Top 10 AI Prompts and Use Cases and in the Healthcare Industry in New Zealand

New Zealand's healthcare system is fast moving from cautious pilot projects to practical change as the Government's new AI Strategy (released 8 July 2025) pushes adoption, governance and ethical use of ready‑made AI tools - the report even estimates AI could add about NZD76 billion to the economy by 2038 DLA Piper analysis of New Zealand's AI Strategy (July 2025).

Healthcare uptake is already notable - roughly 65% adoption in health and social assistance - and agencies such as Pharmac and Medsafe are exploring AI to speed medicine assessments while a new 24/7 virtual service connects patients to NZ‑registered clinicians, effectively putting “a night‑shift GP in the pocket” for rural and busy families OpenGovAsia report on New Zealand's AI-driven 24/7 digital healthcare services.

That practical shift makes workforce training urgent: short, work‑focused courses like the AI Essentials for Work bootcamp - practical AI training for clinical teams help clinical teams learn prompt design, safe workflows and productivity gains before scaling pilots.

“For many New Zealanders, pharmaceuticals are life or death, or the difference between a life of pain and suffering or living freely.” - Associate Health Minister David Seymour

The top ten prompts and use cases were chosen by applying New Zealand's practical, ethics‑first filters: projects had to map to the National Ethics Advisory Committee's standards on

health data and new technologies,

show a clear plan for data quality, bias mitigation, Māori engagement, and human oversight as described in NEAC's AI life‑cycle guidance (NEAC guidance: Health data and new technologies (New Zealand National Ethics Advisory Committee)); they also needed sector validation through stakeholder feedback and learning hubs such as the AI in Primary Care working group, which runs surveys, webinars and e‑learning to surface what's realistic for clinics and communities (GPNZ AI in Primary Care working group resources).

Each candidate use case completed an Algorithm Impact Assessment to identify likely harms, accountability, and monitoring needs before listing in the top ten (Algorithm impact assessment user guide (data.govt.nz)).

The result is a pragmatic shortlist that balances clinical value, measurable risk (think SaMD risk tiers), and the

so what? test -

would this actually reduce a patient's waiting time or diagnostic delay by a visible margin?

Intraoperative computer‑vision is moving from promising research into tangible theatre tools: evaluations of object‑recognition models like the YOLOv8 system highlight that

“accurate recognition of surgical instruments is essential”

for real‑time assistance (YOLOv8 evaluation for real-time recognition of robotic and laparoscopic surgical instruments), experimental work has shown automated detection and counting is feasible in practice (proof-of-concept study: automated surgical instrument detection and counting), and commercial solutions demonstrate how OT‑fitted cameras plus dashboards can flag missing tools and store timestamped video to reduce retained items and infection risk (commercial surgical instrument tracking with Vision AI).

For NZ operating teams, the

“so what”

is concrete: dependable instrument counts, instant alerts when a clamp or sponge goes astray, and a verifiable audit trail that helps catch human slips before they become harm - practical safeguards that combine validated models, theatre‑grade imaging and clinician oversight rather than replacing it.

Pre‑operative planning and route selection are prime candidates for PACS integration in Aotearoa: by fusing 3D CT/MR volumes with EHR data and clinical notes, radiology teams can build interactive, patient‑specific roadmaps that guide incision planning, vessel navigation and implant sizing long before a patient reaches the theatre; frameworks like NVIDIA‑backed MONAI Multimodal explicitly target this use case with a Radiology Agent that links DICOM imaging, structured records and VLM/LLM reasoning to speed interpretation and produce actionable pre‑op summaries (MONAI Multimodal Radiology Agent framework).

The academic literature underlines the same idea - multimodal integration systematically combines imaging with clinical and genomic data to improve disease management and decision support (JMIR 2025 review of multimodal integration in health care), while deep‑learning image‑fusion research shows how CT, PET and MRI can be combined into richer, clinically useful maps.

For New Zealand clinicians and the University of Auckland's radiology teams, the practical “so what” is tangible: a slice‑by‑slice, annotated 3D plan that reduces surprises in theatre, cuts avoidable repeat scans, and hands surgeons a verifiable, PACS‑linked audit trail for safer, faster procedures.

“By integrating diverse data streams through advanced multimodal models, we're not just improving diagnostic accuracy - we're fundamentally transforming how clinicians interact with patient data,” said Tim Deyer MD.

Autonomous and robot‑assisted surgery is leaping from lab demos to real clinical validation overseas - most recently a platform achieved the world's first clinical validation of autonomous surgery in a high‑profile press release - an important benchmark New Zealand teams should watch closely (Cornerstone Robotics autonomous surgery clinical validation press release).

For University of Auckland robotics labs, that global momentum means two practical priorities: build robust, competency‑based training and prove value safely in local pilots.

A 2025 systematic review of multi‑specialty robotic curricula shows exactly what that work looks like - didactic teaching, dry‑lab skills and VR simulation dominate curricula (present in roughly 70%, 70% and 67% of studies respectively), yet most programmes remain only partially validated and assessment methods vary widely (2025 systematic review of essential components and validation of robotic training (International Journal of Surgery)).

That evidence‑first roadmap fits NZ practice: small, PACS‑linked pilots that couple validated simulation training with staged clinical cases and local evaluation (see pilot implementation guide for New Zealand hospitals: using AI in healthcare (2025)) will help turn global breakthroughs into safe, credentialled tools for Aotearoa surgeons.

Waitlist triage and prioritisation are now frontline policy in Aotearoa: an “equity adjustor” algorithm - deployed in parts of Auckland since February and covering roughly 10% of the population - reweights ethnicity, time on the list, location and deprivation to push Māori and Pacific patients higher on elective surgery lists where clinical need allows, an explicit attempt to correct decades‑long outcome gaps (The Guardian report on New Zealand's equity adjustor prioritising Māori and Pacific elective surgery patients; ABC News coverage of the waitlist algorithm used in Auckland).

At the same time Health NZ is trying to clear backlogs fast - planning to outsource many low‑complexity cases to private hospitals and to lift throughput so a stated target of 95% seen within four months can replace the roughly 60% currently achieved - moves that clinicians warn could shift complexity rather than solve workforce pressures (RNZ coverage of Health NZ's plan to cut surgery waitlists by outsourcing to private hospitals).

New Zealand's history with prioritisation tools (see the 2018 general surgery prioritisation tool study) shows these systems can standardise access but also spark ethical and political debate; the “so what” is concrete - small point shifts in a scoring tool can meaningfully shorten a person's wait and change painful months into weeks, so governance, transparent criteria and careful local evaluation are essential.

“clinical need ‘always takes precedence and the equity adjustor doesn't interfere with that'.”

AI scribes are moving fast into New Zealand clinics as practical tools to shave hours off paperwork and restore eye contact - platforms like Heidi Health AI scribe platform, Sunoh.ai clinical transcription tool and enterprise speech solutions promise to transcribe visits, draft letters and even populate billing codes so clinicians "get home on time" and reclaim potentially hours a day; however a University of Otago survey of NZ primary care flags a mixed picture that matters locally: about 40% of respondents already use AI scribes, many report big time savings but others say editing and accuracy issues can erase gains, and te reo Māori, New Zealand accents, consent processes and Māori data sovereignty remain unresolved governance risks.

Health New Zealand has begun endorsing specific ambient tools with strict privacy, legal and training caveats, underscoring the practical “so what?” - scribes can reduce administrative load and improve patient rapport, but only when systems are locally validated, clinicians retain legal responsibility for notes, and clear consent, security and cultural‑data protections are in place (Sunoh.ai clinical transcription tool; University of Otago survey on AI-scribe use in New Zealand).

“Today someone said, 'I've got pain here', and pointed to the area, and so I said out loud 'oh, pain in the right upper quadrant?'”

Automated specialist reports and referral letters promise to shave admin time and speed patient journeys across Aotearoa by turning clinic audio and imaging insights into structured, EHR-ready summaries - think concise specialist letters, clear patient one‑pagers and prefilled orders.

Proven documentation tools in other markets already draft referral letters and embed orders into records (see how enterprise solutions like Dragon Copilot automate summaries and referrals - Review of Ophthalmology: Dragon Copilot automates clinical summaries and referrals), while multimodal copilots built for eye care show how imaging, notes and history can be fused to autosuggest clinical text and plan language for specialists and GPs (Yale Ventures multimodal AI copilot for eye doctors).

In New Zealand the payoff is practical: fewer delays from handoffs and clearer, audit‑friendly referral trails - provided systems are locally validated, EHR‑integrated and safe for te reo Māori, accent variation and Māori data sovereignty so clinicians retain legal responsibility for every note.

“AI isn't the future - it's now. From triaging urgent cases to detecting retinal disease, AI is changing how we deliver care.”

Anomaly‑detection models are emerging as a practical ally for retinal screening in Aotearoa, using unsupervised deep‑learning and knowledge‑distillation techniques to flag OCT scans with any pathomorphological changes rather than only the diseases the model was trained on; recent work demonstrates this approach on retinal OCTs and shows promise for catching unusual or novel presentations that traditional classifiers can miss (Anomaly Detection in Retinal OCT Images (Translational Vision Science & Technology, 2025)).

For New Zealand eye services and rural screening hubs the “so what” is immediate: an automated triage layer that highlights a single suspicious slice in a stack of images so a clinician can prioritise review, speed referrals, and cut costly repeat scans.

Real‑world studies and reviews also suggest these systems work even with limited labelled pathology - useful where curated local datasets are small - and can broaden screening beyond diabetic retinopathy to detect unexpected findings (AJMC article on AI-based anomaly detection for retinal disease screening), making population screening safer, faster and more efficient for stretched NZ ophthalmology teams.

“When abnormal (diseased) data, i.e., referable diabetic retinopathy in this study, were not available for training of retinal diagnostic systems wherein only nonreferable diabetic retinopathy was used for training, anomaly detection techniques were useful in identifying images with and without referable diabetic retinopathy,”

Workflow automation tied into practice management systems is where real day‑to‑day gains show up for New Zealand clinics: by piping intelligent document processing and LLM-powered extraction straight into an EHR/PMS like Incisive, scanned discharge packs, referral letters and lab PDFs can be turned into verified, machine‑readable records that prefill bookings, orders and billing codes - saving clinicians from repetitive copy‑paste and giving managers accurate, audit‑ready trails.

Tools such as Unstract's LLMWhisperer demonstrate how diverse formats (handwritten notes, complex tables, images) can be parsed, validated and output as structured JSON or ETL streams for downstream workflows (Unstract: data extraction in healthcare), while published methods for combining structured and unstructured EMR data show how those outputs can feed cohort creation, decision support and reporting without losing clinical nuance (BMC study on combining structured and unstructured EMRs).

The “so what” is practical: fewer delayed referrals, fewer repeat tests, and a single verifiable record that travels with the patient across primary, specialist and hospital systems - making everyday care smoother and safer across Aotearoa.

Smart capacity planning and provider dashboards turn raw appointment chaos into clear, usable signals for New Zealand health services: platforms that combine real‑time patient routing, predictive analytics and configurable scheduling rules can match demand to the right setting, shift low‑acuity visits to virtual care, and keep ORs filled rather than idle.

International case studies and vendor suites demonstrate the mechanics - Epic capacity optimization tools for hospital flow and OR scheduling, Clearstep Capacity Optimization Suite showing AI-driven workflows to balance provider workloads, and front‑door orchestration from DexCare capacity optimization and front-door orchestration case study illustrate how smarter search and routing can shave days off waits (case studies report a 5‑day reduction in time to appointment and note the potential when just 5% of low‑acuity visits move to virtual care).

For Aotearoa the so what is tangible: fewer empty clinic slots, shorter patient queues and dashboards that show when to flex staff or open nearby capacity - turning guesswork into provable gains for clinicians, managers and patients alike.

Data governance is becoming the backbone of safe AI in Aotearoa: regulators and clinicians now expect provable chains of custody, clear privacy controls and cross‑border safeguards before models support clinical decisions.

New Zealand guidance stresses data minimisation, redaction and human‑in‑the‑loop checks under the Privacy Act and IPPs, so every automation needs a scoped purpose, short retention and searchable audit logs as standard - practical guardrails well explained in the OPC‑aligned AI safety playbook for NZ AI safety for New Zealand businesses - privacy, data handling and practical guardrails.

At the same time, traceability demands an immutable, tamper‑proof audit trail: industry pieces argue that every input, model version and decision should be linkable and reconstructible so audits aren't guesswork but verifiable evidence (What is an AI audit trail and why it is crucial for governance).

Practical tech options range from append‑only cryptographic logs and HSM‑protected keys to enterprise blockchain proposals for a verifiable chain‑of‑custody, backed by Zero‑Trust deployment and continuous model monitoring.

The “so what” is immediate for clinicians and managers: when a triage call or automated report goes to the wrong place, an immutable audit trail shows who, what and when - like a sealed, timestamped chain that visibly breaks if anyone tampers with it - so governance moves from promise to proof.

The path forward for Aotearoa is practical and staged: start small, prove value, and protect people - deploy focused pilots that use clinical decision‑support to cut repeat tests and streamline pathways, then scale only after local validation and iwi engagement; a handy starting point is the step‑by‑step pilot implementation guide for NZ hospitals.

Pair those pilots with short, work‑focused training so clinicians and managers know safe prompt design, consent rules and when human oversight must intervene - programmes such as the AI Essentials for Work bootcamp teach those practical skills and help staff move from curiosity to confident use.

Protect trust by embedding clear organisational policies on data handling, model transparency and staff use - see the practical guidance in Workday's guide for NZ healthcare leaders on harnessing the power of AI.

Finally, plan for workforce transition now: retrain documentation teams into QA and annotator roles, measure outcomes that matter to patients, and treat every deployment as a governance exercise - small, audited wins will convert sceptics faster than sweeping, untested change.