Workers at Australian ports have demanded a 28-hour working week as part of ongoing industrial negotiations explicitly tied to the deployment of AI and automation technologies on the waterfront. The claim, reported by BBC Business, forms part of a broader pattern of labour-market disruption attributed to artificial intelligence: in the same week, a report highlighted that women and university graduates face the highest job displacement risk from AI, with telemarketers, advertising professionals, and accountants among the most exposed occupations. The dock workers’ demand is noteworthy because it is not a demand for protection from automation but a demand for a share of its productivity gains — a shift in framing from “don’t automate us” to “if you automate us, the hours we do work should reflect the new reality of human-machine labour ratios.” The Australian waterfront has a long history of militant unionism and protracted industrial disputes, but the 28-hour claim places the current negotiations in a specifically twenty-first-century context.
The received wisdom
The progressive union-movement reading of the dock workers’ claim is that it represents a reasonable response to a genuine imbalance in how AI productivity gains are being distributed. The argument runs as follows: when automation reduces the number of human hours required to move a given volume of cargo, the firms deploying that automation capture the entire productivity dividend in the form of reduced labour costs and higher margins. Workers, by contrast, face either redundancy or — for those who remain — a working environment in which they are increasingly adjuncts to automated systems, performing residual tasks that machines cannot yet handle. The 28-hour demand is, on this reading, a claim that human workers should share in the time freedom that productivity growth creates — a modern version of the case that drove the reduction of the working week from sixty hours to forty hours across the twentieth century. BBC Business coverage noted the negotiations are explicitly framed as AI-related, marking a threshold moment in which organised labour formally names the technology as the subject of industrial bargaining.
This framing has historical resonance. The reduction in working hours over the past century was not a voluntary gift from employers; it was won through sustained industrial and political pressure and was made possible by sustained productivity growth.
A different read
History cuts both ways here, though. The twentieth-century reduction in the working week from sixty to forty hours was accompanied by sustained real-wage growth — workers worked less and earned more in real terms, partly because of union power, partly because of genuine productivity gains that raised living standards broadly. The question for the current moment is whether AI-driven automation follows that historical pattern or whether it represents something structurally different: an automation wave that concentrates gains more narrowly, raises productivity without a corresponding rise in broad-based wages, and produces a bifurcated labour market between highly paid AI-complementary workers and a larger group of workers in AI-exposed occupations facing stagnant earnings or displacement.
The evidence on this is genuinely contested. The argument that automation historically creates as many jobs as it destroys — the “lump of labour fallacy” inverted — has strong empirical support over long time horizons. The concern raised by economists studying the current AI wave, including work by Daron Acemoglu and Pascual Restrepo, is that not all forms of automation are equal: automation that replaces labour in existing tasks without creating significant new tasks may not trigger the compensating job creation that previous technological transitions produced. Whether large language models and their successors fall into that category remains genuinely uncertain.
What the dock workers’ claim illustrates, whatever one’s view of the economics, is that organised labour has identified the right strategic question faster than most political parties have. Rather than opposing automation — a rear-guard action that history suggests is inevitably lost — the maritime unions are asking how its gains should be shared. The 28-hour demand may be an opening position in a negotiation that ends somewhere different; but the framing of that demand is a significant cultural moment. It asserts that there is no natural law by which the productivity dividend from a new technology must accrue entirely to capital.
The right-of-centre response to this is not reflexive opposition to union demands but a question about mechanism. Union bargaining at the firm or sector level is one mechanism for redistributing productivity gains; it tends to work best in sectors with high union density and significant barriers to offshoring, both of which the waterfront exemplifies. The broader challenge is that most AI-exposed workers are not in heavily unionised sectors — they are the telemarketers, accountants, and advertising professionals identified in the Australian AI job risk report. For those workers, collective bargaining is largely unavailable, and the distributional question has to be addressed through other instruments: tax policy, investment in retraining, and the design of social insurance systems that do not penalise workers for leaving AI-disrupted occupations.
Samsung’s 1,800 percent profit jump driven by AI chip demand is, in this context, more than a business story. It is a data point in a distributional argument. The gains from the AI infrastructure build-out are flowing with remarkable concentration to a small number of firms and shareholders. The dock workers’ 28-hour demand is an early and relatively crude attempt to route some of that gain back toward labour. The question for policymakers is whether better-designed instruments can do the same job more efficiently and at broader scale.
The AI labour question is also a political question. The parties best positioned to benefit from working-class voters displaced by automation are, counterintuitively, often parties of the right — if those parties can credibly offer industrial policy, trade protection, and immigration control as substitutes for the wage-protective functions that union density used to provide. The risk for the centre-right is that it defaults to a pure free-market framing that reads as indifference to displacement, ceding that constituency to either the radical left or to ethno-nationalist movements that offer cultural grievance as a substitute for material remedy.
What to watch
Watch the Australian waterfront negotiations for the settlement: the final number of hours agreed will be a leading indicator for similar claims in other AI-exposed industrial sectors. Watch the AI job-risk reports being produced by governments and international organisations — the Australian findings on women and graduates are likely to be replicated in studies across OECD countries and will shape the political salience of the AI-labour question going into the next electoral cycle. Watch Samsung and Nvidia earnings for signals about the duration of the AI infrastructure investment boom — if capital expenditure on AI infrastructure plateaus, the political urgency of distributional questions intensifies. And watch for the first significant AI-linked redundancy wave in a white-collar sector, which will test whether the political consensus on AI remains as sanguine as it currently is among governing parties.
— J