Macro● Neutral
Rajan on AI's job threat and the corporate response
Raghuram Rajan argues AI displacement will be gradual but firms must retrain workers. He proposes a tax on AI tokens and training credits to ease the transition.
Continue with
Artificial intelligence will displace jobs, but no one knows how fast or how far, Raghuram Rajan wrote in a recent essay. The University of Chicago professor and former Reserve Bank of India governor laid out a case that is neither apocalyptic nor dismissive. The pace of adoption, he argued, depends on how quickly firms integrate AI into existing workflows – and that is proving harder than the technology's raw capability might suggest.
A U.S. Census Bureau survey cited by Rajan found that only 20% of firms with fewer than 20 employees use AI today. Among businesses with at least 250 workers, the share rises to 37%. Even that figure is low, Rajan noted, given the survey's low bar – it asks whether AI is used "in any business function."
The bottleneck is integration. Rajan said companies need models trained not just on existing data but adaptive to new data generated through everyday use. Uncertainty over cost is another drag. Many large firms are running pilots and delaying hiring or firing decisions until they see a clearer picture. Competitive pressures will eventually force broader adoption, he predicted.
But the employment picture is not uniformly grim. If companies keep producing the same goods and services, AI will behave like any historical technology. Some jobs vanish; others become more productive and interesting as drudgery is removed; entirely new roles emerge, Rajan wrote. He invoked the Jevons effect: if AI adoption raises productivity, firms can cut prices and boost sales, which in turn creates more jobs.
AI may also spawn new businesses. A would-be entrepreneur who needs a web programmer and an accountant can now use AI for both, lowering startup costs. Rajan pointed to a pandemic-era rise in new business formation that has continued in recent quarters.
David Autor's work, cited by Rajan, suggests AI can equip moderately skilled workers with higher-order abilities – a nurse practitioner using medical AI to diagnose more conditions is one example. Given the enormous global demand for healthcare and other services, further job creation is possible.