Amazon.com Inc.’s latest global layoffs should come as a singular warning to India. For policymakers dealing with the world’s largest youth population, AI suddenly poses a very real risk to jobs, wages, and a white-collar future.

The e-commerce and cloud services giant’s elimination of 14,000 corporate positions worldwide may not have a large direct impact on its sizeable Indian workforce. The more worrying thing is the kind of occupations at risk: Generative artificial intelligence is starting to affect more than just entry-level computer programming.

Outsourcing hubs like Bengaluru and Hyderabad are already feeling the pinch from AI. But Amazon’s cuts may affect finance, marketing, human resources and tech employees, according to local media reports. That puts many more sectors on notice and validates a growing body of academic work.

After parsing nearly 200 years of data on labor markets and technological change, finance scholars at Northwestern University and the Massachusetts Institute of Technology have concluded that advances in natural-language processing may favor occupations that are lower-educated, lower-paid, and more male-dominated, such as construction and trucking. 

It would be a dramatic departure from how previous innovation affected demand for workers. As Huben Liu and his coauthors explain, until the 1980s IT revolution, most advances in automation supplanted manual effort while supporting cognitive tasks. Take, for instance, Irving Colburn’s early-20th-century invention of a machine to substitute hand-blown glass in window panes. The blowers’ wages fell 40%. Within one generation, mechanization drove an entire class of artisans out of business.

By contrast, the arrival of electronic calculators in the 1970s helped accountants and auditors to become more productive. It didn’t replace them. The tilt toward services such as finance and health care favored women, facilitating their entry into the workforce as 20th-century innovations also eased the burden of domestic chores.

Over time, these improvements went global, but the hard-won gains may now reverse. With the capital costs of implementing AI expected to become cheaper each year, cognitive tasks that don’t require at least five years of specific vocational preparation will be at risk from automation, the researchers say. That includes many entry-level jobs, such as analyzing financial statements at Wall Street firms.

Mechanized production of sheet glass did little to hurt women. At the cusp of automation in 1900, they held few of the 53,000 jobs in the US glass industry. Employers preferred men. (In 1900, the industry employed twice as many children under 16 as women.) But to lose out now to Lilli, McKinsey & Co.’s proprietary AI tool that’s drafting client proposals and preparing slide decks? That would certainly rankle, especially since it’s named after the first woman professional hired by the consulting firm in 1945. 

All this may come as a particularly harsh blow to the 375 million Indians who are between 10 and 24 years old. At 18.5%, youth unemployment in cities is alarmingly high. Young women’s participation in the labor force is abysmally low at under 22%. Large-scale adoption of AI tools by companies will further muddy the picture. In a separate paper, London School of Economics professor Luis Garicano and his coauthor examine a realistic scenario: If AI does away with entry-level grunt work, which employer will bother to train fresh graduates? How will they rise up the career ladder to higher-wage positions?

Artificial intelligence may still surprise us by creating new tasks that don’t yet exist. It’s also possible that young people will invest in their own AI training. But if Amazon is any indication, the technological exposure of higher-educated, better-paid, and more women-oriented occupations is indeed high.

This won’t be the first shock to India’s labor market in modern times. Its cotton spinners and weavers, among the world’s best in the early 18th century, took a large hit from the Industrial Revolution. As the economy struggles to move from lower-middle to higher-middle income, AI is threatening its biggest advantage: the youth bulge it enjoys against other countries that are rapidly aging. 

The right approach to AI would contain both carrots and sticks. The preponderance of Chinese large language models among the world’s top 20, as highlighted by my colleague Catherine Thorbecke, makes it obvious that India isn’t doing enough fundamental research. This must change. The government also needs to read the riot act to outsourcing firms. They have to halt share buybacks and invest in meaningful AI projects, not just data centers.

Finally, the broader corporate sector should be given generous tax breaks for research and development. Instead of coming up with generic copies of drugs going off patent in the West, pharmaceutical companies must be encouraged to use AI to discover new molecules.

The next quarter-century offers the most-populous nation a chance to get rich before it grows old. Ending up on the wrong side of technological change for the second time in 300 years won’t be a good outcome — either for India, or the world. Amazon’s job cuts are the proverbial canary in the coal mine. The time to act is now, before the outlook for white-collar work turns more toxic.