General-purpose AI assistants such as ChatGPT, Gemini and Claude remain the most commonly used AI technology at 68%.

But despite widespread adoption, businesses are struggling to move to the next phase of maturity.

Only 15% of organisations are at the stage of scaling AI organisation-wide, up slightly from 12% in 2025.

Those using AI to transform core operations actually dropped from 8% last year to 4%.

(The research was conducted by Curia Market Research between July 28 and August 3, surveying 207 New Zealand senior business managers).

Why is serious use of AI low and falling?

One element is bill shock.

The big AI players have shifted larger customers from set fees per employee per month to charging for how many AI “tokens” are actually used.

“As AI platforms increasingly move towards consumption-based pricing models, understanding where AI is creating measurable value, and establishing what AI tool is best for a specific category or task, is likely to become increasingly important when making future investment decisions,” Datacom NZ managing director Peter Nelson says.

“You wouldn’t hire a 50-seater bus to transport two people from Hamilton to Wellington, so why use an expensive frontier AI model to improve an email or create basic content for a sales proposal?”

There have been some well-publicised examples – including Uber – of companies putting caps on AI use because of bill shock.

But Compagnone noted that encouraging staff to use AI as much as possible was even out of fashion at tech firms like Microsoft and Amazon, which have moved away from “token-maxing” (encouraging as much AI use and experimentation as possible), complete with leaderboards, to more deliberate use of the technology.

Organisations needed to get a better handle on when to use expensive “frontier” AI and when to use cheaper, more meat-and-potatoes alternatives, or more industry-targeted, fine-tuned “small language models”.

Lack of leadership

Another is lack of leadership – and a related lack of progress in governance and strategy.

Only 16% of respondents had a clearly defined standalone AI strategy, while 23% have formally embedded AI within a broader business, digital or technology strategy.

A further 23% say their strategy remains high-level or has not yet been fully operationalised.

Only 22% of businesses have a dedicated AI leadership role, including 13% with a chief AI officer and 9% with another dedicated fulltime AI leadership position.

The Herald put it to Compagnone that a firm’s chief technology officer should handle AI. In earlier tech crazes, we haven’t had, say, a “chief cloud officer”.

“For me, it’s a socio-technical technology. It’s something that requires transformation skills because of its impact on people and roles,” Compagnone said.

“Having that role just as part of the IT department or an extension of the CTO ignores the fact that actually it needs to consider governance.

“It touches customer experience … it actually needs to be a horizontal role to allow for that sort of wider transformation. The chief AI officer needs to be an orchestrator across the C-suite.”

The leadership gap extends to Government

“In Australia, every single government department now has a chief AI officer,” Compagnone said.

“And when we surveyed more broadly in our chief AI officer white paper across public and private sector organisations [in Australia], we found that 42% already have a chief AI officer, and that’s going to increase by about 21% in the next 12 months.

“If you contrast that to the 22% in New Zealand, there’s less regulation, a less clear plan and not necessarily an emphasis on the importance of really good senior AI leadership.”

The survey found nearly two-thirds (64%) believe NZ should introduce specific AI legislation and controls. Compagnone said clear rules would give businesses more confidence about where and how they should invest.

And it found a majority (74%) support a dedicated national AI framework and co-ordinating body, similar to Australian Prime Minister Anthony Albanese’s establishment of the Office of AI, which included the mandate for every government agency to have a chief AI officer, measures to co-ordinate AI adoption across Government, and a new requirement for new data centres to underwrite new power generation.

Prime Minister Christopher Luxon earlier said he met with Albanese shortly after his announcement and said new measures were in the works in New Zealand.

Details are pending, in the context of Finance Minister Nicola Willis recently leaning on civil servants to make greater use of AI – though with few specifics.

A Tuanz (Telecommunications Users Association of New Zealand) survey of business leaders found many supported the idea of a Government-backed “clearing house” that could provide the public and private sectors with guidance around AI developments to prevent duplicated effort, or just helping to fill skills gaps.

It could give a verdict on, say, whether the latest Claude model had any security risks, or which cheap Chinese models were safe to use amid data privacy fears, and for what tasks as organisations deal with “token shock” and where it was appropriate to use no-cost open source models.

That concept, which Compagnone backs, has not found traction with our Government so far. Compagnone said Denmark was the model for creating frameworks that made it easier for both the public and private sectors to adopt AI.

Growing investment

Despite only one-third seeing payback, the survey found most (79%) New Zealand businesses have increased their AI investment over the past 12 months, while 73% plan to invest more over the year ahead.

Chris Keall is an Auckland-based member of the Herald’s business team. He joined the Herald in 2018 and is the technology editor and a senior business writer.