Ask most leaders to describe how their team feels about AI and you’ll get two groups. The keen ones already using it, and the ones holding out. The plan that follows usually is too: give the keen ones more tools, and work on the holdouts.
One NZ’s research suggests that picture is wrong in a way that matters.
Using it and worried about it
In March 2025, Perceptive surveyed 1,000 New Zealand adults for One NZ’s AI Trust Report. 77% said they had knowingly used an AI-powered service from a business or organisation in the past year. Asked what concerned them about businesses using AI, 65% picked the loss of human jobs.
Those two numbers come from the same 1,000 people. 77 plus 65 is 142. There’s only room for 100, so at least 42 in every 100 people both used AI and named job losses as a worry. The real overlap could be higher. The report doesn’t publish it, so 42% is the floor, not the figure.
Even at the floor, that’s plenty of people who don’t fit either camp. They’re using it, and they’re uneasy about where it leads.
The biggest group is still finding its feet
One NZ ran the survey again in March 2026. Use held steady at 76%. The job-loss question wasn’t repeated, but the report sorted people into three groups:
- AI Actives, 29%: confident, frequent users at home and at work.
- AI Dabblers, 48%: have tried it, use it irregularly, and don’t feel confident yet.
- AI Avoiders, 24%: little recent use, low trust.
The largest group isn’t the enthusiasts or the hold-outs. It’s the people in the middle who’ve had a go and aren’t sure yet. One NZ calls them the persuadable middle.
What that changes inside a business
If you picture your team as adopters and resisters, two mistakes come easily.
The first is reading use as comfort. The person who has ChatGPT open all day may be the one most worried about what it means for their role. They’re just not saying so. Using the tool and trusting where it’s heading are different things.
The second is reading hesitation as refusal. Someone who hasn’t picked it up may have tried it once on the wrong task, got a poor answer, and moved on. Nobody showed them how it fits the work they actually do. That isn’t resistance. It’s a fair response to an experience that didn’t help.
Neither mistake is fixed by more enthusiasm from the top. What helps is quieter:
- Answer the job question out loud, before anyone asks. If 65% of the public name it as a worry, it’s in your building too. Say what AI is for in your business, and what it isn’t for.
- Don’t sort people by how much they use it. Ask each person what part of their week they’d most like back. Start there, whatever their level.
- Let people go at their own pace. The dabblers are the biggest group because confidence takes repetition. One good result in their own work does more than a demo.
- Treat the heavy users as people too, not just champions. Ask them what worries them. You may hear the most useful answer in the room.
The caveats
This is a survey of the public about how businesses use AI, not a survey of staff about their own jobs. The 77% counts anyone who knowingly used an AI-powered service, including customer-service chatbots and voice assistants, not only work tools. The 77% and 65% are from March 2025, and the 2026 edition didn’t repeat the job question. The 42% is a minimum I’ve worked out from the two published numbers, not something One NZ reported. And One NZ sells AI and connectivity services to businesses, so it has an interest in the topic.
Declaring an interest
Helping businesses bring AI in one person at a time, at each person’s pace, is what we do at Coreshift. Weigh my conclusion with that in mind.
But the point doesn’t depend on us. The people in your business who use AI and the people who worry about it are, for a lot of them, the same people. Any plan that treats them as two groups will miss a lot of your team.