AI in Public Sector in Aotearoa New Zealand: A Living Whitepaper

Updated: July 13, 2026

Introduction

AI adoption in New Zealand’s public sector has continued to move from pilot-era experimentation toward a more formal operating model built around shared infrastructure, central coordination, and tightly governed frontline use. Since the previous June 10, 2026 edition, the most important shift is not a single new tool but a sharper system design: the Digital Government Target State is now the dominant architecture, the Service Modernisation Roadmap has been decommissioned, and a July 2026 rapid review has explicitly identified “AI for government” as one of the small number of foundational common-capability projects that should be prioritised centrally. (digital.govt.nz)

The practical pattern remains conservative. New Zealand agencies are still using AI mostly for summarisation, search, documentation, workflow support, pattern recognition, and service navigation rather than autonomous state decision-making. But the environment around those use cases is becoming more structured: the Government Digital Delivery Agency now sits inside the Public Service Commission, regulator-specific guidance has arrived, Health NZ has tightened clinical AI guardrails, and local government is building more explicit public AI governance. (digital.govt.nz)

Executive Snapshot

• AI is now embedded in the state’s central digital operating model. The Digital Government Target State requires agencies to align to shared digital public infrastructure, and that infrastructure explicitly includes an AI broker/gateway, AI platform services, semantic search, and data and AI safeguards. (dns.govt.nz)
• July 2026 brought a stronger centralisation signal. A rapid review released on July 2, 2026 recommended a reset of digital delivery, including identifying a small set of foundational common-capability projects, with the reviewers’ hypothesis explicitly naming digital identity, data exchange, and AI for government. (publicservice.govt.nz)
• The core whole-of-system evidence base is still the 2025 AI survey. The latest published cross-agency snapshot remains 70 agencies reporting 272 AI use cases, including 55 operational deployments, up from 108 use cases across 37 agencies in 2024. (digital.govt.nz)
• Public-sector AI is still mostly assistive, not autonomous. Survey findings, Health NZ guidance, and current case studies all point to bounded uses such as drafting, summarising, documentation, pattern analysis, and workflow support under human oversight. (digital.govt.nz)
• Health remains the clearest frontier for scaled operational AI. Emergency-department AI scribes are live nationwide, Health NZ has published formal GenAI guidance, and HealthX has added an international partnership to strengthen AI and digital innovation assessment and scaling. (beehive.govt.nz)
• Capability is improving, but depth is still limited. In the 2025 Public Service Census, 33% of public servants had used AI for work and 14% used it regularly; meanwhile 88% said they felt confident learning new digital skills. (publicservice.govt.nz)
• Trust is still the biggest adoption constraint. General trust in public services remains high, with 84% trusting the most recent government service they used as of March 2026, but Kantar’s AI-specific research found only 4% of New Zealanders felt well informed about how government uses AI. (publicservice.govt.nz)
• Biometrics is the sharpest near-term compliance pressure point. The Biometric Processing Privacy Code transition period for existing biometric processing ends on August 3, 2026. (privacy.org.nz)

What Changed Since the June 10, 2026 Edition

1) The Digital Government Target State has clearly superseded the old service-modernisation frame

On July 1, 2026, official digital-government material confirmed the Service Modernisation Roadmap was decommissioned in June 2026 and that work now continues through the Digital Government Target State. Official guidance also says AI will become increasingly important in all aspects of Target State delivery. This is a meaningful shift because it places AI more firmly inside a whole-of-government architecture and investment model, rather than as a parallel innovation stream. (digital.govt.nz)

2) A July rapid review reframed AI as a foundational common capability

The Delivering Digital Government – Reset Plan, released alongside the Public Service Commission’s July 2 announcement, is the strongest new system-level development since the last edition. The review argues digital delivery is fragmented and recommends tighter prioritisation, a narrower and stronger GDDA role, and a small number of foundational common-capability projects. Its explicit hypothesis includes AI for government alongside digital identity and data exchange. (publicservice.govt.nz)

3) Health NZ added a new international AI innovation channel

On July 8, 2026, Health NZ announced that its HealthX programme had entered into an agreement with UCLPartners in the United Kingdom to support the assessment, testing, and scaling of digital and AI innovation. The official description emphasises reducing administrative burden, supporting clinical workflow, and improving patient experience, which is consistent with New Zealand’s broader human-in-the-loop posture. (healthnz.govt.nz)

4) The innovation pipeline is becoming more visible, even where deployment is still emerging

The Spirit of Service Awards 2026 finalists, announced on July 2, 2026, included HealthX programme of AI and digital innovation and a Ministry of Justice initiative on using AI agents in Commercial Services. That does not prove large-scale production rollout, but it is a strong public signal that AI initiatives are now mainstream enough to be recognised inside the sector’s own performance and innovation ecosystem. (publicservice.govt.nz)

Current State of AI Adoption

Governance and Operating Model

The governance stack remains anchored in the Public Service AI Framework, the Responsible AI Guidance for the Public Service: GenAI, and the Public Service AI Work Programme. The work programme still describes a 2-year plan with 15 initiatives across 4 focus areas: common use tools, safe and responsible AI, customer and partnerships, and AI workforce. Its published deliverables include a Public Service AI Hub, annual AI use-case measurement, strategic supplier relationships, public-sector AI events, and refreshed executive and practitioner training. (digital.govt.nz)

Institutionally, the biggest structural change is that the Government Digital Delivery Agency (GDDA) was established on April 1, 2026 inside the Public Service Commission and now carries the functions of the former Government Chief Digital Office. That matters because AI stewardship is now sitting inside a broader machinery-of-government reform effort rather than a looser digital transformation setting. (digital.govt.nz)

Governance is also becoming more domain-specific. The Ministry for Regulation released dedicated AI guidance for regulators on May 27, 2026, and version 1.1 adds resources on Māori data and Māori data sovereignty. The guidance emphasises starting small, applying safeguards, and keeping people responsible for decisions. (regulation.govt.nz)

Health NZ has gone further with sector-specific operational rules. Its guidance on generative AI and large language models, reviewed in May 2026, says staff must not enter sensitive patient or organisational information into unapproved tools and must not use LLMs for clinical decisions or personalised patient advice. (healthnz.govt.nz)

Scale and Maturity

The latest published whole-of-government baseline is still the 2025 cross-agency AI survey. It found 272 AI use cases across 70 agencies, including 55 operational deployments, which is a sharp rise from 15 operational use cases in 2024. The official highlights say AI uptake is increasing, the ecosystem is growing, more use cases are reaching the operate-and-use stage, and agencies are already reporting productivity and efficiency gains. (digital.govt.nz)

At the same time, the maturity profile remains uneven. The same survey says the main barriers are lack of skills and capability, anticipated costs, and security. That is consistent with a sector that has moved beyond curiosity but has not yet industrialised AI adoption evenly across agencies. (digital.govt.nz)

Workforce readiness is improving, but still shallow. In the 2025 Public Service Census, 33% of public servants said they had used AI for work and 14% said they used it regularly. Yet 42% agreed their organisation takes advantage of technology to deliver better public services, suggesting the sector’s institutional maturity still lags individual experimentation. (publicservice.govt.nz)

Shared Infrastructure and Service Delivery

The Digital Government Target State is now the strategic centre of gravity. It says agencies must align to a shared Digital Public Infrastructure layer and explicitly shows AI broker/gateway, AI platform services, semantic search, and data and AI safeguards as core components. This is the clearest published sign that New Zealand is designing for reusable, cross-agency AI capability rather than a proliferation of isolated agency stacks. (dns.govt.nz)

The Govt.nz app continues to advance as the state’s common digital front door. Official programme material says Version 2.0 is live with a digital wallet, the Government Credential Issuance Platform is released, digital credentials are expected to begin appearing progressively from mid-2026, and secure messaging and notifications are positioned for initial agency onboarding from July 2026. The app page still presents these as staged rollouts rather than fully completed capabilities. (digital.govt.nz)

Procurement and infrastructure are also being centralised. The C3 (Common Capabilities and Cloud) Programme says new Marketplace channels for infrastructure, telecommunications, and managed security services are intended to support agencies from the 1 July 2026 launch window onward, with transactions able to begin for approved listings from that date. (digital.govt.nz)

Public Trust and Social Licence

New Zealand’s public sector still benefits from strong general trust. As of March 2026, the Kiwis Count framework reported 84% trust in the most recently used public service and 64% trust in the Public Service brand overall. (publicservice.govt.nz)

But AI-specific trust remains much thinner. Kantar’s November 2025 research found only 4% of New Zealanders felt well informed about how government uses AI, 55% were comfortable with government AI use, 66% wanted the option to deal with a human rather than AI, and 60% supported an independent body to oversee AI use. The trust story, then, is not that government is broadly distrusted; it is that AI transparency and AI-specific consent still lag the wider trust base. (kantarnewzealand.com)

Recent News and Policy Developments

Digital delivery has entered a reset phase

The July 2, 2026 rapid review found digital investment is fragmented, prioritisation is weak, the GDDA has limited influence over funding and procurement decisions, and stronger system leadership is needed. The attached reset plan recommends urgent reprioritisation, stronger central strategy and assurance, and a tighter all-of-government governance model for foundational digital capabilities. (publicservice.govt.nz)

AI remains linked to public-service productivity reform

The May 19, 2026 public-service overhaul announcement still matters because it explicitly connected machinery-of-government reform with increased use of AI and other digital tools. That signal has now been reinforced by the July digital reset rather than displaced by it. (beehive.govt.nz)

Health NZ is broadening from deployment to evidence-led scaling

The HealthX-UCLPartners agreement adds a new layer to the health AI story: not just adopting tools, but building an institutional pathway for horizon scanning, validation, and scaling. Officially, the partnership is framed around careful assessment and evidence-led uptake. (healthnz.govt.nz)

Public-facing and citizen-service AI remains cautious

Health NZ’s AI-supported symptom checker is still framed as an exploration rather than a live autonomous triage system. Official material says it is intended to help people navigate to the right health service and support, not replace, clinical decision-making, and that governance arrangements are still being established. (healthnz.govt.nz)

Research Overview

• Latest whole-of-government quantitative baseline: 2025 cross-agency survey, with 272 use cases across 70 agencies and 55 operational deployments. (digital.govt.nz)
• Latest system-level stewardship view: State of the Public Service 2025 says the sector must reduce fragmentation, accelerate and harmonise technology and AI adoption, and create a unified digital front door. (publicservice.govt.nz)
• Latest workforce evidence: 2025 Public Service Census shows rising individual AI experimentation but modest regular use. (publicservice.govt.nz)
• Latest public sentiment evidence: strong general trust in public services, but limited understanding of government AI use and a clear demand for human fallback and oversight. (publicservice.govt.nz)

Case Studies

Case Study 1: Health NZ emergency-department AI scribe

Health New Zealand’s emergency-department AI scribe remains the clearest example of scaled frontline AI in the country. As announced on February 28, 2026, the tool is live in all emergency departments, with rollout completed to 1,250 doctors and frontline staff. Pilot results indicated doctors using the tool saw one additional patient per shift on average. (beehive.govt.nz)

Why it matters: this is operational AI at national scale, but still in the augmentation category: documentation support that reduces administrative burden while leaving clinical judgement with people. That model continues to define the public sector’s comfort zone. (beehive.govt.nz)

Case Study 2: HealthX programme and clinical AI governance

Health NZ’s HealthX programme is now publicly visible as a national innovation vehicle for clinician-led AI and digital tools, and it was named a finalist in the 2026 Spirit of Service Awards. Health NZ’s new UK partnership suggests the programme is being built as an assessment-and-scaling function, not just a collection of pilots. (publicservice.govt.nz)

Why it matters: HealthX indicates a maturing model for public-sector AI adoption: validate first, scale second, and connect deployment decisions to governance, external evidence, and clinical workflow impact. (publicservice.govt.nz)

Case Study 3: Govt.nz app and the Digital Public Infrastructure model

The Govt.nz app is no longer just a concept. It now has a live core app, a digital wallet, emergency-management updates and alerts, and a released credential issuance platform. The larger significance is architectural: it is being built as an all-of-government service channel connected to shared digital infrastructure. (digital.govt.nz)

Why it matters: the public sector appears to be favouring a common service-delivery layer over fragmented agency-by-agency customer AI and mobile experiences. That is partly an inference from the published architecture, but it is a strong one. (dns.govt.nz)

Case Study 4: Hutt City Council and the local-government AI model

Hutt City Council remains the strongest visible local-government adopter. Its public AI page describes custom assistants, a public AI register, an AI risk-management framework, governance oversight, emergency-response experimentation through CERA, and workflow tools for property and consent functions. It also reports measurable administrative gains, such as 3 to 5 minutes saved per invoice in one automation example. (huttcity.govt.nz)

Why it matters: Hutt shows what mature sub-national adoption looks like in New Zealand: public disclosure, internal governance, narrow but useful workflow redesign, and visible experimentation tied to community outcomes. (huttcity.govt.nz)

Case Study 5: Environment Canterbury’s governance-first posture

In May 2026, Environment Canterbury re-established its Artificial Intelligence Working Group to examine how AI is being used to analyse data, support decision-making, and serve local communities. Workshops are intended to be open to the public. (ecan.govt.nz)

Why it matters: not all adoption is about tools in production. In some parts of the sector, governance capability itself is now being built as a formal public function. (ecan.govt.nz)

Adoption Patterns and Trends

1) AI is becoming expected productivity infrastructure

The combination of the public-service overhaul, the Digital Government Target State, and the July reset plan makes it clear that AI is now being treated as part of the expected productivity stack for government, not simply a discretionary innovation layer. (beehive.govt.nz)

2) Shared capabilities are winning over agency-by-agency buildouts

The architecture, procurement changes, and reset recommendations all favour reusable platforms, common assurance, and central coordination. New Zealand’s strategic model is increasingly “shared AI rails, agency-specific use cases.” (dns.govt.nz)

3) Human accountability remains the hard boundary

Across regulator guidance, Health NZ rules, and active public deployments, the dominant norm is that AI supports decisions rather than replacing accountable humans. (regulation.govt.nz)

4) Health is the most advanced public-sector AI domain

Health now combines national deployment, formal AI governance, new evaluation pathways, and a visible innovation programme. No other public-sector domain currently shows that same mix of scale, governance, and pipeline maturity. (beehive.govt.nz)

5) Local government is becoming an important experimentation layer

Hutt City Council and Environment Canterbury show that councils are no longer just watching central government. They are building their own governance models, public disclosures, and operational use cases. (ecan.govt.nz)

Pressure Points and Risks

Skills, cost, and security are still the main inhibitors

The official cross-agency survey continues to identify skills, cost, and security as the main barriers to adoption. That remains the cleanest published explanation for why growth in use cases has not yet translated into evenly distributed maturity. (digital.govt.nz)

AI-specific social licence remains thinner than general trust in government

The gap between high trust in public services overall and low public understanding of AI use is still the sector’s most important political and design constraint. Without stronger disclosure and communication, more deployment may widen that gap rather than close it. (publicservice.govt.nz)

Biometrics is an immediate governance test

The Biometric Processing Privacy Code 2025 transition deadline of August 3, 2026 is now close. At the same time, the Public Service Commissioner has announced an investigation into integrity concerns linked to MBIE’s failed Biometric Capability Update project, and MBIE has acknowledged the project consumed $33 million without delivering anything. While biometric processing is not the whole AI story, it is a live reminder that trust can be damaged as much by poor delivery and governance as by the technology itself. (privacy.org.nz)

Overall Assessment

As of July 13, 2026, AI adoption in New Zealand’s public sector is best described as centralising implementation. The sector has moved beyond a framework-only phase: it has a formal AI work programme, a clearer shared architecture, a central digital delivery agency, visible frontline deployments, and stronger domain-specific operating guidance. The July reset plan has also made AI more explicitly part of the state’s common-capability agenda. (digital.govt.nz)

But the operating model remains deliberately bounded. The most mature New Zealand public-sector examples still use AI to compress paperwork, improve service navigation, support analysis, and reduce workflow friction while keeping humans clearly accountable. That is visible in emergency departments, regulatory guidance, health innovation governance, local-government AI registers, and the architecture behind the Govt.nz app. (beehive.govt.nz)

The key change since the last edition is that AI is now more tightly tied to the state’s core redesign agenda. The next phase will be decided less by whether agencies want to try AI, and more by whether New Zealand can deliver shared platforms, trusted governance, and credible public transparency at the same pace as adoption pressure rises. (digital.govt.nz)