NZ Government AI and Decision Governance: When Copilot Isn't Enough
- David Harvey
- May 21
- 4 min read
Updated: Jul 15
A response to Phil Nevell, Andrew Hamilton — and to New Zealand's AI moment.
Phil Nevell asked the question most AI commentary avoids: who actually benefits, and who just becomes more dependent? Andrew Hamilton answered it more precisely than most policy papers manage — build capability first, so people arrive at this new world as partners, not passengers.
Then, the same week, the New Zealand Government handed us the cleanest possible test of that question.
The contradiction at the centre
Health New Zealand bans staff from using free AI tools for clinical notes — on accountability and data privacy grounds. Corrections bans AI for sensitive offender assessments for the same reason. MBIE banned AI outright, then rolled out Copilot for productivity. ACC uses it to summarise documents. Parliamentary Service allows ChatGPT, with a fact-check mandate.
And the same government is using AI productivity gains to help justify cutting roughly 9,000 public-service roles — about 14% of the workforce — over two years.
Read that twice. AI is judged too unaccountable to summarise a clinical note, but accountable enough to help reshape 9,000 careers.
That is not hypocrisy. It is a missing layer. There is no governed record sitting between what the AI produces and the consequential decision a human makes with it. No receipt. No trail that survives scrutiny after the savings have been booked. The contradiction is not a failure of intent — it is a failure of infrastructure.
Who the missing layer is actually for
Here is the part nobody in Wellington is saying out loud — and it is Phil's question, answered concretely.
The people losing those roles are not, in the main, people who cannot make decisions. Many have spent entire careers doing exactly that: structured analysis, evidence assessment, stakeholder navigation, policy reasoning. Governed thinking is their craft. It is precisely the skill the AI era claims to want more of, not less.
The tool cited to remove them offers them nothing on the other side of that removal. It does not help a senior policy analyst figure out what comes next, or navigate the most consequential decision of their professional life with evidence, structure, and a record of their reasoning.
Ask Omega does. With a platform that turns hard-won domain expertise into governed decisions other people can rely on, a senior analyst displaced in 2026 could be a Decision Governance Practitioner — in their own field — by 2027. Their expertise is an asset. The platform is what keeps it commercially alive in the AI era. Not replaced by AI. Working above it, as the layer that makes its output accountable.
That is what people-first looks like once it stops being a slogan.
The category the Minister couldn't name
In Parliament, asked whether the government's AI strategy would lean on local or overseas technology, Minister Goldsmith said he was not aware of any local AI provider at the scale of OpenAI, Anthropic's Claude, or Copilot.
He is right that New Zealand does not need to build its own large language model at that scale. That was never the missing piece.
The missing piece is a New Zealand–governed decision layer that sits above any AI tool — Claude, Copilot, whatever comes next — preserves the evidence, records who decided what and why, and survives an Office of the Auditor-General inquiry, a Select Committee hearing, or a judicial review.
A probability distribution doesn't survive that. A receipt does.
Minister — that layer exists. It was built here. It runs now. And it is not for sale to the companies that built Copilot.
What it is — and what I'm not going to explain here
It is called Ask Omega. It sits above the model, not inside it. Every decision it processes leaves an evidence trail and a receipt that belongs to the person who made the call — not to the platform that helped them make it.
It is not a prototype. The billing is live — real transactions, processed and settled, with receipts delivered. The decisions are real. Reasoning belongs to humans, and so does accountability.
I am not going to explain the engine underneath in a blog post. That is not the point of this piece, and the people who need to see how it works will ask to see it.
Andrew's point about sequencing applies directly here: build capability before the displacement arrives, and the safety net stays small, because fewer people need it. Ask Omega is that capability-first model in practical form — not a safety net, not a redistribution scheme, but a way for individuals to arrive at the AI era with something to offer rather than only something to lose.
There is also a way for institutions to take part rather than simply procure: a four-year governed study, beginning this year. A study that turns the decisions being made right now into evidence that lasts. But that is a part I would rather show than describe.
The process behind the front
So, to the question that started this.
Phil asked who is actually first in the Intelligence Age. Andrew answered it: capability, partnership, human accountability, not distributed dependency. This is the production-level answer to the question they posed. It is real, it is running, and it was built here.
New Zealand does not need to own every model it uses. It does need to own the decision layer above them — built here, governed by New Zealand law, answerable to New Zealand institutions.
That layer now exists.
If that question matters to you — whether you are shaping AI strategy in Wellington, sitting on a board making consequential calls, or facing a transition you did not choose — the door is open.
The work is on the wall. Come find me for the rest.
Omega* Sensing is part of the Omega* Unified Ecosystem, developed by Design By Zen, an NZ-based AI Lab. Omega* is the algorithmic engine beneath the ecosystem. SHE ZenAI is the brand of a governed clinical intelligence framework designed for high-trust domains where evidence, not confidence, is the currency of care. Version 1.0, April 2026.
Comments