The cheque is running out of road. Australia needs a stake.

AI is quietly moving the returns of work from the people who do it to the people who own it. Cash transfers will not fix that. Ownership will. Here is why the shift from a cash economy to an ownership economy might be what keeps capitalism standing, and why we are starting in Australia.

For a hundred years the deal was simple. You traded your hours for money, and money bought you a life. Ownership was for other people: founders, investors, the occasional employee who got in early enough to hold options that meant something.

That deal is being renegotiated whether we like it or not. The renegotiation is global. Its consequences will be local. And the country that works out how to spread ownership widest, fastest, will be the one whose economy keeps renewing itself while others hollow out.

The signal is coming from the people building the thing.

In May 2025 Anthropic's chief executive Dario Amodei told Axios that AI could eliminate half of all entry-level white-collar jobs and push unemployment to between 10 and 20 per cent within one to five years. The sectors he named were finance, consulting, law and tech. His message to lawmakers was that most of them had no idea it was coming.

Eight months later at Davos he sharpened the point. AI's cognitive breadth means it lands on many industries at once, and that is what makes this different from every previous wave of automation. A factory hand displaced by a robot could move into an office. If AI is moving into the office at the same time, there is no obvious lane to switch into.

The IMF had already run the numbers a year earlier. Almost 40 per cent of jobs worldwide are exposed to AI, rising to around 60 per cent in advanced economies. Roughly half of the exposed roles may benefit. The other half may not. The line from Kristalina Georgieva that matters most for this essay is the one about where the gains go: in most scenarios AI worsens inequality, because productivity gains flow to high earners and to capital, widening the gap between the two.

50%

of entry-level white-collar roles at risk within five years. Amodei, May 2025.

60%

of jobs in advanced economies exposed to AI. IMF, January 2024.

40%

of employees globally fear losing their job to AI, up from 28% in 2024. Mercer, 2026.

Steven Bartlett has spent the last 18 months circling the same question on The Diary of a CEO. Sitting across from Yoshua Bengio in December he ran a thought experiment: put two versions of himself in the room, one with an IQ of 100, one with an IQ of 1,000 that never sleeps and never gets sick. Who do you hire to drive your kids to school, to teach them, to run your factory? He could not think of many jobs left for the first Steven.

His guests do not agree on timing. Scott Galloway thinks the "job apocalypse" is partly a fundraising story told by AI chief executives. Amjad Masad and Daniel Priestley think half the global workforce is exposed. But even the sceptics agree on the shape of it. The gains are concentrating, and ordinary people can feel it.

Australia is early, not immune.

Here is the honest local picture, because the honest version is more useful than the scary one.

In July 2026 the federal Department of Employment and Workplace Relations published its first monitoring report on AI and employment. It found no broad upheaval. Overall conditions remain strong, youth outcomes have mostly held, and occupational churn has not accelerated. But occupations most exposed to generative AI have grown more slowly than the rest since late 2022. The department called the result suggestive, not definitive, and an early indication of modest slowing in the most exposed roles.

Modest is the right word for the data. It is not the word for the mood. ServiceNow's 2025 survey found six in ten Australians fear losing their job to generative AI, the highest share of any country it measured. Australian enterprises had also fallen ten points in AI readiness in a year, with only a third holding a clear AI vision for their business. We are anxious and under-prepared at the same time, which is the worst combination.

6 in 10

Australians fear losing their job to generative AI, the highest of any country surveyed. ServiceNow, July 2025.

0.3%

average annual labour productivity growth in Australia over the past decade. EY, August 2026.

$116bn

AI could add to Australian economic activity over the next decade. Productivity Commission, August 2025.

Then there is the upside, which is real and which the loudest voices in Australian tech are pushing hard. Scott Farquhar, Atlassian co-founder and chair of the Tech Council of Australia, used his National Press Club address in July 2025 to argue that if Australian businesses simply adopted the AI tools that already exist, the productivity lift could be worth up to $115 billion a year. He wants Australia to be the data centre capital of the Asia-Pacific. He also, on the same day, oversaw Atlassian cutting 150 roles it said AI could do.

That is not hypocrisy. It is the whole story in one afternoon. The productivity is real, the displacement is real, and the two arrive together. Farquhar's own answer to the displacement was "tech trades" and apprenticeships, which is a fair response for the electricians and battery installers he named. It is not much of an answer for the 40-year-old finance manager whose team just went from eight people to three, or who now reviews the work of six AI agents instead, day in, day out.

The answer everyone reaches for, and why its biggest backer walked away from it.

There are two standard responses to all of this. Retrain the workers, and if that fails, pay them.

Retraining assumes there is somewhere to retrain into. That is exactly what Amodei's cognitive-breadth argument undermines.

Universal basic income is the bigger idea, and it has had its best-funded test. Sam Altman put $14 million of his own money into a $60 million OpenResearch study that gave 1,000 low-income Americans $1,000 a month for three years. The results were reassuring on the fear people had. Recipients kept working. They spent the money on food, rent and transport. They reported more agency, and some started businesses.

Then Altman changed his mind.

His reasoning is the hinge of this whole argument. A fixed cash payment, he said, is useful but does not get at what is needed as the balance between labour and capital shifts. What he wants instead is collective ownership, whether in equities, in compute or something else. OpenAI's own industrial policy paper proposes a Public Wealth Fund to give every citizen a stake in AI-driven growth.

Read that carefully. The person who funded the largest basic income experiment in the United States concluded that cash is a floor, not a future. If AI does the work and a handful of companies own the AI, a monthly cheque keeps you fed while the actual wealth compounds somewhere you are not.

When innovation stops being human-led, the fork appears.

The third pressure is quieter and probably the biggest. Both Altman and Amodei have written about AI systems that run experiments, generate discoveries and compress decades of scientific progress into years. Take that seriously and something strange happens to the economy. The returns to innovation, which have always been capitalism's engine, stop accruing to the people who innovate and start accruing to the people who own the machines that innovate.

From there the road forks.

Paint the left-hand road forward. It is 2035. Agents handle most analysis, most drafting, most first-pass creative work. GDP is up. Unemployment is stubborn. Governments run some version of a stipend. A few thousand companies and their shareholders own the capability layer. Everyone else has a comfortable, subsidised, slightly purposeless existence with no route into the upside.

That is not a socialist dystopia and it is not a capitalist utopia. It is fragile, because capitalism has only ever stayed legitimate while enough people had a stake in it. Last year's Nobel in economics went to Aghion, Howitt and Mokyr for their work on creative destruction, the idea that growth depends on new entrants constantly displacing incumbents. Creative destruction needs founders. It needs people who can afford to take a risk. A stipend class does not produce founders.

Ownership works. It just does not reach far enough.

The right-hand road is not theory. It has a track record, and the most convincing version comes from the least sentimental corner of finance.

Pete Stavros at KKR spent 14 years granting equity to every worker in the industrial companies the firm bought, from the factory floor to the front office. When KKR sold garage door maker CHI Overhead Doors in 2022 for $3 billion, around 650 employees shared roughly $360 million. The longest-serving received payouts worth more than six times their salary. Under the ownership model margins had risen from 20 to 35 per cent, and the quit rate had collapsed.

$360m

shared by around 650 CHI Overhead Doors employees on exit. HBS case study.

6.5×

annual salary paid out to the longest-tenured workers. Just Capital.

20→35%

margin under employee ownership, unheard of for a garage door maker.

180,000

workers reached by Ownership Works programs across 123 companies by end 2024.

Stavros founded Ownership Works in 2022 to spread the model, with a goal of $20 billion in new wealth for working families by 2030. Blackstone has since committed to granting equity to most employees in future US buyouts. Private equity, of all people, has worked out that ownership makes companies better and workers wealthier.

The problem is reach. Equity still mostly flows through two channels: you found the company, or you get hired by one big enough to have a plan. For the experienced professional whose role is being hollowed out, and for the early-stage founder who cannot afford to hire her, neither channel is open.

The Australian who called it.

The clearest articulation of what happens next comes from an Australian. Daniel Priestley grew up in Melbourne, built his first company there at 21, and now runs Dent Global and ScoreApp from London. In March he sat down with Bartlett for two hours on AI, and Fortune ran the headline that plumbers will soon out-earn lawyers.

Priestley's view is not that jobs vanish and nothing replaces them. He leans on the Jevons paradox: when something gets cheaper, people use far more of it. As the cost of building a product collapses, he expects millions of small, specialist businesses rather than a handful of giants. But he is precise about where the human sits in that business.

Bartlett runs his own companies the same way. He told Fast Company in March that as agents absorb routine labour, his businesses are shifting towards specialists with expertise that cannot be replaced, and he invests in platforms built on the belief that anyone with a skill should be able to work for themselves.

Put Priestley and Bartlett together and you get a specific prediction. The future has more companies, smaller companies, and each one needs a small amount of very senior judgement it cannot afford to employ full-time. That is a market. In Australia, it is a market with nothing serving it.

Why Australia, and why now.

Australian startups raised $5.1 billion in 2025, the third-largest year on record and up 24 per cent on 2024. That is the headline from Cut Through Venture and Folklore's State of Australian Startup Funding report. The next three numbers are the story.

Deal count fell 17 per cent. The top 20 deals took 58 per cent of the money. Capital is concentrating in a small number of AI-native winners while the number of funded teams shrinks. Chris Gillings, who compiles the report, put the gap in one line.

Underneath the capital problem sits an older one. Airtree's John Henderson wrote in 2019 that the number one priority of every promising Australian startup he spoke to was hiring great people, and specifically people who had been part of a scale-up journey before. No amount of funding, he said, solves a talent gap in the short term. Seven years later, with fewer teams getting funded and more experienced operators being restructured out of large companies, the mismatch has become absurd. The people who know how to build are sitting on one side of the market. The people who need them are on the other. The only bridge between them is a salary neither side can afford.

The structural gap, in one paragraph. Australia's employee share scheme rules were written for employees. The startup concession under Division 83A is generous, deferring tax until sale for unlisted companies under ten years old with turnover below $50 million. But it assumes an employment relationship, and the share-based version requires the offer to go to at least 75 per cent of permanent staff with three years' service. There is no clean, standard, tax-sensible way for a founder to grant a stake to a fractional finance director, a growth lead or a former Series A operator who is not on payroll. So it happens informally, on trust, or not at all.

What it looks like for one founder and one expert.

Bring it down to two people in Melbourne.

She has a product, a team of four and about eight months of runway. She needs a fractional CFO, someone who has run a growth function, and someone who has actually closed a Series A in this market. She cannot pay any of them what they are worth. Every week she spends learning what they already know is a week of runway gone, and her seed investors are watching the clock.

He is 44, spent 20 years in finance, and his employer restructured his division in March. He does not have to choose between a salary and a stake. He can take the next role, keep the pay cheque, and put that pattern recognition to work across three or four early-stage companies on the side, paid in something that compounds.

Some of his stakes will not pay out. That is the honest part. But the ones that do have asymmetric upside, and the ones that do not still return relationships, a track record and proof that his expertise creates value outside a job description. Multiply that by tens of thousands of professionals and founders and you have something the stipend economy cannot produce: a broad base of Australians with a direct financial interest in new companies succeeding.

The outcome we are building towards.

By 2030 we want three things to be true in Australia.

First, trading expertise for equity is as normal, as standard and as safe as trading it for a salary, with paperwork a founder can sign in an afternoon and a tax position an expert can understand.

Second, the number of Australians holding a stake in an early-stage company is measured in the tens of thousands, not the hundreds, and most of them got there through their skills rather than their chequebook.

Third, the funded-teams number in the Cut Through report goes up, not down, because more founders reach the milestones that make them fundable before the money runs out.

None of that requires the AI companies to slow down. It does not require a stipend. It requires ownership to be spread wide enough that the system keeps producing the founders it needs to renew itself. That is what safeguards capitalism: not a fence around the machines, but a stake for the people.

Everything above is easy to say and hard to do. Equity is a legal instrument, not a handshake. It has vesting, tax consequences, disclosure obligations and a rulebook written for a world where the only people receiving shares were employees and investors. Trading expertise for equity at scale has never been done cleanly in this country, which is why most of it still happens on trust and falls apart when it matters.

We have spent the last year working out how to change that, and we are starting in Australia, because this is where the gap is widest and the rules are clearest. The first piece goes live on 19 October.

More soon.

Join the Australian waitlist at getsweaty.com

Sources

1. Axios, "AI jobs danger: Sleepwalking into a white-collar bloodbath", May 2025; TheStreet on Amodei at Davos, April 2026.
2. IMF, Kristalina Georgieva blog and staff analysis on AI and the global labour market, January 2024, via CNBC.
3. Mercer Global Talent Trends 2026, via CNBC.
4. The Diary of a CEO with Steven Bartlett: Yoshua Bengio (December 2025), Scott Galloway (May 2026), Daniel Priestley (March 2026), AI agents debate with Amjad Masad, Bret Weinstein and Daniel Priestley (May 2025).
5. Department of Employment and Workplace Relations, "AI and employment in Australia", July 2026.
6. ServiceNow Australia Enterprise AI Maturity Index, July 2025.
7. EY-Parthenon, "AI productivity gains could deliver up to $116bn boost to Australia's economy", August 2026; Productivity Commission, "Harnessing data and digital technology" interim report, August 2025.
8. Scott Farquhar, National Press Club address, July 2025, via Tech Council of Australia and The Nightly.
9. Business Insider and The Atlantic, Altman interview with Nicholas Thompson, April 2026; OpenResearch unconditional cash study findings, July 2024.
10. Harvard Business School case, "Ownership Works: Scaling a Profitable Social Mission"; Just Capital and Stanford GSB coverage of CHI Overhead Doors; ImpactAlpha Q&A with Pete Stavros, May 2025; Wall Street Journal on Blackstone, May 2024.
11. Fortune, "Plumbers outearning lawyers", March 2026; Fast Company, "3 insights into the future of business from Steven Bartlett", March 2026.
12. Cut Through Venture and Folklore Ventures, State of Australian Startup Funding 2025, via Forbes Australia and SmartCompany, February 2026.
13. John Henderson, Airtree, "Impressions of #startupaus after 100 days", 2019.
14. Income Tax Assessment Act 1997, Division 83A startup concession, via Viridian Lawyers and Carta Australia.