Employment in the information and communications industry and the professional, scientific and technical services industry fell by about 147,000 as of February 2026 compared with the same month a year earlier. An examination of the microdata from the National Data Office's Economically Active Population Survey shows that most of this decline is concentrated in specific age groups. Workers in their 20s accounted for 97,000 and those in their 30s for 34,000. Together that is 131,000, or about 89% of the total decline in the two industries.
By industry, professional, scientific and technical services fell by 105,000 and information and communications by 42,000. The former includes research and development, architecture and engineering, and legal and accounting services, while the latter includes software development, computer programming and information services. On a February basis, this is the first time in five years since 2021, when the COVID-19 shock was severe, that both industries have moved backward at once, and the decline is the largest since the 2013 revision of the industrial classification.
When the economy worsens, employment usually falls across all age groups together. That was not the case this time. Over the same period, workers in their 40s fell by 32,000, but those in their 50s rose by 12,000 and those aged 60 and over by 2,000. The share of workers in their 20s and 30s among those employed in the two industries dropped from 51.7% to 49.5% in one year.
Breaking the ages down further sharpens the outline. The late 20s bracket (ages 25-29) alone lost 81,000, while the early 20s declined by only 16,000. The early 30s fell by about 50,000 in professional services but rose by 14,000 in information and communications, and the late 30s moved in opposite directions across the two industries, with an increase of 15,000 in professional services and a decrease of 13,000 in information and communications. That means the late 20s bracket, which corresponds to the first job after graduating from university, was carved out particularly deeply.
The Bank of Korea's report "The Spread of AI and the Contraction of Youth Employment," released last October, points to the same place. Over the three years from July 2022, after the release of ChatGPT, to July 2025, employment among those aged 15-29 fell by 23.8% in information services, 20.4% in publishing, 11.2% in computer programming, systems integration and management, and 8.8% in professional services. The industries with high rates of decline are generally those where people had been doing work that generative AI does well.
The explanation for why it begins with new hires lies in the nature of the work. Oh Sam-il, head of a team at the Bank of Korea's Economic Research Institute, explained that the work handled by junior staff consists of standardized tasks learned in school and is easily replaced by coding, while senior staff's interpersonal relations, organizational management and work coordination are not being replaced. AI has absorbed most quickly the kind of "learning while doing" work such as organizing requirements, drafting documents, and compiling precedents and materials. One law firm stopped the new hiring it had carried out every year, beginning this year, after its processing times shortened with the introduction of AI, according to a lawyer at the firm.
There is a caution here. These statistics do not distinguish between people who could not get in because hiring was reduced and people who left the places they had been working. Nor has any quantitative analysis from the government or research institutes separated economic factors from AI factors.
In the United States as well, of the 1.17 million announced job cuts from January to November last year, those citing AI as the reason numbered 54,694, or about 5%. Even allowing for the fact that companies do not go out of their way to record AI as the reason for job cuts, the grounds are thin for asserting causation.
Even so, the implication left by the age distribution is clear. Skills accumulate through the repetitive work of the first three years, and if those three years are reduced wholesale, the reservoir of people who would become middle managers five years later dries up as well. AI, which cannot replace senior staff, is cutting off the path that produces them. The signal that the number of young people looking for two or more jobs has reached its highest level since COVID-19 does not appear unrelated to this trend.
The training needed by those about to enter the job market is the kind that moves them quickly into the role of reviewing and coordinating the output AI produces. Learning skills that avoid AI is not the answer. Companies need accounting that counts new hiring as workforce inventory five years from now rather than treating it as a cost.
When the March employment statistics come out, it will become clear whether February's figures were a one-month fluctuation or a trend. A jobs forum with participants including Seoul National University professor Kim Dae-il will be held in Jung-gu, Seoul, on April 8.
