The Real Impact of AGI on the Future of Work
Ever since artificial intelligence began to reshape industries, a persistent worry has haunted the workforce: will AGI job automation eliminate most human jobs? However, a new perspective from Yale economist Pascual Restrepo suggests that this fear may be misplaced. Rather than replacing every occupation, advanced AI could bypass the majority of jobs simply because they’re not significant enough to warrant automation. This insight, based on Restrepo’s latest research, reframes the debate on the future of work in an era of artificial general intelligence (AGI).
Why Most Jobs Won’t Be Automated
The traditional assumption is that as AGI job automation becomes possible, only the most creative or deeply human work will remain. Yet, Restrepo’s working paper, “We Won’t Be Missed: Work and Growth in the AGI World,” published by the National Bureau of Economic Research, argues that most jobs are unlikely to be automated—not due to technical limitations, but because they’re not vital for economic progress. According to Restrepo, the focus of compute (the computational power fueling AI systems) will be directed toward bottleneck tasks that are crucial for growth, such as addressing existential risks or advancing science, rather than replacing everyday roles.
Bottleneck vs. Supplementary Work
Restrepo categorizes work into two types: bottleneck and supplementary. Bottleneck work includes roles essential to economic growth—like energy production, infrastructure maintenance, and scientific advancement. These are the tasks most likely to be automated using AGI. In contrast, supplementary work encompasses jobs that the economy can function without, including arts, hospitality, customer support, and even some academic research. Since automating these roles would require vast computing resources but offer little economic payoff, AI may simply ignore them.
For example, jobs in hospitality, live entertainment, and other socially intensive fields could remain largely human. Not because AGI can’t do them, but because it’s not worth the substantial computational cost to automate them. This offers a reassuring outlook for baristas, performers, and designers—work in these areas may survive the era of AGI job automation, not due to some irreplaceable human spark, but because AGI’s resources will be reserved for higher-impact tasks.
The Changing Value of Human Skills
One of Restrepo’s key arguments is that AGI does not make human skills obsolete; instead, it revalues them. In a post-AGI world, the new scarcity is not labor or intelligence, but compute. The worth of human work is set by the opportunity cost of the compute needed to replicate it. If compute and human skill are the only scarce resources, average wages could even be higher. However, as AGI takes over bottleneck roles, the relative role of human labor shrinks, and the share of national income that goes to workers declines.
This decoupling of wages from GDP growth is a significant shift. Today, economic expansion usually means higher wages and better living standards. But in an AGI-driven economy, growth is powered by computational resources, not human effort. The ceiling for labor income becomes what it would cost to automate human work—a figure that shrinks as compute becomes more abundant. According to Restrepo, “labor’s share of GDP goes to zero” over time, with most income accruing to the owners of compute.
Winners, Losers, and Inequality
The question of who owns the compute becomes central in the AGI era. As BlackRock CEO Larry Fink noted, AI could further concentrate wealth among those positioned to control it, widening the gap between the top 1% and everyone else. Restrepo suggests two possible responses: redistributing gains through universal income or treating compute as a public resource, similar to land or natural capital.
Restrepo also distinguishes between two types of automation transitions. A compute-binding transition, limited by hardware availability, would allow for gradual adjustment and steady wage changes. An algorithm-binding transition, like the current surge in AI capabilities, brings abrupt shifts in demand and pay, rewarding workers whose roles remain unautomated—such as electricians and tradespeople—while others face sharp wage drops. This pattern is visible today, especially in data center construction where certain skilled workers command significant premiums.
AGI Job Automation: Not a Universal Threat
Despite the challenges, Restrepo offers a measure of reassurance: workers, as a group, are not made worse off by AGI job automation. Because AGI expands the economy’s productive potential, total labor income after AGI could be higher than before. The risk is that these gains are unevenly distributed, with a growing divide between those exposed to capital markets and those left behind.
The core message of Restrepo’s analysis is clear: the arrival of AGI may not make us collectively poorer, but it could sever the longstanding connection between work and social recognition. In an AGI world, many jobs may simply be ignored, not missed. The real question is not whether AGI will take your job, but whether your job was ever essential enough for it to matter in the first place.
This article is inspired by content from Original Source. It has been rephrased for originality. Images are credited to the original source.