Somewhere in your business this week, someone finished in twenty minutes a job that used to take them most of a morning.
You will never hear about it.
New Zealand research published in July — Employment Hero surveyed 256 business leaders and 820 employees across businesses from five staff to 1,500 — found that 28% of workers are using AI tools their employer doesn’t know about, and 32% present AI-generated work as entirely their own.
Then there’s the finding that I haven’t been able to put down. Even the workers most competent with AI aren’t spared: 41% of them feel replaceable, barely different from everyone else. Getting better at this doesn’t make the fear go away. It just gives someone more to hide.
Sit with what that does to your visibility.
This is not a morale problem
Every business owner I talk to is asking some version of the same question: where does AI actually create value in my business? It’s the right question. Datacom’s 2026 AI Index found 91% of New Zealand organisations using AI in some form, but only 4% using it to transform core operations. Previously it was 8%. It halved.
The usual explanations are that the technology’s immature, or the strategy’s missing, or people need training. All of those might be true. But I think something simpler happens first, and it’s upstream of all of them.
You cannot find your best AI use cases, because the people who found them have a personal incentive not to show you.
That’s not a soft problem sitting off to the side of the ROI problem. It is the ROI problem. Fear isn’t costing you morale. It’s costing you measurement.
What the hiding actually costs
Four things, none of them abstract.
You can’t reuse anything. One person saves three hours a week and thirty-nine people don’t, because nobody knows. The whole economic argument for AI in a business is that a good method spreads. Concealment is the thing that stops it spreading.
You can’t govern what you can’t see. You don’t know which tool, which account, or what went into it. Every organisation worrying about AI risk is worrying about the use it knows about.
You’re investing on assumptions. Buying more licences, or fewer, based on a picture of your own business that your own team is editing.
Your pilot studies the wrong people. Run a trial and you’ll learn about the volunteers — the confident ones who were never hiding anything. The behaviour you most need to understand is, by definition, absent from your sample.
Why they hide
Three fears, and none of them are silly.
Will I be replaced? Forty-two per cent of NZ workers in that survey believe heavy AI use signals they’re no longer needed.
Am I training my own replacement? If the work I do gets easier to automate every time I do it well, what exactly am I building?
Is this cheating? Thirty-eight per cent say using AI to complete part of their job feels like cheating. Thirty-seven per cent feel guilty doing work they know is better than what they’d have produced alone.
Now add the context. Only 47% of New Zealand employers encourage their staff to use AI at all. So in more than half of workplaces, people are operating in a vacuum, against a backdrop of international headlines about AI and jobs, and filling that silence with the worst available interpretation.
They’re not being paranoid. They’re reading the room accurately. The room has said nothing.
A policy won’t fix it
The instinct is to write an AI policy. Most policies answer the question “what am I allowed to do?” That is not the question being asked. The question being asked is “what happens to me?”
Until that one has a credible answer, disclosure doesn’t follow — and the order matters. You don’t get visibility and then reassure people. You answer the real question, in a way people believe, and visibility follows as a consequence.
Credible is the operative word. “AI won’t replace anyone here” is not credible if nobody can say what it will do. What people can actually believe is something more specific: here’s what we’re using it for, here’s what we’re not, here’s what happens to the time it frees up, and here’s who decides.
The honest version
I’m not going to pretend one all-hands fixes this, or that saying the right words makes a 58-year-old in dealer support stop worrying about her job. Trust gets rebuilt by what happens over months, not by an announcement.
But I’d put it this way. If you’re trying to work out where AI belongs in your business, and you haven’t yet made it safe for someone to say “I used AI for this” — you are doing archaeology on a site your own team is quietly filling back in.