For years the honest answer to "what is AI actually worth to my business?" was "it depends." New Zealand finally has a local number. Deloitte Access Economics, in research commissioned by 2degrees, estimates the average AI-adopting SME earned around $400,000 more in FY25 than a comparable non-adopter. For large businesses, the estimated gap was about $59 million.
What You Need to Know
- The first NZ-specific estimate of AI's firm-level impact. Published April 2026. This is local evidence, not an extrapolated global average.
- The average adopting SME earned ~$400,000 more in FY25 than a comparable non-adopter. Large businesses: ~$59.1 million more.
- 82% of NZ businesses report using AI, but most are using AI features inside tools they already own, not solutions built around their own data and workflows.
- The gap is widening, not closing. Larger businesses are moving faster than SMEs, and early adopters are compounding their advantage.
A Number You Can Take to the Board
The report, titled Productivity Propelled, is the first serious attempt to measure the relationship between AI adoption and firm-level productivity with New Zealand data. Deloitte surveyed businesses across the country and compared the financial performance of adopters against comparable non-adopters.
The headline finding matters because of who it comes from. This is not a vendor whitepaper. It's Deloitte Access Economics working from NZ survey data, commissioned by a telco with an SME customer base and no AI platform to sell.
And the finding is consistent with what we see in delivery: the businesses getting value from AI aren't the ones with the most tools. They're the ones that redesigned specific workflows around it and measured the result.
The 82% Problem
Here's the uncomfortable part. The same research shows 82% of New Zealand businesses now report using AI. If four in five businesses are "using AI" and only a fraction are seeing anything like the productivity gains above, then "using AI" is the wrong measure.
Most reported adoption is passive: Copilot in the Office suite, AI features switched on inside existing SaaS tools, individual staff using chat assistants. Useful, real, and nowhere near the ceiling. The firms in the top of the distribution did something different. They picked processes that matter to the P&L, put AI to work inside them with their own data, and tracked the before and after.
That's also why the average is worth reading carefully. An average of $400,000 across adopting SMEs means many earned far less, and some earned far more. The distribution is the story: value follows depth of adoption, not the fact of it.
What This Means If You're Deciding Right Now
If you haven't started: the cost of waiting now has a number attached. A year of "let's see how it plays out" is no longer neutral. Start with one measurable workflow, not a platform decision.
If you're at the 82% baseline: you're in the biggest cohort and the most exposed one. Tool-level AI is easy for competitors to replicate because they can buy the same tools tomorrow. Workflow-level AI, built around your data and your processes, is not.
If you're measuring nothing: you can't tell which cohort you're in. Deloitte could only produce this estimate because it compared measured outcomes. Do the same inside your own business: baseline the workflow before AI touches it, then measure again after. We've written before about how to measure AI ROI properly, and every part of that advice got more urgent this quarter.
The question boards were asking twelve months ago was "what's our AI risk?" The question this report puts on the table is sharper: what did not adopting cost us last year, and what will it cost us next year?