Trust in artificial intelligence in New Zealand is eroding as the technology becomes embedded in everyday life.
That makes responsible, transparent deployment of AI a critical issue for every organisation, from telcos to government agencies.
The 2026 One NZ AI Trust Report*, based on a nationally representative sample of 1001 Kiwis, paints a picture of a country that has moved past the novelty phase with AI and is now actively judging how it is used.
While 76% of New Zealanders say they’ve interacted with an AI‑powered service in the last 12 months, trust in organisations to use AI ethically and responsibly has fallen across all major sectors.
Fewer than two in five Kiwis now believe AI has the potential to deliver better outcomes for society, down from 43% last year.
If leaders were hoping that familiarity would naturally breed trust, the report serves up a reality check.
One NZ’s research identifies a 29% ‘AI actives’ segment comprising confident, frequent users, alongside a large middle of ‘dabblers’ and a stubborn quarter of ‘avoiders’ who still want little to do with AI.
When the computer says ‘no’
Crucially, those active users are as likely as anyone else to walk away if they are uneasy about how AI is used.
Seven in ten AI users say they’ve experienced at least one problem with AI‑powered services in the past year – from incorrect information and clunky chatbots to difficulty reaching a human and privacy worries.
“Just because you’re experiencing AI, doesn’t necessarily mean that you’re going to trust it more,” said Summer Collins, One NZ’s chief AI & business services officer, launching the report at Parliament in May.
“The more you know about the technology, the more you question.”
Collins argues that concerns around AI are not a perception problem, but one of poor implementation.
Badly designed systems that don’t understand the customer, don’t hand off cleanly to a human, or feel cold and impersonal are actively teaching people to distrust AI – and not just that particular deployment.
“This technology is not easy to implement, and it’s not deterministic,” Collins says. “It behaves in ways that you don’t predict. It’s not like a normal technology implementation.”
With AI evolving from chatbots that serve up information in mere seconds, to agentic systems that automate tasks and make decisions on your behalf, eroding trust in AI signals a looming problem for businesses – and for the Government, as it embarks on an AI-driven transformation of the public service.
The research shows that 62% of Kiwis say they would stop using a product or service if they had concerns about how the organisation was using AI. Trust in multinationals’ use of AI is especially weak.
Public comfort with AI varies by sector. It is highest in retail and telecommunications, but much lower in healthcare, finance, banking and government – precisely the areas where AI could have the most impact, but where legitimacy will be hardest to earn.
Human oversight is non‑negotiable
The key trust‑building action that needs a greater focus is providing evidence of human oversight, with 68% of people saying having a clear option to talk to a human would make them more comfortable with AI‑powered customer interactions.
At One NZ, where over 1000 staff have access to enterprise-level AI tools, the technology is used in everything from customer service and marketing to network management and board governance.
One NZ has adopted a simple internal principle: “We’ve got a philosophy which is think AI first but really be deliberate where humans matter most.”
In practical terms, that means designing escalation paths and ensuring that when a conversation hands off from a bot to an agent, the context flows with it so the human can fix the problem quickly, rather than forcing the customer to start again.
One NZ’s AI Hub
One NZ is trying to show what an AI-enabled business can look like. The company has spent more than a decade deploying robotic process automation and, more recently, large language models across the company – but with what Collins stresses is a deliberate focus on culture and guardrails, rather than pure speed.
Central to that has been company-wide training to get staff hands‑on with AI tools, paired with explicit acknowledgement that job fears are “real and rational”, and a commitment to help people build the skills they need as work changes.
Collins set up a centre of AI excellence that offers advice and guidance to every area of the business. One NZ has also drawn a line in the sand on high‑stakes issues, such as employment and credit decisions, where full autonomy for AI is off the table.
The launch of the One NZ AI Trust Hub is also hard‑wiring transparency into how the telco uses AI.
“It is sharing transparently how we’re using AI throughout the business, and for what reason,” Collins explains.
One NZ’s AI governance approach also bakes in technical and commercial safeguards, insisting that customer data remains under One NZ’s control. It has designed systems so the data and customer interface are owned by One NZ, while preserving the flexibility over the AI models underneath.
The organisations that will help build trust in AI are not those that deploy it the fastest, but those that can move quickly while proving where AI helps, where it stops and who is accountable when it misfires.