool to three more care homes after a four-month trial at Parkside Village. (metlifecare.co.nz)
  • The privacy environment tightened further from 1 May 2026. The Office of the Privacy Commissioner says the Health Information Privacy Code was amended in March 2026 to reflect IPP3A, with the updated code in force from 1 May 2026. (privacy.org.nz)

Current State of AI Adoption

1. Public hospitals: documentation remains the most mature live deployment

The most mature public-sector AI deployment in New Zealand healthcare remains the nationwide emergency department scribe rollout announced on 28 February 2026. The Government said the technology was live in all emergency departments, reaching 1,250 ED doctors and frontline staff, and that more than 1,000 additional licences were being progressed, predominantly for mental health teams. (beehive.govt.nz)

That rollout still matters strategically because it shows where Health NZ is currently most willing to scale AI: workflow-embedded, clinician-reviewed documentation support, rather than autonomous diagnosis or treatment. Health NZ’s privacy statement reinforces the same operating model by saying AI outputs affecting records or decisions must be reviewed by the responsible clinician and that AI is not used to make automated decisions about care. (beehive.govt.nz)

There is also now more public documentation around that deployment. On 28 April 2026, Health NZ published an OIA release on Heidi AI that it said provides an overview of governance, risk management, privacy safeguards, and early evaluation findings related to Heidi’s use in clinical settings. That publication is notable less for signalling a new rollout than for showing a stronger transparency-and-assurance layer around an already live deployment. (healthnz.govt.nz)

2. Primary care: rapid uptake is becoming more integrated and more governed

Primary care remains the fastest-moving non-hospital environment. A New Zealand survey paper published on 26 March 2026 described AI scribes as having had “rapid uptake” in primary care. In the search summary, the study reported that 40% of respondents had experience with AI scribes, while only 59% had sought patient consent and 66% had read the tool’s terms and conditions. (pubmed.ncbi.nlm.nih.gov)

Commercial offerings are also maturing. Medtech says its Medtech AI platform, available in New Zealand from 23 February 2026, is integrated directly with Medtech Evolution, stores all data locally in New Zealand, does not retain consultation audio, does not store patient names, dates of birth, or NHI numbers, and requires a clinician confirmation that explicit patient consent has been obtained before recording begins. The platform also positions itself as broader than a scribe, with structured notes, referral letters, patient summaries, and clinician-reviewed outputs written back into the PMS. (medtechglobal.com)

A useful signal of diffusion is the follow-on education layer. Medtech ran a nationwide 12+ location “education lab” roadshow across April and May 2026, explicitly aimed at clinicians and practice teams. That suggests the market is moving from simple experimentation toward repeatable onboarding, workflow redesign, and mainstream practice enablement. (medtechglobal.com)

3. Aged care: ambient scribing is spreading beyond hospitals and GP clinics

One of the clearest newer case studies is in residential aged care. Metlifecare said on 12 May 2026 that it would roll out the HEIDI ambient scribe tool across three more care homes after a four-month trial at Parkside Village involving 12 registered nurses. It said nurses reported less time spent on documentation, better continuity of care and handovers, and growing uptake to full adoption within the Parkside nursing team over a few weeks. Metlifecare also stressed that nurses retain responsibility for reviewing, editing, and approving all documentation, and that the tool is used with resident consent. (metlifecare.co.nz)

This matters because it broadens the New Zealand adoption story. AI documentation is no longer just a hospital ED or GP-clinic phenomenon; it is now visibly moving into aged residential care, where staffing pressure, handover quality, and time spent documenting are also major operational issues. (metlifecare.co.nz)

4. Screening and imaging: now the strongest next-wave clinical scale-up lane

Breast screening is now the most important post-April development. On 27 May 2026, the Government said procurement was underway for a preferred AI mammogram-reading tool, with testing and validation planned ahead of a rollout from early 2027. The proposed design is cautious: AI would perform one of the two independent reads already required in the screening process, with qualified professionals remaining central to all diagnosis and follow-up decisions. (beehive.govt.nz)

This is more significant than it may first appear. BreastScreen Aotearoa already screens about 270,000 women annually, and the programme’s age extension to 74 is underway, so the combination of rising volume and workforce pressure creates a strong operational case for AI-assisted reading. The Government is explicitly framing AI here as a capacity and early-detection tool, not a clinician substitute. (beehive.govt.nz)

The broader imaging lane remains credible beyond breast screening. Health NZ’s privacy statement confirms that AI is already being used to read, analyse, or review clinical information for preliminary results, including images, x-rays, scans, or mammograms, while noting that many such uses are still in pilot phase. Meanwhile, the Ministry of Health’s updated precision health page still cites the AI-equipped South Auckland diabetes retinal screening pilot as an example of current national activity while stating that New Zealand remains in the early stages of safe adoption. (info.health.nz)

Governance, Regulation, and Digital Foundations

1. The official stance is still “assist clinicians, do not automate care”

Health NZ’s updated privacy statement is unusually clear. It says AI is used in a closed environment, that personal information is not used to design commercial AI or third-party generative models, that short-term disclosures to providers are deleted shortly afterwards, and that anything generated by AI that could affect the record or clinical decisions must be reviewed by the responsible clinician. It also states plainly that Health NZ does not use AI to make automated decisions about a person’s healthcare. (info.health.nz)

Health NZ’s separate May 2026 guidance on generative AI and LLMs adds the staff-operating rules: unapproved LLMs cannot be used for patient data entry, clinical decisions, or personalised patient advice, and the organisation says it does not currently have private LLMs available. That is a strong signal that broad public-LLM adoption inside Health NZ remains tightly constrained. (healthnz.govt.nz)

The Medical Council’s guidance, published on 10 March 2026, remains the profession-level counterpart to this system stance. It says doctors remain responsible for all clinical decisions, must check AI output, must tell patients when AI is being used in their care, must obtain informed consent in specified circumstances including AI consultation recording, and must not use AI to impersonate themselves through avatars or chatbots. (mcnz.org.nz)

2. Governance is becoming more formal and more standardised

Health NZ’s National Artificial Intelligence and Algorithm Expert Advisory Group remains the core governance mechanism. Health NZ says all AI development or implementation plans must be registered with the group, which oversees ethical, technical, clinical, and operational standards and is intended to maintain an AI register and publish summaries of recommendations and a national register of AI in use. (info.health.nz)

The newly publicised evaluation framework strengthens that picture. According to the University of Auckland’s TRANSFORM programme, a New Zealand-specific framework was endorsed by Health NZ’s Executive Leadership Team in March 2026 and is now being used to assess all proposed AI tools across Health NZ services. Because it was designed after reviewing 31 international frameworks and adapted to the local health system, it suggests Health NZ is moving from ad hoc review toward a more repeatable national method for AI assessment. (transform.auckland.ac.nz)

3. Privacy and cybersecurity are now part of the adoption story, not background issues

The regulatory baseline also hardened on 1 May 2026, when the updated Health Information Privacy Code reflecting IPP3A came into force. For healthcare AI, that matters because New Zealand’s compliance environment is becoming more demanding around how indirectly collected personal and health information is handled. (privacy.org.nz)

Budget 2026 adds another important layer. On 28 May 2026, the Government announced $153.6 million for Health NZ cybersecurity improvements, including national monitoring, stronger data security processes, critical IT safety upgrades, and AI-enabled assessments to improve cyber maturity across primary care. This is not clinical AI adoption, but it is part of the infrastructure needed for AI adoption at scale. (beehive.govt.nz)

Research and Innovation Pipeline

New Zealand’s research pipeline remains stronger than its current live deployment footprint. The Health Research Council said in August 2025 that it had funded 10 AI in Healthcare studies worth $4.6 million, and those projects continue to shape the near-term implementation pipeline. (hrc.govt.nz)

The portfolio remains notably practical:

  • Acute stroke: a Health NZ-led project is measuring whether AI-assisted rapid brain-scan interpretation improves intervention rates and reduces treatment delays across NZ hospitals. (hrc.govt.nz)
  • Digital pathology for gastrointestinal cancers: Otago researchers are building AI tools that combine biopsy images and patient data to improve treatment decisions and reduce unnecessary surgery. (hrc.govt.nz)
  • Postoperative monitoring: Auckland researchers are using digital tools and AI to detect deterioration earlier after surgery, with explicit attention to privacy, equity, and data sovereignty. (hrc.govt.nz)
  • Healthy ageing: the NZ–Singapore programme includes AI-assisted interRAI assessment and predictive tools for ageing-related risks, designed with explicit attention to Māori, Pacific, and other diverse communities. (otago.ac.nz)

Taken together, this pipeline suggests that New Zealand’s next material advances are likely to come less from broad autonomous AI and more from high-need, measurable, clinician-supervised use cases in imaging, risk prediction, monitoring, and workflow support. That is an inference from the current deployment pattern and funded portfolio. (hrc.govt.nz)

Trust, Equity, and Social Licence

Recent New Zealand research continues to show that adoption will be constrained as much by trust as by technical capability. A February 2026 NZMJ viewpoint argued that patient trust is central to the successful implementation of healthcare AI in Aotearoa New Zealand and highlighted the importance of public benefit, clear governance, data protection, choice, clinician responsibility, and Māori representation. (nzmj.org.nz)

That concern is echoed in more specific community research. A 2026 study on Pasifika perspectives on AI for asthma management found interest in AI’s potential, but also persistent concerns around privacy, accuracy, access, and age-related digital divides. (openrepository.aut.ac.nz)

These findings line up closely with Health NZ’s own caution on public LLMs, its emphasis on clinician review, and its explicit references to Māori data sovereignty and bias risk in both governance and staff guidance. (healthnz.govt.nz)

1. Documentation-first adoption is no longer a pilot story

Ambient scribing has moved from isolated experimentation to a cross-setting pattern spanning emergency departments, general practice, and now visible aged-care deployments. (beehive.govt.nz)

2. Imaging is now the clearest clinical expansion lane

Breast screening procurement, continuing retinal-screening activity, and HRC-funded stroke and pathology projects all point to imaging and interpretation tasks as the most likely next area for scaled clinical AI in New Zealand. (beehive.govt.nz)

3. Governance is shifting from principle to operating rules

The Medical Council guidance, Health NZ privacy statement, Health NZ LLM rules, NAIAEAG process, and the new evaluation framework together amount to a much more operational governance stack than New Zealand had even a few months ago. (mcnz.org.nz)

4. Private and community providers are moving faster than high-risk hospital AI

Primary care vendors and aged-care operators are visibly shipping and rolling out tools faster than the public system is scaling higher-risk diagnostic AI. (medtechglobal.com)

5. Cybersecurity and privacy are becoming adoption bottlenecks and enablers

The combination of stricter privacy expectations, staff restrictions on public LLMs, and new cyber investment shows that safe AI scale in healthcare is now inseparable from data governance and security maturity. (privacy.org.nz)

Overall Assessment

As of 10 June 2026, AI in New Zealand healthcare has progressed from cautious experimentation to selective operationalisation with harder governance. The most mature deployments are still in documentation and workflow support; the most important new public development since April is the move of AI mammogram reading into procurement and validation for BreastScreen Aotearoa; and the strongest private-sector momentum is in integrated primary-care and aged-care documentation tools. (beehive.govt.nz)

But the sector remains early-stage overall. The Ministry of Health still says New Zealand is in the early stages of assessing how AI can be safely adopted, and Health NZ itself says many current uses are still being tested on a limited basis. The evidence therefore supports a clear conclusion: New Zealand’s healthcare AI model is maturing, but it is maturing as supervised augmentation rather than automation. The country is building the rules, procurement pathways, and trust architecture needed for broader adoption, while keeping clinicians visibly accountable for care. (health.govt.nz)

← Back to all articles View as Markdown