Editor's Note: Whether AI will bring about an 'employment apocalypse' is becoming one of the most pressing questions in the business world. David M. Solomon, Chairman and CEO of Goldman Sachs, argues in an article for The New York Times that this concern is exaggerated. AI will indeed impact the labor market, particularly the repetitive tasks in white-collar jobs such as accounting, banking, law, software engineering, and customer service. However, it is more likely to change the nature of work rather than simply eliminate a vast number of jobs.
Solomon's core argument is: AI is not automating 25% of jobs, but 25% of work hours. In other words, some inefficient, repetitive tasks will be taken over by machines, pushing humans toward more complex tasks that rely on judgment and client interaction. Meanwhile, new demands such as data center construction, AI workflow management, and compliance verification are already creating new employment opportunities.
What this article truly aims to address is an old question within technological transformation: each new technology brings pain, but historically, economies have often been able to recreate jobs after the initial shock. The risk of AI does not lie in it necessarily causing unemployment, but in whether society, businesses, and the education system can promptly help workers complete the transition.
The following is the original text:
Over the past few months, having spoken with hundreds of business leaders, I’ve noticed a sharp divergence in their views on artificial intelligence. One camp believes an 'employment apocalypse' and mass unemployment are imminent; the other believes AI will propel society forward in a great leap.
I belong to the latter group—with some caveats, of course. Will AI disrupt the labor market? Undoubtedly. This transition, like other major shifts in history, will present new challenges, especially as AI detaches labor from productivity on an unprecedented scale. But the U.S. has a longstanding ability to create new jobs following technological shocks, from electrification in the early 1900s to the digital revolution of the 1990s. I see no reason to believe this dynamic will stop today.
There is no doubt that AI will reshape our daily lives. Goldman Sachs economists estimate that over the next decade, AI could automate up to 25% of current work hours. The impact on professions requiring hands-on, on-site operations—like food preparation, construction, and services—remains hard to judge. But in white-collar jobs, including accounting, banking, and law, many tasks are likely to be automated. A Stanford University study shows that in the occupations most susceptible to automation, such as software engineering and customer service, employment for entry-level positions has already fallen by 16% compared to the least affected roles.
Yet, looking at jobs or industries less linked to automation, the picture changes. Our economists estimate that since 2022, the growth in data center demand has created over 200,000 construction jobs. While AI may eliminate jobs in some sectors, it could also spur job growth in others. For instance, Goldman Sachs may eventually need fewer people handling regulatory reports or client onboarding processes. But this could free up space for us to hire more bankers, traders, and asset managers who engage in continuous client interaction.
Of course, we cannot ignore the real human cost behind this disruption. The Industrial Revolution did raise living standards, but only after society endured grueling labor in factories and mines, and the foul slums brought by rapid urbanization. In recent decades, the significant decline in manufacturing employment due to automation and global outsourcing has caused immense hardship for many American families and communities, such as in Gary, Indiana, and Greenville, South Carolina.
Yet despite these challenges, I keep returning to one reality: the standard of living for the vast majority of Americans is significantly higher than in the past. I was born in 1962, when the average American adult did not have air conditioning; later, as its price fell, nearly everyone enjoyed cool air. In the 1950s, only large companies like IBM owned computers; today, roughly 90% of American adults hold a supercomputer in their hands. In 1900, global life expectancy at birth was 32 years; today, it exceeds 70.
Perhaps more crucially, job growth has outpaced population growth. Since 1962, U.S. civilian employment has increased by about 145%, while the civilian population aged 16 and over has grown by approximately 128%. During this period, some new industries emerged, while others expanded or declined. Manufacturing employment fell from 15.5 million to 12.5 million, with textile and apparel manufacturing losing nearly 2 million jobs. Meanwhile, the healthcare industry now employs over 18 million people. The U.S. economy remains the world's most innovative, dynamic, and entrepreneurial.
It's true that even the most reliable historical patterns can be broken. But I believe the U.S. economy will remain resilient and dynamic as ever, for three reasons.
First, if our estimates are correct, AI will not eliminate 25% of jobs. What is more likely is that people will find more efficient ways to allocate their time. When I was a first-year banking analyst, creating a simple stock performance chart took six hours, searching through microfilm archives of The Wall Street Journal for prices. Today, a first-year analyst can do it in seconds, and in recent years, we've hired more people than ever before. As tools advance, so does the natural complexity of work. Despite the convenience brought by Excel, email, and Zoom, do any of us really feel like we have less to do now?
Second, even if a job can be replaced, it doesn't mean it will be. Television didn't eliminate the demand for live entertainment, and the internet hasn't made real estate agents or fitness coaches unemployed. Instead, these technologies highlighted and reinforced the value of these professions. Technological change and cultural change do not advance in lockstep. After all, even after decades of ATMs, digital banking, and industry consolidation, employment in commercial banking today is roughly at the same level as it was in the mid-1990s.
Third, the U.S. labor market itself is dynamic. While annual net job creation is at most a few million, the gross flows are much larger; American businesses destroy and create 25 to 35 million jobs each year. One can imagine this pace accelerating as AI drives more innovation, and we are already seeing the economy begin to adapt. Businesses are now seeking talent who can manage so-called 'agentic AI' and apply it across a wide range of scenarios, from execution and workflow to compliance and verification. All of this requires human judgment.
If AI does destroy jobs, and possibly faster than before, then public policy must respond: either by funding large-scale retraining or by encouraging the development of AI that supports workers rather than replaces them.
This must be a joint effort between the public and private sectors. The public sector should provide incentives and resources where necessary, including increased investment in vocational schools and community colleges; the private sector should help employees upskill and redesign on-the-job training systems.
The historical pattern is clear: the U.S. economy can and will adapt to major technological advances. It's also clear that even the gravest predictions from the brightest minds often prove inaccurate. In 1930, John Maynard Keynes famously predicted that by 2030, people would need to work only 15 hours a week. Although his envisioned future of abundant leisure hasn't materialized, it remains a good reminder: fears of an 'employment apocalypse' likely underestimate AI's potential to drive an economic and productivity renaissance.
In addition to leading Goldman Sachs, David M. Solomon is an electronic dance music producer known as DJ D-Sol.
