Skip to main content
Advanced search
The Point - home
DonateSubscribe
HomeWhat's NewOpinionOff the ChartsExplainersFactcheckAbout Us
The Point — home
DonateSubscribe
The Point — home
Advanced search
  • Home
  • What's New
  • Opinion
  • Off the Charts
  • Explainers
  • Factcheck
  • About Us
DonateSubscribe
Subscribe
  • Home
  • What's New
  • Opinion
  • Off the Charts
  • Explainers
  • Factcheck
  • About Us
The Point - home
Privacy PolicySitemapComplaints & CorrectionsNewsletter

The Point recognises the ancestral connections and custodianship of Traditional Owners throughout Australia. We pay respect to Aboriginal and Torres Strait Islander cultures and to Elders past and present.

©2025–2026 The Point, an initiative of The Australia Institute

OPINION

Productivity, New Technologies, Working Time

This is an edited version of a speech delivered by Greg Jericho (Chief Economist at The Australia Institute) at the Academy for Social Science in Australia Annual Symposium. 

Greg Jericho

Greg Jericho

Tue 17 Nov 2026 09.40 AEDT

Economy
Productivity, New Technologies, Working Time

Photo: AAP Image/Dean Lewins

To paraphrase Paul Keating, if you walk into any pet shop in Australia, what the resident galah will be talking about is AI. And if the galah likes economics, they no doubt will be talking about AI and productivity.

We know AI most likely will lead to improved productivity measured by GDP or GVA (gross value added) per hour worked. It has to, surely? If not, then what are we even doing building all these data centres that the Reserve Bank keeps telling us is the main driver of excess demand in the economy?

For the current enormous investment in AI to make any economic or commercial sense, AI must produce productivity benefits, and assuredly most of it will be done by replacing labour. While some output will be increased by AI assisting workers, if that is all it does, the benefits are nowhere near enough to warrant the investment that we are currently seeing.

A 500% increase in the annual value of building commencements in the AI sector in two years isn’t just to help people already being paid to do work.

No, it must replace labour costs, and increase profits, or it’s a dud.

Now fine – that will look good in the productivity statistics. And economist will then hold Paul Krugman close to their hearts and think that’s a good thing great because as he famously said, “Productivity isn’t everything, but in the long run it is almost everything”.

But of course, in the long run, we are also all dead.

I say this because not only is Krugman always invoked when talking about productivity, but Keynes as well hovers over us.

His words from his 1930 essay “Economic Possibilities for our Grandchildren”, where he wrote of “the new disease of which some readers may not yet have heard the name, but of which they will hear a great deal in the years to come—namely, technological unemployment” haunt the debate over AI.

Right now, workers in Qantas, NAB and others are experiencing it.

And although Keynes didn’t actually argue in that essay that productivity would mean we would only work a 15-hour week, he did however suggest we would probably only need to work that much to satisfy our desires to work.

Of course, what happened is our wants (and what we perceived to be our needs) grew with productivity, and we kept working to pay for them – even if we now do actually work less than the 44-hour standard week of 1930.

But if you dismiss Keynes as a dreamer, be not worried, for Jeff Bezos has arrived to keep the dream going. Earlier this year, he told Fox News that “AI’s productivity gains could let some people decide, “I can support my family by working three days a week.”

Yes, the man who times his workers’ toilet breaks, who uses AI to track union activity among his workers, is predicting a world in which AI will deliver such wealth that there will be free toilet breaks for all!

Spare me.

We know this won’t happen in the same way we know AI will not mostly be used in the future to cure cancer. AI will inevitably be used to increase profits through exploitation of workers and customers, and for military purposes.

And who will benefit?

Back in 2015, when Paul Krugman was in Australia for the Festival of Dangerous Ideas, I interviewed him for Guardian Australia and asked him about his famous productivity aphorism. He still stood by it, but that, when he wrote it:

I didn’t realise just how much rising inequality could cause living standards of ordinary workers to lag behind productivity growth. I had no notion of just how big a deal that could be.

And it is a big deal.

In the 20 years before I entered the labour force in 1991, productivity grew 43%, while real hourly wages grew 41%, since then productivity has risen 56% while real hourly earnings have gone up 42%.

So, we should be sceptical when Jeff Bezos suggests that “It’s going to actually be difficult to hire people because the productivity in the economy is going to make it so that if people don’t want to work two jobs, they won’t have to.”

We know that will not happen unless the benefits of AI accrue to workers. And that will not happen unless workers are able to bargain collectively in an industrial relations system that is not weighted in favour of corporate interests – things our polity actively works against.

But all of this assumes that productivity is this very essential thing. And maybe it is.

It is good if we can manufacture and construct more things with the same hours of labour.

Get our newsletter

Quality research, analysis, explainers and factchecks from experts you can trust.

And maybe AI will improve productivity in manufacturing, construction, mining and agriculture (assuming the climate crisis doesn’t destroy it).

But the problem with our devotion to productivity is that it leaves us pretending the economy remains largely made of manufacturing and is easy to measure, and we essential in industries we ourselves don’t even seek for it.

So absolutely it is taken as given that productivity is the be all and end all that last year on an ABC Radio here in Melbourne, I mentioned that Victoria has a higher employment rate per population than NSW, and the host told me, “Yes, but they’re not productive jobs”.

But does anyone choose the most productive childcare for their children? When you one day go to an aged care centre, will you seek out the most productive one? Do we care that, as the Productivity Commission noted a decade ago, artisanal bread production is hurting the productivity figures?

Does the Productivity Commission even really know how to measure its own productivity? Is it number of research papers and presentations? Or is the purpose of the Productivity Commission to actually change policy? Should we measure its productivity by policy changes per hour worked?

I say this as someone who has no idea how to measure my own productivity. Sometimes I produce papers that take two months to write that only the Fair Work Commission reads; sometimes I do a 50 second video that that 300,000 people view across social media. How do I quantify that output per hour worked without indulging in absurdities of measurement?

And yet we continue to suggest we can measure productivity across each industry and the economy with wonderful accuracy.

And sure, there has been good work done by economists trying to measure the non-market and services sectors. But all involve subjectivity – do we include things like NAPLAN results to measure output of schools? I would argue no, but I have seen research doing so.

In reality, our debate over productivity has in my view mostly been one of misguided panic due to elevation of what used to just be a residual in an equation, but is now called multi-factor productivity, and ignoring the slumping productivity in the mining sector over the past decade that weighs down the market sector.

And worst of all, we include the non-market sector where greater productivity should not even be the aim – we should have more workers in care, more workers in education, more workers in health, more workers doing public sector tasks.

AI will likely help the service sector – it’s helpful for deep search when researching, for coding if you do R. But the aim of AI is to eventually do that research and not just assist. Just as AI solved Millennial Maths problems, will it one day write economics papers worthy of a Nobel?

Is that a good thing?

I guess it will leave economists with fewer days to work.

Will AI really help with Centrelink? If it is just AI chatbots, I doubt it. Will using AI to conduct hiring interviews deliver better employees for companies?

But with AI, like self-service checkouts, I doubt we will have any choice.

And yes, the productivity benefits will accrue in the data, but they won’t go to workers without a fight.

Does anyone think Coles and Woolworths employees or bank workers are all thinking about what to do with their extra money once the productivity benefits arrive? That is, the workers who are still around to get paid?

Am I a sceptic or pessimist? Most certainly.

Back in 1930, Keynes argued, “Of course there will still be many people with intense, unsatisfied purposiveness who will blindly pursue wealth-unless they can find some plausible substitute. But the rest of us will no longer be under any obligation to applaud and encourage them.”

But alas, we are under that obligation – our entire political economy has been set up to do so – profits are good, real wage growth is concerning, and the AFR still does a rich list, and no one thinks it odd.

We are told we need more gas exports because something sold by a foreign company whose revenue and profits flow overseas is counted towards our GDP, and the sector is exalted despite the fact that its wage bill is the same as the adult education sector.

But we also know from the input-output tables that around $88 of every $100 of output in the LNG sector goes overseas. And some research from The Australia Institute, soon to be published, suggests around $97 of every $100 of datacentre output will flow overseas.

Not much there to deliver the income for a 3- or 4-day week.

But I guess we will be able to say the GDP per hours worked figure looks good.

This is an edited version of a speech delivered by Greg Jericho (Chief Economist at The Australia Institute) at the Academy for Social Science in Australia Annual Symposium.

Related Articles

FACTCHECK

Factcheck: No, a gas export tax would not drive up gas prices for Australians

By Greg Jericho

This morning, Minister for Industry Senator Tim Ayres argued that a gas exports tax would mean higher prices for Australians. Except this is not true. A gas tax would actually lead to lower prices for Australians.

Economy
Not true
Factcheck: No, a gas export tax would not drive up gas prices for Australians

WHAT'S NEW

All gas, no consensus: Albanese bill faces Senate showdown

By Tegan George

The Albanese government is facing renewed calls to introduce a 25 per cent gas export tax as its gas reservation bill comes under pressure from all directions.

Economy
All gas, no consensus: Albanese bill faces Senate showdown

OPINION

From pilot to powerhouse: expanding WA’s sustainable bond market

By Brad Hiller & Brad Pettit

Now a staple of global finance, GSS bonds offer a powerful mechanism to fund pressing environmental and social initiatives that lie outside traditional government budgets.

Environment & BiodiversityEconomy
From pilot to powerhouse: expanding WA’s sustainable bond market

OFF THE CHARTS

If the RBA forecasts continue, real wages won’t recover to 2021 levels until 2044

By Greg Jericho

Last week, when the Reserve Bank announced it had raised the cash rate to 4.85%, it also released its latest forecasts for the economy in its Statement on Monetary Policy. For those hoping for a recovery in lost living standards, the news was very grim.

Economy
If the RBA forecasts continue, real wages won’t recover to 2021 levels until 2044