FREE GUIDE FOR NZ HEALTH & COMMUNITY LEADERS

AI in healthcare

A practical guide to AI adoption for health leaders in New Zealand.

Real-world New Zealand case studies
How chatbots, scribes, and operations tools are saving time
Risks and governance tips
Six lessons from early adopters
Five key questions to ask before using AI

FOREWORD

Ray Delany, CEO, CIO Studio

Artificial intelligence (AI) is rapidly becoming one of the most talked-about developments in modern technology. As with any significant innovation, it can be difficult to separate the real opportunities from the noise. This is especially true in healthcare, where the potential is great, but so too are the risks and uncertainties.

AI promises much. From supporting clinical decision-making and improving patient outcomes to streamlining administration and enhancing operational efficiency, the possibilities are far-reaching. However, alongside these benefits come valid concerns around ethics, data privacy, bias, and the overall readiness of organisations to implement these tools effectively.

This guide is intended to provide a practical and balanced perspective. Our focus is not on abstract theory or future speculation, but on what can be done today with the resources available to medium-sized healthcare providers. We have aimed to give clear, well-researched information that will help you assess the value of AI for your organisation and make informed decisions.

Like many disruptive technologies before it, AI is currently going through a period of intense interest, often accompanied by unrealistic expectations. But as history has shown, the most lasting impact comes from thoughtful, steady adoption built on sound principles.

While the tools will continue to evolve, the fundamental considerations around governance, ethics, and purpose are more stable.

In this publication, we offer a starting point for healthcare organisations looking to explore AI in a meaningful way. Whether you are already experimenting with AI solutions or simply beginning to consider their role, we hope this guide gives you both clarity and confidence.

Our goal is to help you take practical, outcome-focused steps to integrate AI into your organisations, rooted in the realities of your environment and aligned with what matters most. We hope you find it useful.

Real-world value, not science fiction

Artificial intelligence in healthcare has often been framed as futuristic - a technology of tomorrow. But in medium-sized health organisations across New Zealand and Australia, AI is already here, already working, and already improving outcomes. Not in labs, not in billion-dollar systems, but in clinics, hospitals and primary health organisations that make up the backbone of our healthcare system.

Regional momentum is strong, and growing

According to recent surveys, 45% of healthcare SMEs in Australia are already using AI tools - the highest rate across any industry. In New Zealand, 82% of all small-to-medium businesses (including healthcare) report experimenting with AI, significantly above the global average of around 75%.

93% of NZ businesses using AI say it's improving staff efficiency

95% report revenue benefits

Local AI use is driven by real operational needs, not by curiosity or trend-chasing

Why the health sector leads

Healthcare organisations in Australasia — particularly mid-sized ones — have strong incentives to adopt AI:

Workforce shortages: Clinicians and admin teams are stretched thin. AI that reduces load is highly valued.
Growing patient demand: As populations age, so does the burden on systems. Efficiency becomes critical.
Low margins and high compliance: There's limited room to waste time or money. AI helps streamline both.

Medium-sized health organisations are often large enough to feel the pain of inefficiencies, but small enough to implement targeted AI solutions thoughtfully and effectively. This makes them well-positioned to benefit from AI tools, especially those designed for specific workflows.

Tangible results from the region

Unlike many global deployments, AI in New Zealand and Australia is being implemented in tightly scoped, high-impact areas. These aren't abstract innovations. They're working solutions, in place today, producing clinical and operational gains.

Queensland's Patient Admission Prediction Tool (PAPT)

Forecasts emergency department demand with 90% accuracy, allowing better staffing and shorter wait times.

In rural NSW

An AI model flagged 70% of high-risk patients likely to be readmitted within 28 days, giving doctors a chance to intervene earlier.

Volpara Health

A Wellington-based company, built an AI tool to measure breast density in mammograms - now used in 40% of US breast screenings.

Global comparisons: Enthusiasm with mixed results

Globally, healthcare's use of AI is also surging - but the gap between interest and implementation remains wide.

This cautious approach is not without merit. AI tools need strong governance, especially in healthcare. But it also means that when local providers do adopt AI, they do so with a clearer focus on real outcomes, and with frameworks in place to monitor performance.

What sets AI adoption in New Zealand and Australia apart is its orientation toward the practical. Instead of sweeping transformation projects, we see a "start where it helps most" approach. Triage, admin support, diagnostic assistance - each tool targets a clearly defined problem.

The realism is paying off. Across both countries, AI is already:

Reducing patient wait times
Improving diagnostic accuracy
Automating routine tasks and reducing human error
Helping staff focus on higher-value work

In New Zealand and Australia, the AI conversation has shifted from 'what if' to 'how do we implement this responsibly.' Medium-sized health providers are leading the way - not by making big bets, but by solving real problems with well-scoped AI tools. The lesson is simple: this technology is no longer theoretical. It’s practical, proven, and delivering value now.

In one major international survey, 85% of health systems reported using AI in some form. But few have integrated it meaningfully into frontline decision-making.

Australian hospitals remain cautious, with very few deploying AI outside of imaging as of early 2024.
Many US hospitals use AI to predict conditions like sepsis or stroke - yet even there, uptake is patchy and trust is still being built.

How AI is being used today

AI isn’t being implemented evenly across the health sector. Some organisations are strategically advancing with cutting-edge diagnostic tools, while others are quietly embedding assistive bots into patient services. But when we zoom in on medium-sized providers - the PHOs, clinics, and regional hospitals - the pattern becomes clearer.

Together, these tools are helping organisations increase efficiency, reduce errors, and improve both patient experience and staff productivity. But their roles, risks and benefits differ - and understanding the difference matters.

Assistive chatbots and virtual agents that handle patient interactions and admin tasks.

AI-powered medical scribes that automate clinical documentation during consultations.

Operational optimisation tools that streamline supply chains, staffing and scheduling.

Autonomous AI agents that support clinical decision-making and diagnostic accuracy.

1. Assistive AI: Chatbots and virtual agents

For most organisations, the first step into AI comes via conversational tools. Chatbots - text or voice-based - are now widely used across Australasia to manage tasks such as:

24/7 symptom checking
Booking appointments
Providing answers to common patient questions
Medication reminders and health education

In New Zealand, organisations are trialling virtual agents - AI-powered avatars that speak with patients and whānau in plain language.

These bots can be especially useful for individuals with low health literacy or those seeking privacy for sensitive questions.

Early results show real benefits:

Staff workload is reduced
Patients report greater satisfaction and ease of access
Access to support outside normal hours is significantly improved

Importantly, these tools aren’t replacing clinicians. They’re augmenting the service layer - acting as always-available support staff that answer common questions and escalate more complex ones.

2. Medical scribes: The quiet productivity revolution

One of the most practically transformative AI tools in New Zealand healthcare is AI-powered medical scribes. These systems listen to consultations and automatically generate clinical notes, freeing clinicians from documentation - one of their biggest time drains.

Australian emergency departments report reduced documentation burden and improved efficiency, while New Zealand GPs are adopting scribes to see more patients without extending hours. Uptake is strongest in primary care where documentation load is heaviest.

The appeal is straightforward:

Doctors spend 30-40% of their time on documentation - AI scribes cut this dramatically
More face-to-face time with patients during consultations
Reduced after-hours admin work
Improved accuracy of clinical records

One Wellington practice freed up 90 minutes per day per GP using AI scribes. These systems work seamlessly in the background during consultations - exactly the kind of targeted solution medium-sized health organisations need.

3. Operational AI: Scheduling and supply chain tools

Operational AI may be the most overlooked - but arguably the most effective category for medium-sized organisations.

These tools are working behind the scenes to:

Automate medical supply orders based on demand forecasts
Predict surgery backlogs and optimise theatre utilisation
Match staff rosters with forecasted patient volumes

In one Australian example, an AI tool successfully automated 70% of consumable stock ordering, with built-in buffers to prevent shortages. Another health provider used AI to schedule surgeries more efficiently, reducing idle theatre time and cutting patient waitlists.

These systems are especially valuable for medium-sized clinics and hospitals that lack the manpower for manual optimisation. By automating these repeatable decisions, they free up staff to focus on higher-impact work.

However, uptake is cautious. Many Australasian health leaders remain hesitant to deploy fully autonomous systems without extensive evaluation.

Concerns include:

Safety and reliability
Explainability of decision-making
The ethical burden of misdiagnosis or error

To address this, New Zealand health authorities are building evaluation frameworks to ensure any clinical AI tools are deployed safely and effectively.

AI in healthcare isn’t a single solution - it’s a suite of tools, each solving a different class of problem. For most medium-sized health organisations, the current approach is not to leap straight into diagnostic AI, but to layer in AI gradually, building trust and capability along the way.

The most advanced use of AI is in clinical decision-making.

These tools include:

Diagnostic imaging AI (e.g. Volpara Health's breast density measurement tool)
Predictive risk models for patient deterioration, sepsis, or cardiac arrest
Reasoning engines that analyse symptoms and suggest differential diagnoses

Used properly, these systems act like data-literate colleagues - they don't replace doctors, but they surface insights that improve care decisions.

One rural hospital in Australia used a machine-learning tool to identify patients likely to be readmitted within 28 days, giving clinicians a critical intervention window.

4. Decision-support AI: Clinical agents and diagnostics

Lessons from early adopters

Organisations across New Zealand and Australia that are getting real value from AI aren’t necessarily the ones with the biggest budgets or flashiest tech. They’re the ones that treat AI as an operational improvement tool - not a moonshot project. The early adopters of AI in healthcare are showing us a pattern of success that’s pragmatic, incremental and people first.

The lesson from early adopters is not that AI will fix everything. It’s that AI, when carefully scoped and people-centred, can start solving something - and that’s enough to begin. It starts with solving one job that’s already being done inefficiently, augmenting the team doing it, and learning as you go. The rest can follow.

Below are six key lessons that have emerged from their journeys:

Don't try to solve everything at once

Early success comes from narrowing the focus. In both New Zealand and Australia, medium-sized providers are deploying AI in specific, high-impact use cases - like scheduling surgeries more efficiently or answering patient FAQs after hours.

One hospital in Australia used AI to automate 70% of its medical supply orders. Another implemented a surgery scheduling tool to better match staffing with demand. These are not futuristic transformations - they’re highly targeted operational wins.

Start small. Find the pain points staff complain about most often. That's usually the best place to begin.

Get the team involved early

Organisations that engage frontline teams - especially clinical and operational leads - tend to have better adoption outcomes. Involving clinicians in the selection and evaluation of AI tools not only builds trust, it often improves the tool’s performance by surfacing overlooked edge cases or data issues.

In several case studies, nurses and administrators were included in AI pilot planning. This had a direct impact on the quality of implementation and long-term usability.

AI works best when it augments, not dictates. Implementation is smoother when end users feel consulted, not overridden.

Don't underestimate the value of non-clinical AI

Some of the most immediate benefits have come from workflow automation. Chatbots answering basic patient queries. AI scribes handling clinical documentation. Scheduling systems reducing unused theatre slots. Triage bots providing symptom advice overnight.

These assistive tools are far less risky to deploy and can create quick wins - saving clinician time, improving responsiveness, and reducing operational bottlenecks.

For most medium-sized health organisations, non-clinical AI is the best entry point.

Build a feedback loop, not a set-and-forget solution

In every successful AI case study cited, there’s a human still in the loop. Whether it’s a radiologist reviewing AI-generated imaging risk scores or an administrator overseeing predictive demand models, the best AI implementations create systems where people remain the final decision-makers.

This not only helps mitigate risk, but allows organisations to tune the system over time. What works in week one might need adjusting by month six.

The best AI tools are not plug-and-play. They evolve with your organisation.

Use AI to stretch, not replace, your workforce

New Zealand health providers are experimenting with “digital humans” that can help patients understand medication instructions or follow-up care, particularly in rural areas or after hours. These are not replacing clinicians; they’re filling coverage gaps.

Likewise, chatbot therapist tools are being explored to support emotional health, particularly among youth who feel more comfortable seeking anonymous help. While not a substitute for professional care, they reduce pressure on overburdened services and provide access where none might otherwise exist.

AI becomes most valuable when it extends the reach of your team - not when it tries to replicate them.

Evaluate based on outcomes, not features

The standout organisations measure AI success through outcomes: shorter wait times, higher staff efficiency, reduced missed appointments, earlier diagnoses. Volpara Health’s AI wasn’t just built because it was novel - it was adopted globally because it improved breast cancer detection and enabled more personalised prevention plans.

Don't start with the tech specs. Ask: what will this change for our staff or our patients?

Conversations to have before implementing AI

Before diving into implementation, health leaders must align internally. Based on Australasian research and government guidance, here are five questions every executive team should address.

1. What is the problem we're solving?

AI is not a strategy - it's a tool. Is the goal to reduce wait times? Free up admin staff? Improve diagnostic accuracy?

2. Do we have enough data?

AI systems rely on data quality. Are your patient records clean and digitised? Do you know where the gaps are?

3. Who will monitor and review AI decisions?

Even when AI is accurate, errors still occur. Clinical oversight is critical.

4. What are our ethical guardrails?

AI tools can encode bias, especially in areas like mental health or triage. What safeguards are in place?

5. How will we evaluate success?

Define what good looks like. Is it a reduction in call centre load? A rise in early diagnoses? Put metrics around it.

Risks and realities

For all the momentum behind AI in healthcare, the road to successful implementation is neither short nor straightforward. Early adopters are already showing that AI can unlock real value - but only when introduced with a clear purpose, a structured process, and firm oversight.

In fact, some of the strongest findings across the New Zealand and Australian case studies relate not to success stories, but to what can go wrong when AI is poorly governed, rushed, or misunderstood. These risks aren’t a reason to avoid AI - but they are a reason to move thoughtfully.

1. Over-reliance without appropriate human oversight

Many AI tools, especially those supporting decision-making, offer recommendations - not answers. But if staff misunderstand this distinction, there's a risk that human judgement gets sidelined. This is particularly dangerous in clinical contexts, where AI predictions may seem precise but still miss the nuance of patient context, cultural needs or emerging symptoms. In one high-profile example outside our region, a sepsis prediction model failed to detect several life-threatening cases, resulting in widespread mistrust.

The most successful organisations put clear structures in place:

Clinicians stay in control and validate AI outputs before acting
Escalation protocols are in place for uncertain or ambiguous AI recommendations
Staff are trained not just in how to use AI, but in when not to

AI is not a replacement for human expertise. It's a tool that requires human judgement to be safely applied.

2. Black-box algorithms and lack of transparency

Some AI systems - particularly those purchased as commercial off-the-shelf tools - do not reveal how their decisions are made. This is known as “black-box AI.” For health organisations, this can be a major issue. If a tool makes a recommendation that turns out to be flawed or harmful, leaders need to be able to interrogate the process behind it. Without transparency, there’s no accountability - and no way to improve the system over time.

To mitigate this, organisations should:

Prioritise AI systems with built-in auditability and clinician-facing explanations
Demand clear documentation from vendors
Pilot tools in parallel with human decision-making before full deployment

In clinical settings, explainability isn't optional. It's essential to safe, ethical use.

3. Inadequate data or biased data sets

AI systems are only as good as the data they’re trained on. If the training data is incomplete, outdated, or biased (e.g. under-representing Māori or Pacific peoples), the tool may make unsafe or inequitable decisions. In New Zealand, where health equity is a national priority, this issue is especially pressing. Digital tools that don’t reflect the realities of the communities they serve risk reinforcing existing disparities.

Mitigation strategies include:

Assessing whether the AI model was developed using local data - or if overseas data might introduce bias
Involving culturally informed practitioners in the tool evaluation process
Stress-testing outputs for edge cases and unintended patterns

AI can either reduce inequity or deepen it - depending entirely on how it's designed, tested and governed.

4. Rushed implementation and misaligned expectations

When AI projects are rushed - driven by hype, board pressure, or vendor enthusiasm - they often fail to deliver on expectations. Staff aren’t properly trained. Workflows aren’t redesigned. Feedback isn’t collected. The result is not transformation, but frustration. The tide of AI adoption is inevitable, but success comes from channelling that momentum through structured implementation rather than fighting it or diving in unprepared.

With upfront stakeholder buy-in
With well-scoped pilots
With iterative refinement

Organisations that treat AI as “just another IT rollout” tend to struggle. The technology may be new, but the discipline around change management still applies.

5. Regulatory and reputational risk

Healthcare is one of the most tightly regulated sectors. If AI systems breach privacy, make inaccurate predictions, or affect care quality, the consequences can extend far beyond internal operations into the media, into litigation, and into patient trust. And that trust is hard-won. A poorly performing AI chatbot or failed trial can damage not only patient confidence but staff morale and future innovation appetite.

Fortunately, both Australia and New Zealand are building regulatory frameworks to guide safe adoption. In New Zealand, national health authorities are drafting standards to ensure AI systems are evaluated for safety, equity and cultural appropriateness before they’re deployed.

The message is clear: AI can’t be treated as a side experiment. It must meet the same safety and quality standards as any other clinical or administrative system enhancement.

Summary: Proceed with ambition, grounded in governance

AI can, and does, drive better outcomes in healthcare. But only when implementation is grounded in strategy, caution and continuous oversight.

Before moving forward, healthcare leaders should ensure:

There's a strong, clear case for each AI use.

Staff are trained, empowered, and included.

The tool's decision logic is transparent - or clearly monitored.

Equity and privacy are built into the evaluation process.

The final word: Start small, think big

AI isn’t a silver bullet - but it is a practical toolset that’s beginning to make a difference in health services across our region. The most effective organisations aren’t waiting for perfect conditions. They’re identifying a single high-friction workflow - like triage, patient reminders or rostering - and deploying AI tools to support people doing that work.

For PHOs and practice managers, the takeaway is simple: with the right preparation, AI can help you do more with what you’ve got. The key is thoughtful implementation, ongoing governance, and a clear focus on outcomes that matter to your patients and staff.

Need help with your AI strategy?

If you're considering an AI tool and feel like you could benefit from some additional advice or guidance, get in touch for a no-obligation consultation.

Book your free 45-minute conversation

References

AI Forum NZ (2025). AI in Action Report 2. Key findings on AI adoption across NZ small-to-medium businesses. Source: AI Forum NZ | LinkedIn
Department of Industry, Science and Resources (Australia) (2024). Exploring AI Adoption in Australian Businesses. AI uptake across SMEs, with health and education leading at 45%. Source: industry.gov.au
Salesforce (2024). New Zealand SMBs Using AI See Positive Revenue Growth. Insights into efficiency and revenue gains from AI adoption. Source: salesforce.com/au
Office of the Prime Minister’s Chief Science Advisor (NZ) (2023). Capturing the Benefits of AI in Healthcare for New Zealand. Government commissioned report on healthcare AI case studies and guidelines. Source: dpmc.govt.nz
PwC Australia (2021). Six Examples of AI in Healthcare. Case studies covering demand forecasting, readmission prediction and operational AI. Source: pwc.com.au
Medscape & HIMSS (2024). AI Adoption by Health Systems Report. Global survey showing 86% of health systems are using AI in some form. Source: linkedin.com
The Medical Journal of Australia (2024). Why Clinical Artificial Intelligence Is (Almost) Non-Existent in Australian Hospitals—and How to Fix It. A perspective on the cautious uptake and governance gaps. Source: mja.com.au
RNZ News (2024). Young People Turning to AI Therapist Bots. A look at the social impact and user perceptions of AI chat tools. Source: rnz.co.nz
The Medical Journal of Australia (2024). Potential of Medical Scribes to Allay the Burden of Documentation and Enhance Efficiency in Australian Emergency Departments. Research on medical scribe implementation in Australian healthcare settings. Source: mja.com.au