The catastrophe narrative around artificial intelligence and jobs may be overdrawn. Rather than wholesale automation eliminating workers, a growing body of research suggests AI is quietly reshaping labour in ways that are subtler and potentially more insidious: the unbundling of jobs into narrower, lower-paid tasks.
Research from Luis Garicano, a professor at the London School of Economics, along with Jin Li and Yanhui Wu from the University of Hong Kong, proposes a framework that cuts through the polarised debate. Their argument rests on a fundamental insight: jobs aren't neat lists of tasks – they're bundles.
Consider radiologists. They don't just read scans; they interpret edge cases, talk to clinicians, and sign off on decisions people act on. Replace the image-reading bit, and you haven't necessarily replaced the job. That distinction matters. The authors identify what they call "strong bundles" (jobs that lose value if broken apart) and "weak bundles" (roles that can be decomposed without much loss).
In weak-bundle occupations, AI automates some tasks and narrows the boundary of the job, whilst in strong-bundle occupations, AI improves performance inside the job but does not remove the human from the bundle. Think of customer support agents handling tickets or coders writing predictable functions. When AI takes those fragments, workers don't simply vanish; they concentrate on the remainder. And here lies the trap.
Once AI takes over part of the work, the human stops dividing their time. They go all-in on what remains, which means output per worker jumps, prices fall, and suddenly you don't need as many workers as before. In other words, the hit to employment doesn't come from AI doing the job outright, but from humans becoming too efficient at the leftovers.
This framework reconciles two seemingly contradictory observations. AI is reshaping jobs, not wiping them out. Tasks move around, productivity may go up, yet employment and hours haven't shifted much – at least yet. The warning systems and forecasts predicting millions of job losses sit uncomfortably beside the reality that total headcounts haven't crashed. Both can be true if the mechanism is narrowing, not elimination.
The implication, though, is uneven. If you're in a strong-bundle job – something heavy on judgment, context, or responsibility – AI is more likely to make you faster and better paid. Lawyers reviewing contracts gain an assistant; senior diagnosticians gain a scanner. The high-touch, high-judgment roles become more valuable.
If you're in a weak one, it may quietly hollow out your role until there's not much left to defend. The risk falls heaviest on entry-level workers and those in routine cognitive or administrative roles. For Australia specifically, the picture shows both progress and peril. Generative AI is more likely to augment jobs than replace them, with early adoption varying across industries and many workers using tools independently in their work, according to Jobs and Skills Australia's Generative AI Capacity Study.
Yet the picture for new entrants is darker. Recent employment decline in AI-exposed sectors is particularly pronounced for those under age 25, with employment totals for older workers not declining. The job market is getting very tough for new graduates in AI-exposed fields. Australia has experienced notable tech layoffs in 2026, with about 9,238 layoffs linked directly to AI adoption, and Sydney now ranking third globally in absolute job cuts.
Counterbalancing this, there is now a focus on technology-led growth, with AI becoming part of everyday work and leadership, and AI engineering now the number one job on the rise in Australia. Still, the problem is distributional. Winners and losers aren't evenly spread. Fiscal prudence demands policymakers watch the transition period. Growth in profits will be stronger than growth in wages, and governments will need to address emerging issues through tax policy, competition policy and industrial relations.
The Garicano framework offers neither uncritical optimism nor apocalyptic doom. It suggests that AI's harm may arrive not as a cliff but as a slow decomposition of roles, with some workers gaining and others quietly squeezed into ever-narrower lanes. The question isn't whether the bundle holds, but for whom and for how long.