Datacom’s AI director has a phrase for most of what gets counted as AI adoption in this country. He calls it “AI confetti.”
Individual people, doing individual things, some of them brilliantly, none of it joining up.
Here are the numbers behind the phrase, from Datacom’s 2026 AI Index. Ninety-one per cent of New Zealand organisations use AI in some form. Eighty-one per cent are still in exploration or implementation. Fifteen per cent are scaling it across the organisation. Four per cent are using it to transform core operations.
That last number used to be 8%. It didn’t slow down. It halved.
And from Deloitte’s work for 2degrees: 82% of New Zealand businesses use AI, over a period in which labour productivity fell 0.7%.
Three things we’ve been calling one thing
I think the confusion is structural. Three completely different states get reported under the same heading.
Having it. Licences bought, accounts created, access granted. This is a procurement fact. It tells you nothing.
Using it. People completing tasks faster than they used to. This is real and valuable — and it’s invisible, unevenly distributed, and mostly undocumented.
Working differently. The process itself redesigned around what’s now possible. The handover that no longer exists. The report that writes itself. The role that changed shape.
Adoption measures the first. We’ve been putting it in board packs as though it measured the third.
Ninety-one per cent is a procurement statistic wearing a transformation costume.
We’re not behind on tools
The comparison that should sting is this one, from the Employment Hero research earlier in the year. Thirty-four per cent of New Zealand employers say AI is driving more innovation in their business. In Britain it’s 55%. In Australia, 45%. Asked whether AI is core to how the business operates: 15% here, 24% in Britain.
Same tools. Same vendors. Same models, released here on the same day. A fraction of the reported return.
Whatever is going wrong is not a technology access problem, and it’s not a pricing problem. It’s what happens after the licence is issued.
Why the 4% is falling
Two explanations I find credible, and they compound.
The first is cost discipline. AI pricing is shifting toward consumption billing, and businesses that can’t say which of their AI use is productive are responding the only way they can — by pulling back. Seventy-nine per cent of NZ organisations increased AI investment last year, while only a third said returns exceeded costs. Spend up, confidence down. That combination ends in caps.
The second is harder. Getting from “using it” to “working differently” requires someone to change a process. Processes belong to people. Those people did not ask for this, weren’t consulted, and in many cases quietly suspect that a redesigned process needs fewer of them.
Stage two can be done by an individual in private. Stage three cannot be done without their cooperation. That’s the wall, and no amount of licensing gets you over it.
Three numbers I’d report instead
If I were sitting in a New Zealand leadership meeting this quarter, I’d stop reporting adoption entirely and report these.
How many roles have at least one task that is now permanently done differently. Not faster in one instance — permanently, structurally different.
How many of those changes have been picked up by a second person. A change that one person does is a personal productivity story. A change two people do is the beginning of a method.
How many are written down anywhere.
That third one is the killer, and it’s where most New Zealand businesses will score close to zero. Most of the genuine AI value in this country right now lives in one person’s head and goes on annual leave with them.
What confetti actually is
The gap between 91% and 4% isn’t a technology gap. It’s an organisational one.
Confetti is what you get when forty people each solve their own problem, privately, in their own way, and nobody collects the answers. Every individual piece of it is real. It just doesn’t add up to anything, and you can’t hand it to the next person.
The work isn’t more adoption. We’ve done adoption. The work is collection.