AI Adoption in Healthcare in New Zealand: A Living Whitepaper
Updated: 2 April 2026
Executive Summary
AI in New Zealand healthcare is now best described as selective operational deployment with tightening governance. The clearest live deployment remains ambient documentation in public emergency departments, where AI scribe tools are now available nationwide. Beyond that, the strongest activity is in screening and imaging, workflow support, and mental health navigation, while most other use cases remain in pilot, exploratory procurement, or funded implementation research rather than routine national service delivery. The Ministry of Health still says New Zealand is in the early stages of assessing how AI can be safely adopted, and that remains an accurate system-level description. (beehive.govt.nz)
The main shift since the previous edition on 9 March 2026 is not another large-scale rollout, but a stronger rules-and-oversight layer. On 10 March 2026, the Medical Council of New Zealand published final guidance on using AI in patient care, formalising expectations around clinician accountability, transparency, consent, record-keeping, privacy, and bias. Health NZ’s public privacy statement and advisory-group framework also reinforce that AI is being positioned as a clinician-support tool, not an autonomous decision-maker. (mcnz.org.nz)
What Has Changed Since 9 March 2026
- Professional guidance is now final, not draft. The Medical Council’s guidance, published on 10 March 2026, says doctors remain responsible for all clinical decisions, must check AI output, must tell patients when AI is used in their care, and must obtain informed consent in specific situations including AI consultation recording and some high-impact diagnostic or treatment uses. It also states doctors must not use AI to impersonate themselves in practice through avatars or chatbots. (mcnz.org.nz)
- The governance gap between fast adoption and formal rules has narrowed. That matters because the best available NZ primary care survey showed AI scribe uptake had already become real-world practice before formal guidance fully arrived: among 197 respondents, 40% had experience with AI scribes, nearly half had not sought consent, and only 66% had read the tool’s terms and conditions. (pubmed.ncbi.nlm.nih.gov)
- Health NZ’s public AI posture is now more explicit. Its privacy statement says AI is used for transcription, consultation summarisation, document generation, staff knowledge tools, and preliminary review of images and scans; it also says many of these uses are still in pilot phase, that AI-generated content affecting care is clinician-reviewed, and that AI is not used to make automated decisions about patients’ healthcare. (info.health.nz)
- Publicly visible change since early March has been governance hardening more than fresh scale deployment. The most important deployment news remains the 28 February 2026 nationwide ED rollout, while the most important post-9-March development is the formalisation of how clinicians are expected to use AI safely. (beehive.govt.nz)
Current State of AI Adoption
1. National frontline deployment: emergency department scribes
The most mature public-sector use of AI in New Zealand healthcare is still the nationwide emergency department scribe rollout. On 28 February 2026, the Government said AI scribe technology was live in all emergency departments, reaching 1,250 ED doctors and frontline staff. The Government also said doctors in the pilot were able to see one additional patient per shift on average, and that Health NZ was progressing approval of more than 1,000 additional licences, predominantly for mental health teams. (beehive.govt.nz)
This deployment is strategically important because it shows where New Zealand is most comfortable using AI at scale: documentation-heavy, clinician-supervised, workflow-embedded augmentation rather than autonomous diagnosis or treatment. It also creates an operational template for expansion into adjacent services under the same governance model. (beehive.govt.nz)
2. Primary care: rapid uptake, now becoming more integrated
Primary care remains the fastest-moving non-hospital adoption environment. The 2025 exploratory survey of NZ primary care providers found meaningful early uptake of AI scribes, with reported benefits around time saving, reduced multitasking, and better patient rapport, but also significant concerns about privacy, legal compliance, data leaving New Zealand, and documentation errors. (pubmed.ncbi.nlm.nih.gov)
The commercial layer is also maturing. Medtech Global says Medtech AI, launched for New Zealand from 23 February 2026, is directly integrated with Medtech Evolution rather than functioning as a standalone transcription app. Medtech says clinicians stay in control of all outputs, no audio is retained, patient-identifiable fields such as names, dates of birth, and NHI numbers are not stored in the system, data is processed and stored locally in New Zealand, and clinicians must confirm explicit patient consent before recording begins. (medtechglobal.com)
At the sector level, primary care is also building its own governance and literacy infrastructure. General Practice New Zealand’s AI in primary care working group now points to repeated sector surveys, implementation guidance, AI e-learning, and webinars on literacy, live scribing, ethics, and workforce readiness. That is a sign the market is moving beyond experimentation toward operational norms and training. (gpnz.org.nz)
3. Screening and imaging: the strongest next-wave adoption lane
The clearest next expansion area is still screening and imaging. On 12 February 2026, the Government said Health NZ was taking the first formal step toward potential AI support in BreastScreen Aotearoa, inviting organisations with AI image-reading experience to show how the technology could be used safely and effectively. The move was framed as exploratory validation rather than deployment, but it directly linked AI to workforce pressure and service sustainability. (beehive.govt.nz)
The diabetes retinal screening pilot in Māngere remains one of the country’s most concrete clinical AI use cases beyond documentation. The programme uses AI for primary grading of retinal images captured in community settings, with the goal of increasing access, delivering real-time results, and reducing delays for people who are overdue for screening. The Government said more than 26,000 people in South Auckland had missed recommended screening in the previous two years. (beehive.govt.nz)
That pilot is reinforced by earlier New Zealand research rather than appearing out of nowhere. NZ-led studies on the THEIA system found high sensitivity for detecting referable diabetic retinopathy in local multi-ethnic datasets, and a prospective multicentre evaluation reported that it did not miss referable disease in the study setting. That makes diabetic eye screening one of New Zealand’s clearest examples of a research-to-implementation AI pathway. (pubmed.ncbi.nlm.nih.gov)
4. Mental health and navigation
Mental health is emerging as a second major adoption lane. In November 2025, the Government funded Whakarongorau to develop a mental health AI navigation platform intended to act as a digital front door, helping people identify available support, understand options in their area, and in some cases book directly. The stated use case is navigation and triage, not diagnosis. (beehive.govt.nz)
The ED scribe programme also points in the same direction operationally: Health NZ’s next tranche of additional licences is expected to be used mainly by mental health teams, suggesting documentation support will likely arrive in routine mental health workflows before more ambitious clinical AI tools do. (beehive.govt.nz)
Research and Innovation Pipeline
New Zealand’s research base is becoming more substantial and more implementation-oriented. The Health Research Council has confirmed investment in 10 AI-in-healthcare studies worth $4.6 million, and framed them as work intended to support timely access to quality healthcare and meaningful changes in practice and policy. (hrc.govt.nz)
The funded portfolio is notable because it is not just exploratory computer science. It is tied to real service problems:
- Acute stroke: a Health NZ-led study will measure whether AI-assisted rapid brain-scan interpretation improves intervention rates and reduces avoidable treatment delays across NZ hospitals; the project’s lay summary says the expected upside, if successful, includes more patients accessing time-critical treatment and lower health-system costs. (hrc.govt.nz)
- Digital pathology for gastrointestinal cancers: University of Otago researchers are developing AI tools to analyse biopsy images and other patient information to support more precise treatment decisions and avoid unnecessary surgery. (hrc.govt.nz)
- Postoperative monitoring: University of Auckland researchers are using digital tools and AI to detect deterioration earlier after surgery, with explicit attention to privacy, equity, and health data sovereignty. (hrc.govt.nz)
- Youth mental healthcare ethics: AUT-led research is focused on co-designing practical guidance for safe, fair, and culturally responsive use of AI in youth mental healthcare. (hrc.govt.nz)
The healthy-ageing pipeline is also deepening through the NZ-Singapore research programme. Current projects include AI-assisted interRAI assessment for aged care and a dementia risk prediction tool being developed from New Zealand and Singapore longitudinal datasets, with an explicit emphasis on explainability and clinician usability. (otago.ac.nz)
Research Signals on Trust, Equity, and Social Licence
Recent NZ research continues to show that adoption will depend as much on trust as on technical performance. A February 2026 New Zealand Medical Journal viewpoint, drawing on local interview studies, found patients were more comfortable with healthcare AI when data use was for public benefit, governance was strong, data protection was clear, consent or choice was available, clinicians remained visibly responsible, and Māori representation was built into development and governance. (nzmj.org.nz)
That emphasis on trust is mirrored in newer condition-specific work. A March 2026 study on Pasifika perspectives on AI for asthma management in New Zealand found interest in AI’s potential, but also highlighted concerns around accuracy, privacy, access, and age-related digital divides. In other words, enthusiasm exists, but only alongside strong expectations about inclusion and safety. (sciencedirect.com)
Governance, Regulation, and Digital Foundations
Health-system governance
Health NZ’s National Artificial Intelligence and Algorithm Expert Advisory Group remains the centrepiece of system governance. It oversees AI initiatives using Health NZ data, requires that projects align with ethical, technical, clinical, and operational standards, and says all AI development or implementation plans must be registered with the group. Its terms also explicitly include equity, bias, discrimination, Māori health, and Māori data sovereignty, and say a national register of AI in use within Health NZ will be made publicly available. (info.health.nz)
Privacy and clinician responsibility
Health NZ’s public privacy statement makes the operating model unusually clear: AI is used inside a closed environment, patient information is not to be used to build commercial AI or third-party large language models, short-term disclosures to providers are deleted shortly afterwards, and any AI output that affects patient records or clinical decisions must be reviewed by the responsible clinician. Health NZ also states it does not use AI to make automated decisions about a patient’s care. (info.health.nz)
The Medical Council’s new guidance adds the profession-level counterpart to that system stance. It says doctors must understand the intended use and limitations of the AI they use, remain accountable for decisions, document AI use when it influences care, and obtain informed consent in situations such as AI transcription of consultations, identifiable data sharing outside the primary medical record, or clinically significant AI involvement in diagnosis, treatment, or delivery of care. (mcnz.org.nz)
Infrastructure still limits scale
New Zealand’s AI ambitions in healthcare still depend on unfinished digital modernisation. In November 2025, the Government said 65% of hospitals still used paper-based notes and Health NZ was managing more than 6,000 digital systems. The Health Digital Investment Plan responds with long-horizon investment in a national EMR, remote monitoring, national radiology infrastructure, stronger cybersecurity, and virtual-hospital capabilities, while HealthX was set up to accelerate AI and digital tools into frontline settings. (beehive.govt.nz)
This matters because NZ’s adoption pattern is not just about willingness to use AI; it is also about whether the underlying data, interoperability, device, and records environment is mature enough to support safe scaling. In practice, the digital foundation remains a constraint as much as a catalyst. (beehive.govt.nz)
Key Patterns and Trends
1. Admin-first adoption is now established
The most scalable and accepted use case remains ambient documentation and workflow support. That is where New Zealand has moved from pilot to nationwide use, and it is also where primary care vendors and clinicians are moving fastest. (beehive.govt.nz)
2. Imaging is the most likely next clinical scale-up
Breast screening, diabetic retinal screening, stroke imaging, and digital pathology all point in the same direction: AI is most likely to expand next where there are high-volume interpretation tasks, specialist shortages, and measurable outcomes. (beehive.govt.nz)
3. Governance is catching up quickly
The combination of the Medical Council’s final guidance, Health NZ’s privacy rules, and the advisory-group process means AI is increasingly being pulled into formal clinical governance, not treated as a casual software add-on. (mcnz.org.nz)
4. Equity and data sovereignty are design requirements, not side issues
Across Health NZ governance, Ministry principles, HRC projects, and recent trust research, Māori health, bias, fairness, and data sovereignty consistently appear as core adoption conditions. (info.health.nz)
5. New Zealand’s model is guarded augmentation
The public-sector model remains clear: AI can support clinicians, improve throughput, and reduce repetitive work, but it is not being positioned to replace clinical judgement or to make automated care decisions. (beehive.govt.nz)
Overall Assessment
As of 2 April 2026, New Zealand healthcare has moved beyond AI curiosity and into a phase of targeted operationalisation. The evidence is strongest in nationwide ED scribing, fast-moving primary care documentation tools, a credible screening-and-imaging pipeline, and a funded research portfolio that is unusually focused on implementation rather than hype. (beehive.govt.nz)
But the sector is still early-stage overall. Most clinical AI beyond documentation remains in validation, pilot, or research mode; Health NZ itself says many AI uses are still being tested on a limited basis; and the Ministry continues to characterise the country as being in the early stages of safe adoption. The defining feature of the last few weeks is that professional and system governance have strengthened materially, especially through the Medical Council’s guidance. That makes the current NZ model clearer than before: AI first as supervised augmentation, with privacy controls, clinician accountability, and social licence treated as prerequisites for scale. (info.health.nz)