The diagnosis in that piece feels right, New Zealand does have an issue here. Naming a problem correctly doesn't close it though, and the closer I read, the more it felt like the piece was describing something I'd already watched happen live in a conference context. I'm left thinking we need more of a grass-roots revolution here, not a polite wait for more national boldness to fix the issue.
The productivity numbers behind that discomfort check out. Deloitte puts New Zealand at 63rd out of 67 mid to high income countries on productivity, a ranking that lines up with Treasury and OECD data showing the gap to the top half of the OECD has been widening for three decades. That's not spin.
The adoption numbers are real too, but there's a sharper figure sitting right next to the ones getting quoted. Datacom's 2025 State of AI Index is where the familiar stats come from: 87% of NZ organisations using some form of AI, 28% at an advanced stage. What gets left out is the breakdown underneath that. Only 12% have scaled AI across the whole business. 46% are still running pilots. Eight in ten of us have opened the tool. Only one in ten has actually changed how the business runs because of it.
A lot of the public discourse on this, ours included at times, is aimed at getting more people over that same entry-level line. Open Copilot, try the chatbot, summarise those documents. That's not where the gap actually sits anymore. Most people are already there. The gap is what happens in the two years after that, and almost nobody seems to be actively sharing the war stories about that part.
Calling for boldness and kiwi-ingenuity to ignite is easy. But boldness isn't the missing ingredient. I don't believe most organisations are short on ambition. They're short on the unglamorous mechanics of how we go about doing this well, and those mechanics just don't show up in a keynote.
We took a group of colleagues along and watched the gap up close, across a handful of presentations, and a room full of people at varying stages of actually trying this. A few things stuck.
Adoption is not adaptation
The sharpest point of the day: if you automate the way you already work, you bake your inefficiencies into the new tooling. That's adoption, and it's what most organisations are doing while calling it transformation.
Adaptation is asking the harder question underneath. How would you do this work if you'd had these tools from the start? That's not (just) a technology question. It's a question about which parts of a process should exist at all.
We ask ourselves this on our own AI-native transition as much as we ask it of partners who come to us wanting to "add AI" to a process that's already in place.
The enterprise stranglehold
The barrier we hear about constantly from the regulated organisations we work with isn't appetite. It's approval. A small business can have three people trialling a new tool by Friday. A government agency or a regulated enterprise can spend months getting a single piece of software through security review and procurement before anyone's allowed near it (and for good reason given what's actually at stake in those environments). The organisations with the most to gain from AI are often the ones least able to move.
The answer isn't to wave that away. Security and compliance obligations in health, transport, and government aren't red tape for their own sake. The path that's actually worked for us is bundling AI capability into products and partnerships that have already cleared that bar, rather than asking every enterprise to run the same gauntlet from scratch for every new tool. Slower than a keynote allows for. But the one we reckon holds up.
AI confetti
One line from the day earned a knowing laugh from the whole room: you can't go sprinkling AI confetti around an organisation and expect results. It ends up in the corners, under the chairs, and you only find out how deep it is when you move the furniture. Three teams running three different tools with no shared intent isn't transformation. It's mess with a better interface, and right now, senior leadership approval too, because it at least feels like momentum.
Context drag
The same tool produces wildly different outcomes in two organisations, and the tool isn't the variable. In one team it improves speed and decision quality. In another it produces more summaries, more meetings, and a strange sense that everyone is busier than before. AI lands inside whatever pattern system already exists. It doesn't neutralise a dysfunctional one. It usually amplifies it.
Accountability drift
The most useful framing of the day had nothing to do with tools. Accountability drift is the slow, almost invisible movement of judgement away from the person whose name is on the output. It doesn't feel like failure while it's happening. It feels like efficiency. A little less thinking, a little more summarising, a little more relief. The captain stays in the seat even when the autopilot is flying. Nobody accepts "the co-pilot was flying" as an answer when something goes wrong, and that seat is getting harder to sit in.
These aren't abstractions. They're what the absorption gap actually looks like from inside a project, and none of them get fixed by a national mindset shift.
The gap inside the gap
Here's what stayed with us longer than any of the frameworks. The day itself split cleanly in two. The best sessions came from people who had genuinely built something, or who brought a framework or solution they'd tested against real world work. The weakest sessions were pitched squarely at large enterprises, and they were underwhelming in exactly the same way. No depth, no real innovation. Entry-level AI, the kind already sitting inside tools these organisations had owned for years, repackaged with an enterprise price tag and presented as transformation. You could hear the difference within a minute of someone opening their mouth. The gap between "I've done this" and "I've bought this" was audible, and it's the same gap the absorption problem is describing at a national scale, just compressed into a single room. The enterprises stuck behind months of security review are, in the same breath, the ones being sold familiar tooling from familiar vendors as the answer to that wait.
That's what changed something for us. Watching it happen live, all day, made it hard to keep treating this as someone else's problem to solve and ours to support from the sidelines. If the people worth listening to were the ones who'd actually built things, and the summit circuit keeps rewarding the ones who'd only thought about it, then more panels don't feel like the fix.
We've spent enough of this year in rooms like that to notice we're no longer content waiting to be asked to help. Something has shifted from "we help organisations close this gap" to "we should be putting the right people in a room together ourselves." A discomfort we've decided to act on rather than write about again next year.
We don't know yet exactly what that looks like. But we strongly suspect it'll start with fewer slides and more voice given to the people who've actually done the work.
FAQs
What is the AI absorption gap?
The AI absorption gap is the difference between an organisation's ambition to use AI and its actual ability to embed it well. Most New Zealand organisations aren't short on interest in AI. They're short on the practical discipline needed to use it effectively.
What's the difference between AI adoption and AI adaptation?
Adoption means automating existing processes with AI tools, which usually just bakes in old inefficiencies faster. Adaptation means rethinking how the work should be done given the tools now available, which is a harder and more valuable question.
What is accountability drift?
Accountability drift is the gradual shift of judgement away from the person responsible for an output, as AI-assisted summarising and drafting quietly replaces independent thinking. It rarely feels like a mistake while it's happening.
Why doesn't better AI tooling fix the absorption gap on its own?
Because the tool isn't the variable that determines the outcome. The existing patterns of trust, accountability, and decision-making inside an organisation are. The same tool produces very different results depending on what it lands in.
How many New Zealand businesses have actually scaled AI, not just tried it?
According to Datacom's 2025 State of AI Index, 87% of New Zealand organisations use some form of AI, but only 12% have scaled it across the whole business. Around 46% are still running pilots. Most of the country has opened the tool. Very few have changed how the business runs because of it.
