How AI will disrupt workers
Australia's unions have a plan for AI disruption. History suggests it won't arrive the way they're promising — and the workers most at risk aren't the ones they're thinking of.
The question economists and workers will be arguing about for the next decade is not really whether AI destroys jobs. It is whether the productivity gains get distributed or hoarded. History has not given a comforting answer to that question, and the early signs from the current wave of AI adoption are not especially reassuring.
Consultation rights only work when there are comparable jobs to move into
The framing from outgoing ACTU secretary Sally McManus, in one of her final interviews before leaving the role, is instructive. Asked about AI's threat to jobs, she pointed to the standard existing protection: employers must consult before rolling out technology that displaces workers. Her view was that if AI disruption stays contained to the tech sector, that is probably enough. But if it spreads economy-wide, consultation rights become meaningless when there are no comparable jobs for displaced workers to move into. At that point, she said, unions would push for shorter working hours at no pay reduction as a mechanism for sharing productivity gains more broadly.
It is a coherent idea. It was also the idea in the 1970s, and the 1980s, and it has never quite arrived. The 38-hour week was won as a genuine labour movement achievement, but the idea that automation would progressively shorten the standard working week has been deferred at every subsequent technological inflection point. Personal computers did not deliver it. The internet did not deliver it. The smartphone did not deliver it. Each wave raised productivity and, eventually, wages for workers who remained employed, but it did not reduce the hours those workers were expected to put in. If anything, the opposite happened: the professional class, the group that technology augmented rather than replaced, worked more, not less.
The four-day week sounds appealing in the abstract and polls well. The revealed preference of workers who actually gain negotiating power tends to be a pay rise.
Workers given a genuine choice tend to take the money, not the time
The millionaire-versus-billionaire trade-off captures something true about human nature that the shorter-hours argument tends to ignore. If AI genuinely makes workers twice as productive, most of them, given the choice between doing half the work for the same pay or the same work for twice the pay, will take the money. Not because they are irrational or have been manipulated into a bad preference, but because relative wealth is itself the goal. A nicer house, a better school for the kids, a holiday that feels like a real holiday rather than a stretched budget exercise: these things are purchased with income, not with leisure hours. The four-day week sounds appealing in the abstract and polls well. The revealed preference of workers who actually gain negotiating power tends to be a pay rise.
That does not make the distributional problem go away. It just locates it differently. The real risk is not that AI makes workers so productive that they do not know what to do with themselves. It is that AI is productive enough to eliminate entire categories of routine cognitive work, the administrative, the paralegal, the data-entry, the call-centre, while the gains accrue almost entirely to the firms deploying it and the workers skilled enough to direct it. That is not a shorter-hours story. That is a two-tier economy story.
HECS relief targets the wrong workers
McManus also floated HECS debt relief for workers displaced by AI, which is worth examining on its own merits. The argument would be that someone forced out of their field by automation should not also be carrying the debt they incurred training for that field. That is not an unreasonable proposition for a one-off structural disruption. But as we have noted before when examining how HECS changes interact with graduate debt loads, debt relief targeted at already-qualified graduates does relatively little for the workers carrying the heaviest burden relative to their income, and nothing for the workers who never had the credentials to begin with.
The workers most exposed to AI displacement are not, in the main, university graduates with HECS debts. They are people doing jobs that never required a degree, jobs that are now cheaply automatable precisely because they were well-defined enough to be learned without one. Call-centre operators, data processors, junior back-office staff. They do not have HECS debts. They have mortgages, or rents, and not much margin.
The adjustment mechanism everyone is counting on is already broken
What they need is not shorter hours or debt relief. They need a labour market mobile enough to absorb them into new roles, retrained fast enough to make that transition real, and tight enough to give them bargaining power when they get there. Australia's job mobility rate is already near record lows, which means the adjustment mechanism everyone is counting on to make technological disruption manageable is already functioning badly, before the disruption has properly arrived.
The AI productivity boom may well come. The question of who it enriches is a political question as much as an economic one, and the answer will depend on institutional arrangements that are currently being built, or not built, right now. The unions know this. The government knows this. What neither has yet produced is a policy architecture that matches the scale of what might be coming.
Sources
The Bearing — How Changing HECS Payment Timing Reveals the Real Cost of Student Debt
The Bearing — Less people are changing jobs, and that's bad
Frequently Asked Questions
Will AI reduce working hours in Australia?
History suggests not automatically. Every major wave of automation since the personal computer has raised productivity without reducing standard working hours — if anything, the workers technology augmented ended up working more. The more likely outcome is that productivity gains show up as higher pay for skilled workers, not shorter weeks for everyone.
Which Australian workers are most at risk from AI automation?
The workers most exposed are those in routine cognitive roles that never required a degree — call-centre operators, data processors, and junior back-office staff. These jobs are cheaply automatable precisely because they were well-defined enough to be learned without tertiary qualifications, and the workers doing them have little financial margin to absorb the disruption.
Why wouldn't HECS debt relief help workers displaced by AI?
Because the workers most at risk of AI displacement are largely not university graduates carrying HECS debts. They are workers in non-degree roles who have mortgages or rents and no credentials to forgive. HECS relief would primarily benefit a different, generally more financially secure group.
What do Australian unions plan to do about AI job losses?
The ACTU's position, as articulated by outgoing secretary Sally McManus, is to push for shorter working hours at no pay reduction if AI disruption spreads economy-wide — effectively sharing productivity gains by reducing the hours each worker needs to supply. Existing consultation rights are the first line of defence, but McManus acknowledged these become inadequate if there are no comparable jobs for displaced workers to move into.
Who captures the gains from AI productivity — workers or companies?
Based on historical precedent, the gains from automation have flowed disproportionately to capital owners and to skilled workers who direct technology, rather than to workers displaced by it. The distribution of AI productivity gains is ultimately a political question, determined by institutional arrangements — bargaining rights, retraining systems, labour market policy — that are being built or neglected right now.