The increased use of technology and the uptake of artificial intelligence (AI), has long been touted by government as a way the public service can adapt to the country's tight fiscal needs and future proof itself to be financially sustainable. That view coming into clearer view this week after the announcement of operating cuts totalling $2.4 billion for the public service.
The Government expects that to come from mergers, utilising AI and technology and through reducing public servant jobs.
Finance Minister Nicola Willis said the public were using AI every day, "and, while parts of the public sector have seized the opportunity to innovate, others are still locked into outdated ways of doing things that prioritise box-ticking over outcomes."
Willis said that was not acceptable or sustainable so the Government would put a sinking lid on agencies’ operating budgets to; "drive progress on streamlining the number of government agencies and entities, digitising customer-facing and back-office government functions, and restoring public service numbers to historical norms."
But while plans are already underway to accelerate the use of AI in the public service, it is not clear how much this will actually save.
In February the Government chief digital officer Paul James said there were not a lot of data points available yet on how much AI would save New Zealand, and while it would absolutely drive down costs, it would take a while to get there.
Asked by Labour's Reuben Davidson what specific roles AI was expected to replace, Public Service Minister Paul Goldsmith this week said they would be working through this over the next three years. An AI workforce plan was also in process.
Asked if he had evidence showing AI being able to safely carry out public service roles at scale, Goldsmith said, "it depends on what particular roles one has in mind, but currently, already, AI is being used in certain circumstances across the public sector, as it’s used across the wider world."
'An overly simplistic view on AI-use'
Auckland University senior law lecturer Joshua Yuvaraj said the announcement to streamline public services using AI suggested; "an overly simplistic view on AI use without a full appreciation of the sheer scope of work that government departments do, and the different risk levels that different government departments may have."
"When you're saying that this technology is reliable enough for us to downsize the public service by a significant number, then I think that's an extraordinary claim that needs extraordinary supporting evidence of efficacy and accuracy, which the empirical evidence just doesn't suggest that that's the case as far as I'm aware."
Yuvaraj said not enough attention had been paid to the potential for things to go wrong, and whether that had been factored into the plan to significantly downsize government departments in favour of integrating AI.
"One of the main issues with the AI tools that seem to be mooted is the issue of accuracy. If generative AI tools are used, one of the major problems is they are not tethered to reality. They have no conception of right or wrong. They have no conception of true or false. What they have is a conception of what they are fed in their training data right now."
Yuvaraj asked if a mistake is made, "how is that going to be handled? Who is going to be liable, and how are we going to apportion blame? How is the public going to feel?"
"If you're asking an AI tool to approximate what we've typically allocated to human judgment, you've got to expect that there are going to be mistakes, there are going to be inaccuracies."
"The retort will be, well, humans make mistakes, that's true, but humans exist in a sort of pyramid of accountability, and employment accountability, even accountability of criminal law, et cetera."
Yuvaraj said it was very easy to say AI would save time, "but again, the the body of evidence on that is mixed."
24 Comments
Starbucks has pulled back its AI‑driven “Automated Counting” inventory‑tracking tool from all North American stores after roughly nine months, citing persistent accuracy problems and now reverting to manual counts for milk and beverage components.
https://www.cnbc.com/2026/05/21/starbucks-scraps-ai-inventory-tool-acro…
Retired after a well paid career as a computer programmer. never forgotten what one of the rare good bosses said: "Never computerise a good manual system".
DENSO invented the QR code in the early 1990s, and the initial use case was inventory and production tracking for auto parts in manufacturing plants. DENSO described the QR code as the result of “two years of trial and error.”
The West was very late in discovering QR and its application.
AFR
‘Replace low-value human capital’: CEO’s stinker nails AI’s ugly truth
His brutal justification for massive AI jobs cuts will harden opposition to the technology. It also reminds us what the AI investment boom is really all about.
It’s graduation season across American universities and one of the grand traditions in these ceremonies is the commencement speech, where a distinguished speaker delivers some life advice to the young adults setting out to conquer the world.
The University of Arizona seemingly landed a big name for its speech: former Google chief executive Eric Schmidt. But Schmidt was loudly booed when he declared the coming tech revolution that artificial intelligence would usher in would be “larger, faster and more consequential than what came before”.
It’s apparently been a bit of a trend at these graduation ceremonies; when speakers mention AI, they cop the ’ol Bronx cheer. And it’s not just the kids.
Community anger about power and water-hungry data centres is building across America, amid growing concerns about affordability and potential job losses. Opinion polls suggest views on AI are deeply divided.
on Tuesday night when the CEO of one of the world’s largest banks dropped a bombshell.
Bill Winters, who has run the $77 billion emerging market-focused bank Standard Chartered for 14 years, announced a plan to slash up to 8000 jobs between by 2030 by deploying AI.
There will be job “reductions in favour of machines, and that will accelerate as we go forward into AI”, Winters told a press briefing.
“It’s not cost-cutting. It’s replacing, in some cases, lower-value human capital with the financial capital and investment capital we’re putting in.”
The backlash against AI is building across the world, and Standard Chartered bank CEO Bill Winters may have turbocharged it. David Rowe
Read that sentence again. It’s bad enough that you’re cutting jobs, and that your staff will spend the next four years wondering if they are next; Winters said cuts were likely in areas such as human resources, risk and compliance.
But to reduce them to the level of “lower-value human capital” is a gaffe we’d bet Winters will struggle to live down.
Standard Charted shares fell 2.2 per cent after Winters’ big reveal, although the stock is up 62 per cent in the past 12 months.
Regardless, Chanticleer can’t help thinking we owe Winters a little debt of gratitude for saying the quiet part out loud. Banking, which is highly driven by rules, processes and data, has always seemed like the best target for AI deployment, and it’s certainly where the likes of OpenAI and Anthropic have been aiming.
The smoke signals coming out of the Australian banking sector this year suggest the big four are on the cusp of a StanChart moment.
None of our big CEOs would be dumb enough to use the brutal language Winters used, but they are slowly preparing to move from relatively narrow experimentation with AI (mainly in areas such as coding and software development) to broader deployment.
Commonwealth Bank’s recent establishment of a project to map the tasks within the organisation, and create a platform for training and career development, is a perfect example of the preparatory work under way.
In many ways, cutting costs is the great hope of investors holding on to their overvalued bank shares.
And let’s face it, the banks have little choice. Even before US President Donald Trump went to war with Iran and the federal budget changes put the brakes on property investors – which have been running at as much as 40 per cent of home loan flows in recent months – the big banks were struggling in an environment of persistent cost inflation and tepid growth.
That environment has got a lot worse, in very short order, as Morgan Stanley banking analyst Richard Wiles argues.
“Aside from COVID, we cannot recall a time in the past 25 years when the operating conditions for banks have shifted so quickly,” he says. “Three RBA rate hikes, proposed changes to property-related tax concessions in the federal budget, and the potential direct and indirect effects of the global energy shock have created a far more uncertain outlook for the Australian banks.”
In many ways, cutting costs is the great hope of investors holding on to their overvalued bank shares. And AI looms as a crucial tool in that task.
While Winters has captured the great fear of all workers – that one day we might all be lumped into his category of low-value human capital – he has also explained the growing sense of greed driving the rise and rise of any stocks connected to the great AI buildout.
The latest Bank of America global fund manager survey released on Tuesday night showed a record rise in investors’ allocation to stocks in the course of a month, to levels not seen since the uber-frothy markets of early 2022. Why? Because everyone is chasing the most crowded trade in the world: chip stocks specifically, and anything to do with the AI infrastructure buildout more broadly.
Microsoft is canceling or winding down most internal Claude Code licenses and shifting engineers to GitHub Copilot. Too expensive apparently.
https://www.windowscentral.com/microsoft/microsoft-cancels-claude-code-…
OMG they will puke with CLI vs Claude
Replace low-value human capital
This is what almost every technology has done throughout history. Hard to know if AI will be any different in terms of outcomes
It is the first to replace intelligence and it will do it in 3-5 years.
at the same time its replacing drivers in cars and long distance trucking.
The low value human capital guy was just caught saying the quiet part out loud.
Even without AI most large organisations treat their staff as disposable items.
And at the some time they parrot their corporate values (there's usually four of them), but will include something about One Team! Better Together! Etc.
AI will just put petrol on the flames
I use Claude and frankly it’s incredible however at this stage it’s a tool so you still need input at both ends of the process. But I’d say it’s a no brainer that any institution can get a huge efficiency benefit out of it if used for the right things.
No doubt. However, as Microsoft shows, token-based pricing is forcing every enterprise customer to confront the actual cost of running these models at scale. The prices you are seeing for AI will probably not last too long.
For many they will go up 20x at the end of this month, some of us have been smashing tokens... to build skills while they have been cheap, get you hooked then charge you heaps
Yep basically loss leading. The big players have cemented their dominance and now they want to cash in.
Ask a CEO, do you want me to produce 5 times more or produce the same work for 20% of the current cost....
Uber’s CTO said the company effectively exhausted its entire 2026 AI budget within the first few months of the year, largely due to rapid adoption of Claude Code.
https://theagenttimes.com/articles/uber-burned-through-its-entire-annua…
Just make sure you are not in the lower value human capital bucket.
That’s always been a good idea
how much , on average, do you think these poor lower average people get paid at Standard Chartered?
it might surprise you, have you ever worked at an investment bank?
Income distribution at a global bank like Standard Chartered
follows a distinct pyramid structure, where a broad base of analysts and associates earn entry to mid-level salaries, while a significantly smaller proportion of executives, managing directors, and senior specialists account for the highest tier of compensation
.
The Income Pyramid
- Entry to Mid-Level (Analysts & Associates):
- Income Range: \(\$50,000 - \$150,000\) USD.
- Distribution:
This layer makes up the largest segment of the workforce (often the lower and lower-middle pay quartiles). It typically consists of early-career professionals, junior engineers, and operations staff.
- Upper-Mid Management (Vice Presidents, Directors, & Senior Managers):
- Income Range: \(\$150,000 - \$300,000\) USD.
- Distribution:
Makes up roughly \(25-35\%\) of the workforce. Income is bolstered significantly by performance bonuses and deferred shares.
- Executive Leadership (Executive Directors & Managing Directors):
- Income Range: \(\$300,000\) to well over \(\$1,000,000+\) USD.
- Distribution:
The top \(1\%\) to \(5\%\) of employees. These roles are typically located in major financial hubs like London, Singapore, Hong Kong, and New York.
AI has likely made it possible for me to work past retirement age in IT. I use it daily. The other day I quipped, "AI means never having to ask your colleagues for advice".
Staff who leave are very reluctantly, if ever, replaced, because their skills can be replaced by AI tools. Paradoxically this means staff are implored not to retire because AI can bolster the dwindling faculties of aging tech workers keeping them functional. Managers may find it strategic to keep an older staff member rather than go through the hoops of applying to hire a replacement only have it to denied by upper management who believe in the power of AI.
I use a.i for design ideas development. Sometimes it's helpful, other times wastes hours and hours of my time, especially if I wasn't particularly analytical and persistent in testing it with further questions to which it often answers beginning with 'confessions' that it has made a mistake or mistakes. I could have spent and wasted a lot of time and money if I unquestionably followed its 'advice.' Despite its ability to trawl most of the internet it omits finding in some cases better solutions to problems and turns up what I suspect is links and recommendations centered around promoting sales opportunities for its better paying commercial customers, advertising in disguise with a cloak of logic and infallibility.
I reckon widespread implementation too quickly in the public service will lead to shocking errors because as quoted in the article 'one of the major problems is they are not tethered to reality. They have no conception of right or wrong. They have no conception of true or false.'
What models are you guys using ?
Perplexity and Gemini
Joe Rogan pumps perplexity I must try it
If it's applied piecemeal and indiscriminately without examining what's being done and why, and the staff losses are last-on first-off rather than genuinely supernumerary, then it's not going to do much good. It's also likely that those who can leave will, leaving...?
We welcome your comments below. If you are not already registered, please register to comment
Remember we welcome robust, respectful and insightful debate. We don't welcome abusive or defamatory comments and will de-register those repeatedly making such comments. Our current comment policy is here.