Addressing the AI gaps to foster trust in government

Eighteen months on from cabinet announcing its strategic direction on artificial intelligence (AI) in June 2024, and ten months on from the release of New Zealand’s first national AI strategy, the question across government has matured. Where it was once ‘where do we start?’, it is now ‘how do we scale this responsibly, in a way that is genuinely ours?’.

That shift, however modest in framing, represents the most consequential development in this sector since the digital transformation conversations of the late 2010s.

This conversation was at the forefront at the Public Sector Network’s Government Innovation Showcase on 12 May 2026. We spoke to people at the conference about what’s working, what is missing, and why this year’s conference arrived at a decisive moment for artificial intelligence (AI) in the New Zealand public sector.

AI use in government is on the up, but trust is key to continued growth

Despite New Zealanders’ low levels of trust and positivity around AI, the 2025 cross-agency AI survey of NZ government organisations recorded 272 use cases across 70 agencies; more than double the 108 cases across 37 agencies a year earlier. New Zealand is, by international standards, a country whose public sector is moving faster on AI than its public is ready for. That is not an easy place to legislate from, and not an easy place to deliver from. Ultimately, the government requires the trust of New Zealanders to effectively deliver public services. This extends to cover its adoption and use of AI.

The cross-agency survey methodology itself is a good example of how to earn and maintain that trust. Many jurisdictions still rely on press releases and ministerial speeches for their adoption telemetry. Aotearoa, alongside very few other countries, has a publicly available, year-on-year, agency-level dataset that lets us see scale, sector, maturity and barriers in detail. That kind of transparency is the foundation everything else depends on.

We’ve built the policy, now we need to deliver

The 2026 Government budget policy statement promises to address New Zealand’s longer-term productivity challenges and to support the delivery of core public services. A focus on uplifting New Zealand’s digital government services will be crucial to achieving this. That includes investing in AI; through funding and efforts to increase trust amongst New Zealanders in the government’s use of it.

Trust will be built through effective delivery. The assessment is that New Zealand has built a credible policy architecture. To facilitate a coherent delivery system there are five clear gaps for public sector leaders to address.

The trust deficit is structural, not down to lack of information

Kiwis expect a comprehensive, regulatory approach to AI. Government needs to be accountable, and to provide demonstrable guardrails, and a clear answer to the question ‘if this goes wrong, who is on the hook?’.
Capability is concentrated, not distributed

Inside the public service, AI capability is concentrated in a handful of agencies, with the rest doing their best with goodwill and Microsoft Copilot licences. Budget 2025 allocated $213 million for tuition and training subsidies and $64 million for STEM and priority skills, and the New Zealand Institute for Advanced Technology committed $70 million over seven years for AI research and commercialisation. These are necessary, but they are not sufficient. What the system needs is a systematic public service AI capability uplift indexed to specific role families, not a single masterclass repeated.
Data infrastructure is the silent constraint

A consistent message from public sector clients is some version of ‘we could do this if our data were ready’. Legacy systems, fragmented data domains, inconsistent metadata standards, and the unresolved fate of programmes like Hira have left many agencies with the AI ambition of 2026 and the data infrastructure of 2014. Organisations that have managed to embed AI rather than sprinkle it have done so by treating data foundations as a strategic asset, not a back-office concern. Aotearoa cannot afford to under-invest here.
Scaling is harder than starting

Datacom’s 2025 figures show that while 87 percent of large New Zealand organisations are using AI, only 12 percent have scaled it across the enterprise. That bell curve scaling gap is now the defining problem internationally. The next chapter of the story is about orchestration, integration, and the unglamorous work of operating models, change management, and lifecycle assurance.
Te Tiriti has been named, but not yet operationalised in AI delivery

The Public Service AI Framework references Te Tiriti and explicitly acknowledges Māori data sovereignty. The National AI Strategy nods to it. The Algorithm Charter for Aotearoa New Zealand, signed by more than two dozen agencies, commits signatories to embedding a te ao Māori perspective in algorithm design. These are real commitments, and Te Mana Raraunga, Te Kāhui Raraunga and the wider Māori data sovereignty community have done extraordinary work to make the conceptual foundations clear.Yet most agency procurement processes do not yet ask the question ‘who governs the model weights, where does the training data sit, and how have iwi, hapū and Māori data subjects been engaged in design and benefit-sharing?’ Until those questions are standard line items in business cases, the commitment to Te Tiriti in AI is aspirational.

How KPMG is showing up

KPMG New Zealand is investing in this conversation because the next two years will determine whether Aotearoa’s public sector ends up with AI that genuinely serves the country, or merely operates within it. Our perspective is informed by global research, and shaped by local delivery in central agencies, district health systems, infrastructure operators, and Crown Research Institutes.