The Complete Guide to Using AI in the Healthcare Industry in New Zealand in 2025

Introduction: AI is already reshaping healthcare in Aotearoa in 2025 - national surveys report AI adoption surging across sectors and about 65% uptake in health and social assistance, with ambient documentation tools and radiology/pathology pilots (eg, Volpara) trimming admin time and boosting throughput, not just replacing roles; New Zealand's 2025 productivity report documents 82% overall AI use and 93% of businesses seeing efficiency gains, while the new national AI Strategy frames this as a chance to “invest with confidence” and aim for responsible adoption that could add NZD76 billion by 2038.

For health leaders, the priorities are clear: pair clinical pilots with strong governance and upskilling - practical training such as Nucamp's 15‑week AI Essentials for Work bootcamp (AI Essentials for Work bootcamp syllabus (15‑week)) helps clinicians and managers learn promptcraft and tool use to safely embed AI at scale.

prepared with the assistance of AI as a demonstration of the Government “walking the talk” while maintaining appropriate oversight and safeguards for sensitive information.

New Zealand's July 2025 national AI Strategy, framed as “Investing with confidence,” gives health leaders the policy map needed to adopt AI without waiting for bespoke regulation: it champions a light‑touch, principles‑based approach aligned with the OECD AI Principles and is accompanied by practical Responsible AI Guidance for Businesses to help organisations manage ethics, data and governance as they scale tools in clinical settings; the Public Service AI Framework complements this by setting expectations for safe, transparent GenAI use across government.

The Strategy explicitly prioritises adoption over building foundational models - a pragmatic play to capture productivity gains (the government cites a potential NZ$76 billion uplift by 2038) while tackling barriers like regulatory uncertainty and a clear skills gap.

Uniquely Aotearoa, the approach embeds Treaty of Waitangi considerations and leans on existing laws (privacy, consumer protection, human rights) rather than inventing an entirely new regime, giving DHBs and clinics a predictable compliance path.

For practical next steps, clinicians should read MBIE's strategy and the accompanying guidance to align procurement and governance, and study the Government Chief Digital Officer's Public Service AI Framework for examples of testing, accountability and workforce readiness that can be mirrored in hospitals and primary care.

“AI could add $76 billion to our GDP by 2038, but we're falling behind other small, advanced economies on AI-readiness and many businesses are still not planning for the technology,” says Dr Reti.

AI scribes are already delivering tangible, practical wins for New Zealand clinicians: surveys and trials show roughly 40% of primary care providers have tried ambient transcription tools and many report reduced multitasking, faster record‑keeping and a real boost to patient engagement - users estimate time savings of about 30 minutes to two hours per day and note clearer eye contact and improved rapport when documentation is handled by a scribe.

These tools also help standardise notes and free up capacity for more patients or longer, higher‑value consultations, making throughput and referral workflows easier to manage; see the University of Otago coverage of emerging ethical questions and uptake in Aotearoa and the full exploratory survey published via CSIRO for the detailed findings.

Well‑designed rollouts (for example, vendor guides like Heidi's implementation playbook) emphasise clinician review of AI notes, chosen KPIs and staged adoption so the efficiency gains aren't lost to editing overhead or workflow mismatch - practical benefits are greatest where governance, consent and local language training (including te reo Māori and NZ accents) are built into the plan.

Risks and ethics in Aotearoa's AI-driven health services come into sharp focus around privacy, data sovereignty, bias, transparency and accountability - practical concerns flagged by Health New Zealand's guidance on generative AI and large language models and by legal experts analysing AI privacy in Aotearoa (Health New Zealand generative AI guidance for health services, Clyde & Co analysis of AI privacy risks in Aotearoa).

Clinically, the danger is not just a data breach but subtle harms such as biased predictions that worsen inequities, or hallucinated outputs - a real‑world caution where AI-generated legal citations were later shown to be fabricated - so accuracy checks and human review are non‑negotiable.

The National Ethical Advisory Committee's standards map a lifecycle approach: assess data quality and identifiability, evaluate risk using a SaMD‑style matrix, build explainability and monitoring into deployment, and embed Māori Data Sovereignty and Treaty obligations into governance (NEAC guidance on health data and new technologies).

For healthcare leaders the practical takeaway is clear: adopt tool‑specific auditing, explicit consent and storage rules, assign clear accountability for each AI lifecycle step, and require human‑in‑the‑loop checkpoints so technology augments clinical judgment rather than replaces it.

“To err is human, but to really foul things up you need a computer.”

Data management, privacy and security are the non‑negotiable foundations for any AI rollout in Aotearoa's health system: the Health Information Privacy Code 2020 makes clear that identifiable health information collected, used, held or disclosed by health agencies must be treated under bespoke health rules (see the New Zealand Health Information Privacy Code 2020 (official PDF)), while the Privacy Act 2020's Information Privacy Principles set the baseline expectations for purpose, access, retention and overseas transfers that every clinic and hospital must follow (see the Privacy Act 2020 (New Zealand) overview and Information Privacy Principles).

Practical steps that map directly to those laws include appointing a named Privacy Officer and running mandatory training (Health 101 / Privacy 101 and role‑based Health/Privacy ABC modules used in general practice), building role‑based access controls into the patient management system, and enforcing short inactive‑session timeouts (RNZCGP clinical privacy and security guidance recommends automatic lockouts of no more than 15 minutes) so a clinician's notes aren't exposed by an unlocked terminal.

Backups must be encrypted, stored offsite or on secure services, and regularly tested for restoreability; video and telehealth platforms should be chosen only after a Privacy Impact Assessment guidance (NZ) and Cloud Risk Assessment and documented in practice privacy policies.

Rule 12 of the Code and IPP 12 of the Act demand explicit safeguards or patient authorisation before sending personal health data offshore, and any serious breach requires prompt notification to patients and the New Zealand Office of the Privacy Commissioner; embedding these legal checkpoints into procurement, audit and vendor contracts turns compliance from a box‑tick into operational resilience.

Choosing AI tools for New Zealand clinics and hospitals starts with a clear problem statement and a pragmatic checklist: what clinical job must the tool do, what data and systems it must integrate with, who owns the data, and what the compliance and scaling needs are.

For low‑risk admin tasks like appointment scheduling, triage chatbots or scribes can be piloted quickly with off‑the‑shelf products that deliver fast wins and low upfront cost, while for diagnostics, treatment planning or anything touching sensitive patient records Appinventiv's analysis shows custom or hybrid builds often make more sense because they deliver ownership, tailored governance and long‑term scalability.

Run a staged MVP or hybrid rollout (pilot with a vendor, parallel a custom roadmap), cost it over a 3‑year horizon, and budget for data‑cleaning, retraining and user training up front.

New Zealand evidence is already compelling: specialist co‑pilot tools are easing workflow pressure and, when scaled, can add meaningful capacity - one provider reports the tech lets clinicians see two to three extra patients a day and has supported hundreds of thousands of consultations - so pick tools that match both clinical risk and local realities, and pair them with workforce upskilling (see Appinventiv's guide on custom vs off‑the‑shelf choices and local NZ pilots reported in ITBrief).

For clinics wanting immediate tactical wins plus future resilience, start small, insist on auditability and NZ‑friendly data contracts, and keep a clear path to a custom or hybrid system if compliance or accuracy demands grow.

Appinventiv: Custom vs Off‑the‑Shelf AI analysis, ITBrief: NZ specialist scribe pilots and productivity reporting, Nucamp: Capacity planning and productivity analytics (AI Essentials for Work syllabus).

“What this technology really does is free up my human intelligence for where it matters most: my clinical judgment, my focus on the patient in front of me, and complex decision‑making.”

Start with a tightly scoped pilot that matches clinical need and risk: define the job-to-be-done (for example, speeding coded discharge summaries), identify data owners and a named Privacy Officer, and run a Privacy Impact Assessment before any data leaves a local system; Te Whatu Ora's AI‑assisted clinical coding pilot is a practical template - using previously coded, anonymised records in a 2–3 month test to assess speed and accuracy and to create a single shared worklist across sites - and can be reviewed for procurement and technical design guidance (Te Whatu Ora AI-assisted clinical coding pilot overview).

Next, stage the rollout: start with a parallel run (AI output alongside human coders), measure time‑savings, error rates and edit burden, then iterate models and integration points; document consent and audit trails and align local policies with emerging professional guidance and primary‑care resources (practical templates on consent, PIAs and four‑point principles are collected in the Pinnacle Practices guidance for general practice) (Pinnacle Practices AI tools in general practice guidance and templates).

Finish by embedding results into procurement criteria (ICD‑10‑AM certification where relevant), staffing plans (new QA and annotator roles), and a clear escalation path so clinicians retain final accountability - imagine replacing a desk-high stack of paper charts with a single digital worklist that routes cases nationwide, but only after human review and documented safeguards are proven.

“The test is expected to use an anonymised set of clinical records which have previously coded by Te Whatu Ora, the ROI says. ‘We are interested in these records being used to demonstrate the collation of source data, processing, coding and grouping of clinical coding data as well as reporting on the outcomes of this process.'”

Keeping Aotearoa's health workforce ready for AI means treating training and change management as core clinical infrastructure, not optional CPD: a KPMG-informed analysis shows 76% of New Zealand workers have had no formal or informal AI training and only 41% use AI at work, while roughly 60% say they lack confidence - a skills gap that will blunt the tech's promise unless addressed with role‑specific learning and career pathways (see the KPMG overview of New Zealand's AI skills gap).

Practical healthcare upskilling is about more than one‑off courses: embed staged, hands‑on modules that mirror day‑to‑day tasks (ambient‑scribe review, safe promptcraft, audit logging), pair clinicians with supervised sandboxes and mentoring, and create new roles such as Clinical AI annotator and documentation QA specialist to put human judgement into automated pipelines (explore suggested roles and pathways).

Evidence from workforce reports shows employers back this approach - most NZ firms now support upskilling and see AI creating opportunities - so hospitals and primary care networks should budget for ongoing training, protected practice time and clear success metrics; the payoff can be dramatic, with sensible rollouts reclaiming 30 minutes to two hours per clinician each day and shifting attention back to patients, not paperwork (see practical workforce insights from the AI Essentials for Work syllabus).\n\n \n \n \n \n \n \n \n \n \n \n

“The time has come for New Zealand to get moving on AI,”

New Zealand is already collecting practical wins: Te Whatu Ora's pilot of Tuhi, developed with Māori-owned Awa Digital, is testing ambient audio-to-note workflows with twenty clinicians across the motu to ease cognitive load and speed documentation (HiNZ: Health NZ pilots Tuhi), while international vendor evidence shows what scaled adoption can look like - Heidi's implementation guide reports activation rates of 60–80% (well above typical 20–40%) and a Modality Partnership trial using Heidi delivered headline results including a 51% drop in documentation time during appointments, a 61% fall in after‑hours admin and 78% of clinicians reporting better rapport with patients (Heidi: AI medical scribe adoption).

Those numbers translate into a vivid, practical image for NZ clinics: clinicians trading a looming pile of after‑hours notes for a single reviewed, ready-to-sign summary at the end of clinic.

Early endorsements matter too - Health NZ has effectively signalled support by endorsing two AI scribes in local reporting, giving providers a clearer procurement path (eHealth News: two scribes endorsed).

Conclusion: New Zealand's new AI Strategy and Public Service AI Framework have turned a policy question into a practical mandate - adopt responsibly, test quickly, and invest in people - because the upside is real (the Strategy cites a potential NZ$76 billion boost by 2038) and the risk is manageable when governance is front‑loaded.

For health leaders this means three clear next steps: run tightly scoped pilots that keep clinicians in the loop and map to a risk‑based approach; hard‑wire privacy, Māori data sovereignty and auditability into procurement and vendor contracts; and treat upskilling as infrastructure by training clinical teams in promptcraft, human‑in‑the‑loop review and capacity planning.

Public‑sector leadership and the Strategy's light‑touch, OECD‑aligned principles give hospitals a steady compliance compass - read the government launch for the official framing and timelines (New Zealand AI Strategy official launch (Beehive)) - and practical courses like Nucamp's 15‑week AI Essentials for Work bootcamp can turn staff anxiety into everyday skills (AI Essentials for Work bootcamp syllabus (15 weeks)).

Start small, measure time‑savings and error rates, then scale the proven pilots: the smartest path to capture productivity gains is steady governance, visible clinician oversight and training that makes AI a tool for better care, not a leap in the dark.

“AI could add $76 billion to our GDP by 2038, but we're falling behind other small, advanced economies on AI-readiness and many businesses are still not planning for the technology,” says Dr Reti.