Enterprise & Work

Four Years After ChatGPT, the AI Job Apocalypse Still Has Not Arrived

Economist John Quiggin argues NAB, Qantas, Canva and Atlassian layoffs reflect cost-cutting and offshoring, not AI — and the bigger risk lies in stock markets, not jobs.

Is the AI job apocalypse finally here? The evidence suggests otherwise | John Quiggin
Is the AI job apocalypse finally here? The evidence suggests otherwise | John QuigginAlexandra_Koch / pixabay
By Rebecca Stone6 min read

Why it matters

  • Four years after ChatGPT's launch, productivity growth remains negligible despite a datacentre investment boom, mirroring the 1990s fibre-optic era.
  • Qantas is outsourcing 850 jobs to India while invoking AI, which economist John Quiggin calls a case of simple offshoring.
  • NAB, which has announced mass sackings most years since 2010, explicitly denied AI played a role in its job cuts.
  • Open-weight models such as DeepSeek and Mistral run about a year behind leaders like OpenAI and Anthropic at a fraction of the cost.
  • Quiggin quotes economist Robert Solow's 1987 remark: 'You can see the computer age everywhere but in the productivity statistics.'

Four years after the launch of ChatGPT, productivity statistics show the same pattern that puzzled economists in the computer age: enormous hype, enormous investment, and very little measured gain. That is the central claim of John Quiggin, professor at the University of Queensland's school of economics, who argues in a new analysis that recent large-scale job losses at NAB, Qantas, Canva and Atlassian — some of them explicitly blamed on AI — do not amount to the long-predicted AI jobs apocalypse.

Quiggin opens with history. In 1987, the Nobel prize winning economist Robert Solow remarked: "You can see the computer age everywhere but in the productivity statistics." At the time, computers had been around for 40 years. Real productivity gains did emerge in the 1990s after the arrival of the internet, but they were short-lived.

The parallel, Quiggin writes, is direct. Four years after the launch of ChatGPT, the pattern seems to be much the same: lots of hype and excitement but with very little productivity growth. The main economic impacts, as in the 1990s dotcom era, have been a stock market boom and huge investment in infrastructure — fibre optics then, datacentres today — betting on future demand that has yet to emerge.

The stakes are straightforward. If AI-driven automation were genuinely displacing workers at scale, productivity data would show it. Instead, Quiggin argues, the current wave of Australian corporate layoffs has more mundane explanations, and the more serious risk may sit in financial markets rather than the labour market.

What did NAB actually do?

National Australia Bank has specifically denied that AI played a role in its decision to cut jobs, and Quiggin sees no reason to disbelieve this. The sackings, he writes, seem to be business as usual for the bank, which since 2010 has announced a mass sacking in most years, sometimes more than one. Yet its total employment numbers have remained unchanged, even though many employees have quit or retired.

"Quite simply, NAB's business model, like that of other banks and big employers, is to hire employees when it needs them and sack them whenever it is profitable to do so," Quiggin writes. "That's capitalism after all, but it's a far cry from the days when big employers like banks gave and expected loyalty in dealing with their workers."

The implication for readers tracking AI layoffs: a corporate press release that mentions AI does not establish that AI caused the cuts. NAB's headcount record since 2010 suggests cyclical cost management, not a technological break.

Is Qantas automating jobs or offshoring them?

Qantas has invoked AI, but Quiggin argues the airline did so only, in his words, to muddy the waters. The decision to outsource 850 jobs to India is, he writes, consistent with the airline's longstanding strategy of doing over Australian workers while claiming preferential treatment due to being a national flag carrier.

The 850-job figure is the story's largest concrete number, and Quiggin treats the AI framing as a complication in reporting what is in reality a simple case of offshoring. The logic, he argues, cuts against Qantas itself: if the jobs are going to be automated anyway, why go to the trouble of sacking existing staff ahead of time and setting up a new operation overseas?

"Let's hope that members of the Australian political class give this some thought as they enjoy the hospitality of the Chairman's Lounge," Quiggin writes — a pointed reference to the airline's relationships with Australian policymakers.

Why does Canva look more exposed?

Canva, the Sydney-founded design software company, looks more vulnerable to genuine AI displacement, according to Quiggin. The firm operates on a "software as a service" model, charging monthly or annual subscriptions for access to easily usable graphics design software. That model has been highly successful for users who do not want to struggle with more powerful but complex packages.

AI changes the calculus. Anyone who wants, for example, a poster advertising an event can now do so in seconds, specifying the style, colour scheme and so on. The result may be recognisable as "AI slop" — Quiggin's term — but many people will settle for that. For a company whose product is approachable design, fast-generation AI tools attack the core value proposition directly.

Can Atlassian's software be replicated by AI?

Atlassian's business model is also under challenge, Quiggin writes, but its software offerings will be harder to replicate with the use of AI. Atlassian's applications focus on project management and organisation-wide data. Piecemeal use of AI by individual workers is unlikely to replace these functions.

He allows one caveat: it is possible that AI will enable the reverse engineering of software packages, but this remains to be seen. Atlassian's exposure is therefore structural rather than immediate — a question of whether AI eventually erodes the moat around enterprise workflow tools, not whether a chatbot replaces them next quarter.

Are the AI companies themselves safe?

Quiggin closes the argument with the hyperscalers. The AI leaders — firms such as OpenAI and Anthropic, along with Meta (Facebook/Instagram) and Alphabet (Google) — aren't safe either, he writes.

Cheap "open-weight" models such as DeepSeek from China and Mistral from France are running only a year or so behind the leaders, while investing a small fraction of the money. The comparison recalls the 1990s fibre-optic buildout: infrastructure spending justified by demand projections that have not yet materialised.

The endgame Quiggin sketches is deflationary for the industry. "It's entirely possible that LLMs (large language models) will become like email, an essential part of life for many people but not a source of profit," he writes. If that is the case, "we are more likely to see an apocalypse in the stock market than in the jobs market."

What does this mean for the AI discourse?

The analysis assembles four distinct layoff stories into one argument. NAB is cyclic cost-cutting with an explicit AI denial. Qantas is offshoring of 850 jobs dressed in AI language. Canva faces real product-level substitution. Atlassian faces a slower, structural threat. Only the last two involve AI as a genuine commercial force, and neither represents the mass displacement that "job apocalypse" rhetoric implies.

The broader economic backdrop reinforces the scepticism. Solow's 1987 observation held for years before the internet finally delivered measurable productivity gains in the 1990s — gains that were short-lived. Four years into the ChatGPT era, Quiggin sees fibre-optic history repeating as datacentre investment: a bet on future demand that has yet to emerge, financed by a stock market boom.

For policymakers, the piece carries a warning about attribution. Accepting corporate AI framing at face value can misdirect labour policy toward an automation threat while offshoring proceeds under cover of that framing. For investors, the warning runs the other way: open-weight competitors a year behind at a small fraction of the cost could turn today's AI profit expectations into tomorrow's commodity business. Quiggin's bottom line is that the evidence, four years in, points to neither an employment collapse nor a durable AI oligopoly — and that the next correction, if it comes, will show up in share prices first.

Source: The Guardian AI

Share this article:

More from Rebecca Stone

Rebecca Stone

Show full bio

Correspondent covering companies, deals and strategy in artificial intelligence and machine learning at AI In Context.

620 articles

Related articles

  1. Largest Study of ChatGPT Use Shows AI Becoming Everyday Tool
  2. Alphabet's Search Revenue Slows to 12% as AI Fears Hit the Stock
  3. A Federal Job Guarantee Emerges as the Answer to AI Unemployment
  4. Forrester: AI bubble fears premature, but math is getting harder
  5. 39% of Engineering Skills Will Shift by 2030. The Definition of 'Adaptable' Hasn't.

« Previous article