AI will supercharge productivity. Will workers benefit?

Productivity growth, or the ability to produce more per hour, is supposed to make everyone richer. The idea is that greater productivity allows companies to make more money, which workers and owners share through higher wages and more valuable businesses. But since the 1980s, productivity gains have gone almost exclusively to executives and owners of companies, leaving average workers behind and fueling the widest wage and wealth gaps on record.  

Enter artificial intelligence, which promises greater productivity growth than any technology before it. If AI delivers, and those productivity gains continue to elude everyday workers, wage and wealth gaps will widen further, perhaps significantly, adding to the burdens that high rates of economic inequality are already placing on the economy, the labor market and the political and social environment. It doesn’t have to be this way, and now is the time to consider policies that would help everyone share in AI’s anticipated bounty.

The divergence between productivity growth and pay raises, which has swelled over the past four decades, is well known to economists. I’m citing numbers from the Economic Policy Institute, but they’re roughly the same no matter how one slices them. What they show is that productivity and compensation for ordinary workers grew in near lockstep from the end of World War II through the 1970s. Since then, however, productivity has grown nearly four times faster than pay for ordinary workers, the difference going to shareholders and the most highly paid workers.

The results are glaring. Wage ratios — the difference between the highest- and lowest-paid workers — have surged in recent decades. The best known among them, the CEO-to-worker pay ratio, climbed to an astounding 399-to-1 in 2021 from just 20-to-1 in 1965. The US’s Gini index, which measures the degree of income inequality, has trended sharply higher since the 1980s and is now the highest among developed countries. Wealth gaps, too, show a wide and growing divergence between the richest Americans and everyone else.  

The notion of two Americas divided along economic lines is quickly becoming a reality. US private schools bestow a world-class education and a glide path to elite universities on a fortunate few, while public schools struggle to teach basic reading, if they can find teachers. Wealthy Americans are turning to concierge medicine, which largely operates outside the impenetrable US health-care system, while everyone else struggles to find a doctor, if they can afford one at all. Private planes shuttle rich travelers around while ordinary Americans are crammed into ever tighter commercial airlines, assuming they earn enough to travel. Increasingly, the richest Americans have less and less need or opportunity to encounter their less fortunate countrymates. 

There are good reasons to worry that AI will deepen those divides. The previous two productivity booms — led by personal computers and the internet — helped concentrate market share across industries, whether it was Apple Inc. in personal computing and smartphones, Alphabet Inc. in web search, Microsoft Corp. in business software, Meta Platforms Inc. in social media or Amazon.com Inc. in online retailing and cloud computing. Less competition has made the winners bigger, more profitable and more powerful, enriching owners and executives while they push around workers.  

Those same companies have a big lead in the AI race. Only this time, AI also threatens to displace highly paid workers, from engineers at big technology companies to professionals such as lawyers, consultants and money managers. “This is going to be very different from the past 40 years, when blue-collar workers lost out and white-collar workers benefited from technology progress,” Anton Korinek, an AI researcher, told Bloomberg News. “This is a reversal where white-collar workers are the ones that are easier to automate now.”

If true, the number of workers excluded from technology-led productivity growth is set to grow — and possibly to a much greater degree than in the past. Korinek co-wrote a new paper from the Brookings Institution estimating that AI may increase productivity by as much as 2.3% to 3.3% a year over the next 20 years, well higher than the Congressional Budget Office’s projection of 1.5% a year and realized productivity growth of 1.2% a year since 1980. Imagine triple the productivity growth shared by an even smaller percentage of Americans; it would make today’s inequality look like a socialist utopia.  

Some argue that the degree of inequality is unimportant if everyone is doing well. But that’s far from reality because, by any count, tens of millions of full-time US workers don’t earn a living wage. So it’s not just that economic inequality is high; it’s that even before mass adoption of AI, an alarming number of workers struggle to survive and have little hope of building any wealth.

The good news is that there are policy interventions that could help narrow wage and wealth gaps. The US can adopt a system of co-determination like the decades-old one in Germany in which workers are represented on companies’ boards to ensure that they have a say on pay. Financial regulators can require companies to disclose compensation data so that the impact of AI on wages can be measured and addressed. That data would allow policymakers to devise targeted incentives for companies to pay workers a living wage. The US could also establish a sovereign wealth fund that would invest in AI and use some of the profits to aid workers displaced by the bots.

AI may very well usher in the next productivity boom. But if it isn’t shared more broadly than the technology-charged gains of the past four decades, economic inequality — with all its attendant harms — will deepen.