Mass Layoffs Are Scary, but Probably Not a Sign of the A.I. Apocalypse

Link to New York Times article

Despite fears that Amazon and other employers are already replacing workers
with bots, the A.I. transition is likely to play out differently.

Amazon’s announcement last month that it was cutting 14,000 corporate positions included an alarming warning for those worried that the artificial intelligence apocalypse may have arrived.

A.I. is “enabling companies to innovate much faster than ever before,” a senior vice president wrote in a message to employees that the company shared publicly. And to realize this potential, the executive suggested, “we’re convinced that we need to be organized more leanly, with fewer layers.”

The announcement, combined with other layoff news from companies like Target, UPS, Microsoft and IBM, has prompted some to suggest that the economy has entered a period of A.I.-driven restructuring.

But the A.I. apocalypse is probably not here just yet. While the technology may have played an indirect role in some of the downsizing, experts say the transition to an A.I.-powered workplace most likely won’t entail large-scale layoffs in which employers dump humans in favor of machines. Rather, the transition is likely to be more gradual, in many cases occurring as new companies, built to exploit A.I., take market share from more established companies that are slower to embrace it.

“Widespread adoption is going to happen at the new firms,” said Mert Demirer, an economist at M.I.T. “It’s always the case that the smaller the production process, the more the process is easier to change.”

Among established companies, those in the tech industry appear to be furthest along in adopting artificial intelligence. Employers like Amazon, Microsoft and Google have made a number of A.I. tools available to their white-collar employees, such as A.I. assistants that suggest lines of code, so-called agents that can generate whole sections of computer programs, and chatbots that can produce drafts of memos and reports.

But employees at Amazon, where the announced job cuts affected less than 5 percent of corporate workers, say the adoption of these tools has been uneven across different teams and organizations. And thus far, the layoffs and buyouts at big tech companies do not appear to have been driven by the automation of white-collar jobs directly.

“We do think that at some point A.I. tools will allow us to enhance productivity to a point that we’re going to need less labor, but we’re not there yet, not in any significant way,” said Gil Luria, an analyst who covers Microsoft and Amazon for the investment bank D.A. Davidson.

Instead, he added, the companies appear to be making the cuts partly to hold their overall profit margins steady while they spend tens of billions of dollars on A.I. infrastructure like data centers. Cutting back on employees is a way to convince shareholders that the companies are “investing in a responsible manner,” Mr. Luria said.

Amazon said that different companies were funding their A.I. investments in different ways, and that while some couldn’t afford their investments, Amazon could.

The use of A.I. tools appears to be even less intensive outside technology companies. In a recent survey by McKinsey, the consulting firm, almost 80 percent of companies reported using generative A.I., but about the same number reported that the tools had not significantly affected their earnings. A study released this summer by researchers at M.I.T. reached a similar conclusion, finding that industries other than technology and media showed “little structural change” as a result of A.I.

In some cases, laying off workers is less about automating their jobs today than gambling that they won’t be needed in the future. Over the past few years, many employers have engaged in “labor hoarding” — hanging on to workers they no longer need because they may have use for them in a year or two and don’t want to go through the trouble of hiring again. But given the possibility that advances in A.I. will reduce the need for workers on that timetable, these companies feel more comfortable laying them off.

“The whole motive of labor hoarding is that you’re going to need workers when demand picks up again,” said Benjamin Friedrich, a labor economist at Northwestern University’s Kellogg School of Management. “You want to be ready to go.”

It’s unlikely, however, that big established companies will be able to substitute A.I. for large numbers of workers over the next year or two. One reason is that big companies are by their nature plodding and bureaucratic when reimagining their work processes. The McKinsey report observed that many companies’ flirtation with A.I. had involved “a proliferation of disconnected micro-initiatives” that suffered from “limited coordination.”

But a more important reason has to do with a deeper conservatism: Established companies tend to use new technology to do what they’ve always done, only somewhat faster and less expensively, said Andrew McAfee, a principal research scientist at M.I.T.’s Sloan School of Management. They don’t tend to rethink their entire structure.

By contrast, new companies often ask how best to organize themselves when starting from scratch, without the employees or rituals that the new technology renders obsolete.

Dr. McAfee, who is also a founder of Workhelix, an A.I. start-up, cited the electrification of factories that began in the late 19th century as an example. During the first few decades of electrification, he said, many factory owners simply began powering their machines with electricity rather than steam. But they didn’t reorganize their processes. It was only when entrepreneurs reimagined the factory to incorporate new layouts and new processes like assembly lines that electricity brought enormous productivity gains.

Something similar is likely to play out with A.I.: a long period of marginal changes at established companies before new businesses eventually change the way work is staffed and organized.

In the legal profession, for example, large firms have long deployed teams of partners and several associates, billing clients for each lawyer’s services at an hourly rate. But some entrepreneurs have recently started companies that rely on vastly fewer legal experts per client, supporting them with A.I. instead.

“Professional services firms are still generally operating on a time- and materials-based business model, which disincentivizes using A.I. to cut hours,” said Omar Haroun, chief executive of the parent company of Eudia Counsel, a so-called A.I.-native law firm. “We’re actually trying to prove that one knowledge worker can do the work of 10.”