New research from AWS gives us a useful snapshot of where AI adoption has reached in New Zealand.
Half of the businesses surveyed now report using AI consistently, up from 42% a year ago. Only 15% have a formal AI strategy. The research covered 1,000 business leaders and 1,000 members of the public.
The 50% figure will probably get most of the attention. I think the more interesting question is what happens when widespread use starts to require deliberate organisational decisions.
AWS commissioned the research and is itself a major supplier of cloud and AI services, so its commercial interest is worth keeping in mind. The reported business benefits also need to be read carefully. For example, 84% of AI adopting businesses reported increased revenue, but that does not establish that AI caused the increase.
Businesses with a formal AI strategy were more than twice as likely to report advanced AI use, at 41% compared with 19%. They were also much more likely to say they could measure the return on their AI investment, at 63% compared with 34%.
AWS describes this as making AI deliberate. Deciding where it can create value, who funds it and how it will be measured.
That gets close to the governance issue.
At what point does accumulated AI use require an explicit organisation level decision?
Someone using an AI tool to improve the wording of an email is very different from AI being used in customer interactions, employment decisions, professional advice, sensitive information, products and services, or other activities with material consequences.
As the significance of those uses increases, the people formally accountable for the organisation should have a progressively clearer picture of what is happening. They should be able to explain which uses matter, who is accountable for them, what risks have been considered and accepted, what boundaries have been set, and how they know those arrangements are actually operating.
A formal AI strategy does not answer those questions. An organisation could have a good strategy and weak governance, or sensible governance without producing a document called an AI strategy.
So the 15% figure should not be read as evidence that the other 85% lack AI governance. The research does not tell us that.
It does tell us that reported AI use is now far more common than having a formal AI strategy.
That matters because adoption can build gradually. Staff start using useful tools, teams solve local problems, functions introduce new products and existing vendors add AI capability to systems already in use. None of those decisions necessarily looks significant on its own.
The cumulative effect can be different.
An organisation can end up with AI influencing important activities without anyone having stepped back and looked at the exposure as a whole.
The research also found that startups reported higher levels of advanced AI use than large enterprises and SMEs: 31%, compared with 27% and 16% respectively.
That matters because organisation size tells us relatively little about AI exposure. A smaller organisation may have significant AI use in consequential activities, while a much larger one may be using it mainly for low-risk productivity tasks.
The better question is what AI is doing, where it is being used and what happens if it gets something wrong.
For boards and executive teams, there is a simple place to start:
What AI use across this organisation would we expect to know about?
Then ask whether anyone can provide a reliable answer.
If they can, the next questions are about accountability, risk, boundaries and evidence.
If they cannot, that gap is worth understanding before adding more policy, process or technology.
The significance of the AWS research is that AI use is now widespread enough that more organisations will need to know when individual and functional adoption has created an exposure that warrants an explicit organisational response.
That is when AI governance needs to become deliberate.
Source: AWS, Unlocking New Zealand’s AI Potential 2026, released 22 September 2026. AWS commissioned the research; fieldwork was undertaken by Strand Partners.