The top AI-using firms now outproduce their peers by a factor of 14

The number that actually matters

OpenAI’s new New Zealand data is being reported as a novelty. New Zealanders send almost 10 million messages to ChatGPT each day, 40% use it daily and 45% say it saves them at least four hours a week. Interesting, but not the point.

The point is buried further down. The top 10% of New Zealand businesses using ChatGPT now generate 14 times as many output tokens per worker as the typical firm. That is not a gap between adopters and non-adopters. It is a gap between businesses that all have access to the same tools. And it is getting wider.

Gawesha Weeratunga of OpenAI’s Economic Research team put it plainly, telling the BNZ Business Breakfast there is “quite a big gap in terms of the businesses here in New Zealand who are embracing AI and using it for sort of agentic capabilities… And that’s a gap that’s actually widening over time.” On the competitive stakes, he was blunt: “I think there is a risk of getting left behind.”

Access is not the differentiator, depth is

New Zealand is not a laggard on adoption. It sits in the top 40 markets globally for per capita ChatGPT use, and higher still for agentic tools. Nearly half of NZ enterprise output tokens came from agentic products – tools that carry out multi-step work like data analysis, research and testing ideas. Weeratunga said NZ businesses are “really over-indexing on agentic adoption.”

So what separates the top decile? Not the software. “Despite having access to the same capabilities, depth of usage is what differentiates firms,” Weeratunga said. “Agents need the context to understand the work, the tools to act, and the persistence to carry a task through to completion.”

The confetti problem

Datacom’s fourth annual AI Index, published in late August 2026, tells the same story from the other end. It surveyed more than 200 leaders at organisations with 100-plus staff and found 91% were using AI, but only 4% were using it to transform core operations.

Datacom’s Director of AI Lou Compagnone calls much of this “AI confetti”. Her diagnosis is sharp: “Organisations are adopting AI, but not adapting it to the organisation. How can we do things completely differently now, because of AI?” One structural tell: only 22% of NZ organisations have dedicated AI officers, against 42% of Australian businesses.

What the gap costs in dollars

The abstraction becomes concrete in the April 2026 Deloitte Access Economics report commissioned by 2degrees. It found the average SME using AI earned around $400,000 more in FY25 than a comparable non-adopter, while the average large business earned about $59.1 million more. The catch is in the fine print: the productivity lift only materialises where AI is genuinely embedded in operations, not bolted on.

And the spending is already flowing. Datacom found 79% of NZ businesses increased AI investment over the past year, but only a third see payback. Datacom’s NZ Managing Director Peter Nelson warned that as platforms shift to consumption-based pricing, knowing where AI creates measurable value “is likely to become increasingly important when making future investment decisions.” Spending without strategy is how you end up in the bottom 90%.

The honest caveat

Credit where due: OpenAI’s own researcher will not overclaim. Asked whether AI is actually lifting enterprise bottom lines, Weeratunga said “the revenue question is something that we’re also tracking… The question for large enterprises is still open and we’re tracking that very closely.” The output gap is real and measured. The dollar payoff at scale is still being proven.

This is a leadership problem, not a tech one

None of this is new terrain for New Zealand’s SMEs, which have long trailed Australia. When the government published its first AI strategy in July 2025 – making NZ the last OECD country to do so – it cited a 2024 survey showing 68% of NZ SMEs had no plans to evaluate AI, against 38% of Australian SMEs.

In a labour market where filled jobs sat at 2.26 million in the March 2026 quarter and grew just 0.1% on the previous quarter, output per worker is where competitive ground is won or lost. Weeratunga’s practical advice for closing the gap is unglamorous and cheap: model it “from the top down”, give teams time to experiment, appoint internal champions, run hackathons. The tools are already on your team’s desks. The only question left is which side of the 14x gap your firm ends up on, and the window to choose is closing.

Sources

• Kiwis send 10 million ChatGPT messages a day as work use climbs (2026-09-03)
• OpenAI says Kiwis use ChatGPT to save hours weekly (2026-09-03)
• AI adoption hits 91% of NZ businesses, but use to transform core operations falls to 4% (2026-08-28)
• AI in business: ‘How can we do things completely differently now?’ (2026-08-31)
• AI moves from hype to hard results for Kiwi businesses, 2degrees study reveals (2026-04-28)
• New Zealand’s strategy for artificial intelligence: Investing with confidence (2025-07)
• Business employment data: March 2026 quarter (2026-06-09)