[Kangkook Lee's Column] How Many Jobs Does AI Take Away? ... First, 'Youth Employment' Is Shaking

Dr. Kangkook Lee is Professor of College of Economics at Ritsumeikan University in Japan and Affiliated Research Fellow at CUKPE

We often hear in the news that tech companies are laying off workers on a large scale due to the advancement of artificial intelligence (AI). When using generative AI that has become smarter than humans, it seems we have entered an era where people are no longer needed for tasks such as data collection, analysis, report writing, and software development. Furthermore, if robots become smarter thanks to AI, won't the problem of future unemployment become truly severe?

To what extent will AI impact unemployment?

First, despite such news, evidence that AI is taking away human jobs remains scarce to date. A recent report by the U.S. think tank Economic Innovation Group examined changes in unemployment rates by dividing U.S. workers into several groups based on their exposure to AI. The analysis found that the unemployment rate was generally lower for the group with higher AI exposure, and there was no difference in the change of unemployment rates depending on the level of exposure. Even after the launch of ChatGPT in late 2022, the unemployment rate of the high-exposure group did not rise relatively further.

There was no difference based on AI exposure in cases where individuals left the labor market to become economically inactive or changed jobs. The same was true for data on employment by industry, youth, and recent graduates. A study by Yale University’s Budget Lab also stated that no changes in the use of AI on job structure or unemployment have yet been observed. A Goldman Sachs report from April of this year also analyzed industry-specific data and pointed out that AI has no statistically significant association with unemployment or changes in employment.

According to another study based on a survey of companies in the U.S., U.K., and Germany at the end of 2025 , approximately 90% of companies responded that AI had no impact on employment over the past three years since the launch of generative AI. While they predicted a minimal decrease in employment over the next three years, the majority still indicated that there would be no overall effect. In fact, The Economist recently reported that related jobs have actually increased due to rising investment in AI, particularly the boom in data center construction. Jobs created by AI investment, such as electricians, number around 1 million—far exceeding the approximately 200,000 jobs laid off due to AI—leading to the conclusion that we are currently experiencing an "AI job boom." Furthermore, alongside the advancement of AI, jobs in customer service and accounting assistants decreased between 2023 and 2025, while professional jobs such as data scientists and legal assistants increased.

A Federal Reserve report released in July of this year also states that the impact of AI on the overall labor market is not yet distinct when looking at indicators such as youth unemployment rates and labor market participation rates. However, according to this report, layoffs in the information industry, which heavily utilizes AI, have increased relatively significantly since 2025, while job openings have decreased further. Another Fed report released in August revealed that since 2023, it has become more difficult for the unemployed in the U.S. who held jobs with greater exposure to AI to find employment. Nevertheless, the impact of AI-induced unemployment on the labor market as a whole has not been pronounced to date.

The biggest victims of AI are young people

However, it appears that at least youth employment is taking a hit from AI. Studies have been released one after another indicating that as generative AI spreads, jobs for junior employees have decreased relatively significantly. For example, in a paper titled "Canaries in the Coal Mine?", Stanford University Professor Brynjolfsson analyzed vast payroll data from ADP and revealed that as exposure to AI increased, youth employment declined significantly, while jobs for experienced senior employees increased. In the software development sector, jobs for the 22–24 age group decreased by about 20% over the approximately 18 months following October 2022, while jobs for the group aged 30 and older continued to increase. Another study analyzing tens of millions of LinkedIn resumes also reports that entry-level jobs have decreased relatively significantly since the spread of AI, attributing this primarily to a decline in new hiring. A report by Goldman Sachs presents similar results.

The situation is no different in Korea. In a report published last year, the Bank of Korea analyzed National Pension Service data and revealed that for approximately three years following the spread of generative AI, employment declines were more pronounced among the youth group in their 20s as their exposure to AI increased. During this period, 211,000 jobs for young people were lost, with about 99% of these occurring in sectors with high AI exposure. These sectors include computer programming, publishing, and professional services; however, even in these industries, jobs for those aged 30 and older continued to increase. In other words, this indicates that "seniority-biased" changes are appearing in the domestic labor market, particularly in sectors with high AI exposure.

The reason the advancement of generative AI is having a particularly significant impact on youth employment is that it replaces structured knowledge work, such as data collection and report writing, which are primarily handled by these workers. In particular, AI is having a greater impact on highly educated workers by replacing more of their duties and reducing their working hours. On the other hand, senior employees with long careers possess tacit knowledge and social skills that are difficult for AI to replace, such as understanding work context and organizational management; therefore, the active use of AI could actually play a complementary role in their work. This may be somewhat related to the reality of the severity of youth employment issues, as evidenced by the fact that as of September 2026, the youth employment rate has been declining year-on-year for 26 consecutive months.

However, there are also voices suggesting that the reason AI is reducing jobs, particularly for young people, may be a response by companies seeking to hire experienced employees to handle uncertainties such as tariffs, war, and inflation. An article in the Financial Times last July reported that as companies adapt to the use of AI, entry-level employment is not simply decreasing; rather, the nature of work is shifting toward more complex tasks, such as data-driven decision-making and human resource management that effectively utilize AI. Research by Professor Brynjolfsson also reports that jobs with the highest exposure to AI have increased significantly in professions where AI complements and enhances labor. A recent study using corporate data reports that companies that spent heavily on AI subsequently increased employment more, including hiring more entry-level young employees. If so, the issue of worsening youth employment due to AI may require further observation. To prevent a decline in jobs for young people, education and institutional support are needed above all else to help them develop the ability to effectively utilize AI.

Still, won't it lead to mass unemployment in the future?

While AI appears to have had a limited negative impact on youth employment so far, could the future be completely different? In fact, Professor Hinton, the godfather of AI, has predicted that mass unemployment will occur soon due to the rapid advancement of AI technology. Based on this perspective, the AI ​​Futures Group's AI 2040 report forecasts that after 2030, AI will very rapidly replace office work, and robots will replace production work.

According to this institution, the share of AI and robots in total labor is projected to rise rapidly from a maximum of 20% in 2032 to a maximum of 85% in 2035 when considering economic value, and up to 95% when considering only actual labor. Along with this, the authors predict that the economic growth rate will also jump significantly, reaching 50% by 2032. Furthermore, they forecast that when this era arrives, the government will generate massive revenue by granting companies permits to use robots and computing resources, from which citizens will be able to receive a "citizen dividend." This is a highly optimistic outlook and is similar to Elon Musk's argument that everyone will receive a basic high income even if jobs are lost in the AI ​​era.

However, the prospect of such rapid change occurring in the early 2030s sounds rather unrealistic. Looking back at history, while the advancement of automation technology led to the disappearance of some jobs, it did not result in mass unemployment across the entire economy. This is because new jobs continued to emerge even after the introduction of machines, computers, and industrial robots following the Industrial Revolution. According to Professor Otter of MIT, concerns about a so-called "Robocalipse"—the mass loss of jobs due to robots and automation—have always been raised, but they have never become a reality. This is because while automation brings about job displacement, it can simultaneously increase productivity and income, thereby raising the demand for labor, and automation technology also creates a reinstatement effect by generating new tasks that require labor. For example, this means that technologies like computers eliminate some jobs but create other jobs related to using or developing them.

Therefore, a report by the renowned venture capital firm Andreessen Horowitz (a16z) emphasizes that the doomsday of jobs caused by AI is nothing but an illusion. This doomsday theory is based on the mistaken notion that workers and AI are engaged in a "zero-sum competition" because the amount of labor to be performed is fixed. Historically, while the mechanization of agriculture drastically reduced the number of farmers and electrification significantly increased productivity, the economy was reorganized and more new jobs were created. The introduction of Excel reduced simple accounting assistant jobs but led to a significant increase in jobs for accountants and financial analysts, and along with rising incomes, all kinds of new service industry jobs, such as nail salons, emerged. In fact, as Professor Otter's research shows, as of 2018, approximately 60% of all jobs were new positions that did not exist in 1940.

In the case of AI, while it may eliminate jobs such as customer service or insurance clerks, roles in educational administration or management decision-making could increase as they benefit from the use of AI. Furthermore, this report emphasizes that the expansion of new knowledge jobs related to AI may be just beginning. Ultimately, just as history has, I believe that new jobs will continue to emerge alongside the development of AI for a long time to come. Of course, the future may differ from the past, and if AI develops to an extreme degree, it might be able to replace every conceivable human task. However, even if such an era arrives, it would likely be in the very distant future.

So far, we have examined various research findings on whether AI is currently taking away jobs or will lead to mass unemployment in the future. While it is difficult for economic studies to demonstrate, the most important thing is not to forget that the impact of technological change on the labor market is always a human affair. For example, if AI were to eliminate jobs without countermeasures, it would lead to social resistance, which could result in changes to policies and institutions and influence the effects and direction of technological change. Therefore, instead of worrying about mass unemployment, what we need now is an effort to develop pro-labor AI that increases labor demand rather than merely replacing humans.

(This article was originally published as a column in Hankyoreh in Korean and translated into English with the help of Google Translate. The views expressed in this article are those of the author(s) and do not necessarily represent the official stance of the center.)