A tech expert says there are signs showing of AI's impacts in the country's high youth unemployment figures particularly around new graduates struggling to find jobs.

And there are a whole stack of other reasons a big new report out Friday gives for stopping with the ad hoc approach to AI.

On the streets of Wellington some people were worried AI would take their job, some weren't:

• "I'm not concerned about my job because we actually use AI, which I'm quite enjoying."
• "Personally, no, but other people for sure."
• "I work in a medical practice as a receptionist. It's not going to take long till they put a wee robot in there."
• "It's catch-22. I guess there's some big productivity improvements and there are going to be some hits."

But however you might feel about AI, overall the country's doing it wrong, says the Human Rights Commission in a landmark new report out Friday.

Senior advisor Sophie Bradwell-Pollak helped produce the report.

"Ultimately New Zealand has a very light ... approach that's not based on human rights," she said.

"Because it's looking at one part - so cutting costs and looking at AI in terms of efficiency and investment. [That] is only one small part of what we need to be looking at."

The report's 100-plus pages argue the top priority is to establish an overarching AI infrastructure strategy anchored in human rights and Te Tiriti o Waitangi.

It calls too for a new AI safety institute running a system for monitoring wins and harms from the machines.

On jobs, the country could just look sideways, said Bradwell-Pollak.

"Australia for example, the federal government a month ago released a report looking at what jobs are most susceptible to AI interference or AI impacts and that would be a great starting point."

No one country led the way on a human-rights-based approach to AI, she said, but many were doing more than New Zealand.

Futurist Frances Valintine agreed the one-track mind on AI just isn't cutting it.

"We are currently looking almost like through the lens of being afraid of missing out," said the founder and chair of academyEX.

Agentic AI - models that do not just generate content but go out and do jobs and subcontract jobs to other AI agents - are currently setting things up for the arrival of the equivalent four million new highly intelligent "workers". But without the country thinking through the broad consequences.

"We're sort of jumping in boots and all without sort of taking that helicopter view," said Valintine.

She's just been speaking to 800 tech business leaders at the Revved summit in Auckland. One message there - and on Morning Report on Thursday - was that New Zealand must lean in to AI.

But Valintine said the country must make its own path, like it did on going nuclear-free, and choose to partner with AI companies that aligned with NZ values.

"I've had many conversations since I spoke ... people are just saying, they're all measuring it on a cost of a licence, how many tokens [how AI use is paid for] they're using and how much productivity gains they're making.

"And I've had so many people saying, 'I've suddenly realised ... we're not thinking about the consequences of how do we get middle-tier managers if we're not hiring graduates at the same level anymore? How are we going to retain people if people are feeling unsafe and insecure in their jobs?'"

There are early signs of AI impacts in this week's persistent high youth unemployment figure, she added - and maybe a 2-to-3 year window to get the fundamentals right - so this '"absolutely" should make AI an election issue.

Bradwell-Pollak agreed it was not too late for the political parties to step up.

"They're still coming out with their policies, we're still meeting and engaging with MPs.

"So it's not too late. This is a topic that people are concerned with now. And also, you know, the [report's] recommendations will endure past this election and hopefully with the new government."

The Human Rights Commission said Friday's report was deep and technical with lots of expert input - but a simple signal of much more public engagement on AI to come.