Most governments discover they need AI governance once things go wrong. New Zealand has taken a different approach: building the framework first, then standing up the agency designed to live inside it. This month, the Government Digital Delivery Agency, known in te reo Maori as Te Punaha Matihiko, formally opened its doors inside Te Kawa Mataaho, the Public Service Commission. It inherits a Public Service AI Framework that has been in place since January 2025 and a National AI Strategy launched last July. The sequencing matters.
The new agency consolidates digital investment and procurement that was previously distributed across the public sector, with the Government Chief Digital Officer role shifting from the Department of Internal Affairs to sit at its centre. The financial logic is blunt: Minister Judith Collins has pointed to potential savings of up to 30 percent on a projected NZ$13 billion technology spend over five years. That is around NZ$3.9 billion. Centralisation at that scale, with AI baked into the mandate from day one, makes the governance question genuinely consequential rather than academic.
The Public Service AI Framework that underpins the agency's work draws from the OECD's values-based AI principles, which Cabinet formally endorsed in June 2024. Five principles shape how agencies are expected to approach AI: inclusive and sustainable development, human-centred values, transparency and explainability, safety and security, and accountability. The framework is not a binding regulatory instrument. Agencies are encouraged to align with it rather than compelled to. Whether that voluntary posture holds as AI deployment scales up is one of the open questions the new agency will need to answer.
What is structurally distinctive about this arrangement, and largely unreported in international AI coverage, is the explicit embedding of Maori digital sovereignty into the agency's leadership design. A dedicated Director Maori Digital and Crown Relationships role sits within the executive structure, a recognition that how government AI systems are designed and deployed is not a culturally neutral question. For Maori communities, the stakes around data sovereignty, algorithmic decision-making, and access to digital services carry specific historical weight. Treating that as a structural concern rather than a consultation afterthought is a design choice worth noting.
The broader picture is a small country making a deliberate bet on coordination over competition. Where larger governments have tended to let AI adoption race ahead of governance, often with predictable friction, New Zealand is testing whether you can reverse the order: agree on principles, build the institution, then accelerate deployment. The Public Service AI Work Programme commits to a two-year plan of initiatives. The new agency's first task will be demonstrating that governance scaffolding can enable rather than obstruct the ambition it was designed to support.