AI in small business: what's actually happening in NZ
Where adoption sits, what it's used for and what's holding it back.
The headline numbers
Every survey asks a slightly different question, so the "how many NZ businesses use AI" number moves depending on who's counting, and how loosely "using AI" is defined.
of NZ SMEs proactively using AI, up from 32% last year
of SMBs say they're proactively using AI tools
of businesses use AI in some form, even just bundled features
of EMA member SMEs already using AI
The honest read: somewhere between a third and four in five NZ small businesses have touched AI, depending on whether "using AI" means running ChatGPT deliberately or just having Copilot switched on inside software they already pay for. The gap between those two things is the real story, covered below.
Where sole operators buck the adoption gap
Adoption climbs steadily with headcount, which is what you'd expect. The exception is sole operators: they're more likely to use AI than small employing teams, because there's no one else around to hand the admin to.
Dependency scales with size too. 92% of 20+ staff businesses said they'd feel workload or cost impacts if AI use stopped, against 49% of 1–5 staff businesses and 64% of sole operators. Translation: if you're running this alone, AI has already become part of how you get through a week, even if you've never called it a "workflow."
MYOB Business Monitor, reported by IT Brief, Aug 2026
What they're using it for
Across NZ SMEs, AI use is concentrated almost entirely in content and admin, not core operations. If you sell content, copy or admin support into this market, you're not introducing your prospects to AI. They've already been trained by ChatGPT and Copilot to expect it in exactly this lane.
MYOB Business Monitor, reported by IT Brief, Aug 2026
Which of Xero's four types are you?
Xero splits Kiwi SME owners into four mindsets. Worth a minute to place yourself. It changes what a sensible next step looks like.
Bolting AI onto specific tasks, self-taught, motivated by saving time. The largest group by far. 58% expect to use it more in the next year.
Not convinced it's relevant to them. Worried about data privacy and accuracy. 89% have no near-term plans to invest.
Early adopters, using AI extensively: 76% report higher productivity, 71% time savings, 47% better decisions.
Believe it could help, but time, not interest, is what's stopping them. 55% name time as the specific barrier.
Xero, “From curious to confident: how Kiwi business owners are adopting AI”, May 2026
Switched on, not used well
This is the gap that matters most if you're small: NZ businesses have adopted AI faster than they've built anything durable with it. CAIRN's AI maturity data shows the pattern clearly: the parts of adoption that come free score fine, the parts that take deliberate work don't.
Everyday AI use by staff scores 3.4 out of 5, the highest signal in the whole assessment. Training and capability scores 2.0, the lowest and most common gap.
Only an estimated 2.7% of the NZ workforce have AI embedded in how they work. The rest, even inside the "82% now use AI" headline, are using it superficially, or not at all.
“When a business tells a survey it uses AI, more often than not it means someone on the team has started leaning on the AI features baked into tools they already owned. Almost everyone has switched it on. Far fewer have changed how they work because of it.”
Only 13% of EMA member SMEs that use AI have an actual AI policy. Two-thirds are still in the experimentation or pilot stage. Only a third have embedded it into a regular process.
Two numbers worth sitting with
Alongside the headline adoption data, two figures from this research pass are worth knowing on their own.
more revenue, on average, for NZ SME AI adopters in FY25 vs. comparable non-adopters
of SMEs say AI has helped maintain margins; 19% say margins improved
Treat the $400k figure with some caution before quoting it to a client. It's self-reported, doesn't control for the fact that businesses that adopt AI early tend to already be better run, and FY25 was the year AI tooling became table stakes anyway. Still, expect prospects and competitors to start quoting it at you, so it's worth knowing where it comes from. The margin numbers are more modest for smaller operators specifically. For businesses with 20+ staff, 55% report maintained margins and 23% improved, roughly double the SME-wide figure. The smaller you are, the less likely the gains have shown up on your bottom line yet.
On support: Xero and ASB now offer a fully funded 12-week AI bootcamp for NZ SMEs, delivered by academyEX, a free way to move forward if training is your actual blocker. Separately, on 11 August the opposition Labour Party proposed a $2m/year AI small business fund and an expanded AI Advisory pilot. Worth flagging as a policy pitch rather than a live programme: nothing to act on yet, just one to watch.
What's holding the rest back
of non-users cite a lack of expertise as the main reason
cite data privacy and security concerns
don't trust the quality of AI outputs
of Pragmatists say time, not interest, is the barrier
Support asks are consistent across surveys, too: practical training (47%), access to trusted, pre-vetted tools built for small business (46%), and real-world case studies of other owners doing it (43%). Practical support beats more hype or another tool list.
What this means if you're running this alone
At 36% proactive adoption and 64% dependency among sole operators, most solo businesses have already dabbled: Copilot in Office, ChatGPT for a social caption, AI baked into the accounting software. "Hasn't started using AI" is a rare signal now. The sharper question is structure: no policy, no redesigned workflow, nothing that would break if it stopped tomorrow.
Content and admin is still the easiest way in. 38% social/marketing content, 28% copywriting: that's where AI already earns its keep for small teams. You don't need to be sold on the concept of AI. The question is just whether it's saving you time, or quietly adding a new admin task of its own.
"One workflow, done properly" beats "more AI." The 3.4-vs-2.0 gap in CAIRN's data is the whole argument: everyday use is fine, training and embedding isn't. Pick one real workflow, quoting, follow-ups, reporting, and rebuild it with AI designed in. That beats bolting on another tool.
Doing that properly for one real workflow is exactly the AI Operations work I do for clients.
Trust and privacy concerns are real. 40% cite data privacy, 37% distrust output quality. If those are your reservations too, you're in good company. Worth addressing them directly rather than being talked out of them by generic AI-optimism.
Expect the $400k figure to come up in conversation now too. Know the caveats above so you can respond to it honestly: neither dismissing it outright nor treating it as a guarantee for your own business.