Predictions of a looming "jobpocalypse" driven by generative AI have captured headlines, with prominent voices warning that white-collar unemployment could surge dramatically. But a closer look at how past technological waves reshaped labor markets suggests the reality is far more complicated—and far less deterministic—than a simple equation of automation equals job loss.
Many people, including some of the field's leading names, have predicted potentially disastrous impacts on jobs from generative artificial intelligence. A leading voice in AI safety, Dr. Roman Yampolskiy; Anthropic CEO Dario Amodei and Silicon Valley investor Vinod Khosla, for example, have predicted that unemployment in white-collar jobs could rise to 20%-99%.
The term jobpocalypse has gained traction alongside those warnings. But John Burn-Murdoch, columnist and chief data reporter for the Financial Times, argues that the narrative misses critical context. The impact of AI on employment, he writes, depends on a range of economic and structural forces that extend well beyond whether a machine can perform a given task.
One of the most important—and often overlooked—factors is demand. As Burn-Murdoch notes, technology has boosted productivity in areas such as software and professional services, which, in theory, should reduce the need for workers. Instead, employment in these sectors has expanded "because rising consumer demand has more than compensated."
The surge in software development jobs since the 1990s offers a clear example: efficiency gains lowered costs, which in turn fueled an even faster increase in consumption.
A similar dynamic has played out across professional services, where tools have made accountants, architects and advertising creatives more productive without shrinking their ranks. In some cases, the opposite has occurred, with demand growth outpacing productivity gains.
Healthcare provides another instructive case. Advances in lab testing and diagnostic imaging—including AI—have significantly improved efficiency, and in some instances, systems can outperform human experts. Yet employment continues to grow. The National Resident Matching Program reported that in 2025, there were a record 1,208 positions offered in U.S. diagnostic radiology residency programs, while the Works in Progress newsletter noted that radiology remains the second-highest-paid medical specialty in the country, with an average income of $520,000. Regulatory and insurance frameworks also play a role, effectively limiting full automation even where the technology exists.
By contrast, manufacturing illustrates what happens when demand is already saturated. Following decades of post-Second World War expansion, U.S. manufacturing employment peaked at 19.5 million in July 1979 before entering a long decline as automation and global competition took hold. By March 2026, employment had fallen to 12.6 million. In this case, productivity gains translated more directly into job losses because consumption did not rise fast enough to offset them.
The second-order effects of technology can also reshape employment in unexpected ways. The same digital transformation that boosted knowledge work helped hollow out retail employment—not because store workers were directly automated, but because e-commerce shifted activity away from physical locations. At the same time, that shift drove growth in logistics and warehousing as demand for distribution and delivery surged.
Even within the same sector, the impact of new tools can diverge sharply across roles. The spread of spreadsheets in the 1980s reduced demand for bookkeeping and accounting clerks while increasing opportunities for higher-skilled accountants and analysts. This pattern raises the possibility that AI may augment higher-skilled workers while displacing more routine roles, challenging the assumption that top earners are most at risk.
History also shows that the most disruptive effects of technology are often indirect. Bank tellers, for example, initially weathered the introduction of ATMs, only to face steeper declines later as smartphones enabled mobile banking and reduced the need for physical branches. Similarly, the internet's impact on newspapers came less from automation in newsrooms and more from the collapse of traditional advertising models.
Taken together, these examples underscore Burn-Murdoch's central point: asking whether AI can perform a task is only a starting point. Employment outcomes hinge on demand, regulation, industry structure, and the broader economic ripple effects that follow technological change. So far, occupations most exposed to AI have proven just as likely to grow as to shrink—a pattern that suggests the future of work will be shaped as much by markets and institutions as by the technology itself.
© Arc, All Rights Reserved. Request academic re-use from www.copyright.com. All other uses, submit a request to TMSalesOperations@arc-network.com. For more information visit Asset & Logo Licensing.