The warnings are piling up. In the past week, IMF chief Kristalina Georgieva has said that artificial intelligence is “hitting the labor market like a tsunami”; JPMorgan Chase boss Jamie Dimon has predicted that America’s biggest bank will soon need fewer employees; And Dario Amodei, who runs Anthropic, has predicted that the technology his company is at the forefront of developing could eliminate “half of all entry-level white-collar jobs.”
AI can really wreak havoc white collar workforce. But rather than making such jobs less attractive – let alone redundant – there is potential to reshape them. The AI office will look less like a robot and more like a cyborg, combining the best human and computer capabilities: the Six Million Dollar Man instead of the Terminator. To see why, consider what has happened to white-collar work over the past three years, how it compares to earlier technological revolutions, and what these patterns suggest about what’s to come.
Despite all the concerns, white-collar workers are still doing well. The US has added nearly 3 million white-collar jobs through the end of 2022 – including management, professional, sales and office roles – while blue-collar employment has remained stagnant (see Chart 1). Some businesses are often seen as early victims of AI. There are 7% more software developers, 10% more radiologists and 21% more paralegals in the US than three years ago. The slowdown in hiring for some entry-level white-collar jobs recently detected by academic research appears to predate ChatGPT and may therefore have more to do with rising interest rates and the unpredictable global business environment.
Salaries of professionals have also been frozen. Real (inflation-adjusted) wages in professional and business services (think salespeople, accountants and the like) are expected to increase by 5% through the end of 2022. Office and administrative workers earn 17% more. Taking into account education, age, gender, race, and other characteristics, we calculate that white-collar workers now earn one-third more than blue-collar workers (see Chart 2). This is almost three times the premiums in the early 1980s and has continued to rise over the past three years. In other words, AI has not yet stripped office workers of their sustainable wage gains.
These findings will not surprise historians of technological change. The early years of the computer age were also marked by dire predictions of mass displacement. In 1982 Nobel-prize winning economist Vasily Leontief warned that “the relationship between man and machine is changing fundamentally”, as computers began to take on “first simple and then increasingly complex mental tasks”. In such a situation, digital automation proved to be a boon for office work. Employment in management, professional, sales, and office roles has more than doubled since the early 1980s, and their wages have increased by nearly a third after adjusting for inflation.
One reason white-collar work flourished in the earlier digital age was that computers rarely replaced entire jobs at once. They automated routine and repetitive tasks – those that could be codified into clear rules and executed by machines. When a job becomes completely routine and repetitive, it can disappear (as happened with typists). But most professional roles are sets of tasks, only some of which can be automated. The result was not replacement but upgrade: computers increased productivity and directed human effort toward higher-value activities such as analysis and decision-making. Air-traffic controllers illustrate the pattern: Software helped process flight data, humans retained authority over high-stakes decisions, salaries increased.
computer aided dominance
More importantly, by increasing productivity and cutting costs, computers have also expanded the range of activities that companies can perform profitably. E-commerce has created new work in logistics, supply-chain planning and digital payments. Smartphones created app designers. Social media ushered in digital marketers and influencers. As a result, there was a steady increase in white collar employment. According to Daron Acemoglu of the Massachusetts Institute of Technology and Pascual Restrepo of Boston University, about half of America’s employment growth between 1980 and 2010 came entirely from the creation of new businesses.
AI is smarter than older digital technologies. But the same logic of technological change seems likely to apply. For one thing, today’s artificially intelligent systems have what AI scholars call “jagged intelligence,” exhibiting uneven and inconsistent performance. It’s not enough to be 95% good at a task when the remaining 5% involves side cases and discretion. Anthropic’s evidence, based on millions of unknown interactions with its models, bears this out. Only about 4% of businesses use AI in three-quarters or more of their operations; Hardly anything can be fully automated. Like computers, AI is reducing the costs of specific cognitive activities – drafting text, writing code, gathering information or running standard analyzes – rather than replacing entire roles.
Recent labor-market data support this view. We analyzed employment and wage trends at more than 100 large white-collar occupations in the US since the second half of 2022. Employment increased by 4% and real wages increased by 3% across the entire sample. To understand the impact of AI on different roles, we used occupational descriptions to classify white-collar roles into four groups based on the tasks involved: technologists; Managers and Coordinators; care workers; and back-office staff. We then tracked employment in each group starting in late 2022, using a six-month moving average.
Roles that combine technical expertise with oversight and coordination have reaped the greatest benefits. Employment among project managers and information-security specialists has increased by 30% or more. Other occupations that combine deep expertise in mathematics-related fields with problem-solving are also thriving (see Chart 3). So are jobs that involve interpersonal care work and that demand judgment and coordination (see Chart 4). Only routine back-office work has been reduced. The ranks of U.S. insurance-claims clerks have shrunk by 13% and the ranks of secretaries and administrative assistants have shrunk by 20% over the past three years.
AI is already creating new jobs. Companies are hiring “data annotators” to label digital information so AI can parse it, “forward-deployed engineers” to guide customers through AI implementations, and, in the C-suite, “chief AI officers.” Indeed, the fastest-growing white-collar occupations in recent years are those without established names. “Other mathematical-science occupations” have seen their ranks grow by almost 40% since the end of 2022 and their real salaries increase by almost a fifth. “Other computer occupations”, such as systems architects and IT project managers, have also expanded rapidly. Employment among “business operations specialists, all others” – combining process design, coordination and analysis – has increased by nearly 60%, with corresponding solid salary increases.
This doesn’t mean that all white-collar workers can sleep easy. In the subset of task bundles that involve few edge cases and little discretion, AI may soon be able to automate much. Already new models can perform autonomous tasks for many hours by combining coding, analysis and tool use with limited human input. Benchmarks created by METR, a research group, show that AI can write software on its own for up to five hours continuously, and this figure doubles approximately every seven months. Anthropic’s Mr. Amodei thinks AI may be able to do much of the work of software engineers by this year.
Entry-level jobs appear insecure for similar reasons. The same happened with those who were victims of earlier technological upheavals. The share of Americans in clerical and administrative work, which had already declined from 18% to 10% in the 1980s, continues to shrink. New research from Sam Manning and Tomás Aguirre of the think-tank Center for the Governance of AI shows that such workers have the weakest ability to adapt, fewer transferable skills and fewer chances to move into higher-value jobs.
Such disruption would be painful for those affected. But it is far from the labor-market disaster that some had predicted. Combining human intelligence with machine intelligence is likely to generate more value than AI alone for some time. Human accountability and participation will retain a premium in the market. And white-collar workers have shown themselves to be highly adaptable. AI will once again recreate their jobs. But it will not eliminate them.
