This picture taken on November 12, 2018 shows job seekers attending interviews for employment during a jobs fair in Seoul.
JUNG YEON-JE/AFP via Getty Images
Four years after ChatGPT reshuffled the global technology landscape, South Korea’s central bank has released what may be the most granular government-level accounting yet of generative AI’s toll on young workers — and the numbers confirm a generational inversion no labor economist had yet documented at this scale.
A report published Tuesday by the Bank of Korea (BOK) found that 268,000 of the 285,000 net jobs lost by workers aged 15 to 29 between June 2022 and June 2026 were concentrated in sectors with high AI exposure, representing 94 percent of all net youth job losses over that four-year period. The sectors hardest hit — IT services, publishing, computer programming, and professional services — are documented in the BOK study.
At the same time, workers in their 50s added a net 230,000 jobs over the same four years — with 173,000 gains in AI-exposed industries according to the same BOK data. The divergence paints a picture of a labor market being reorganized, not merely disrupted, by artificial intelligence: older, experienced workers are holding ground and even advancing in AI-adjacent roles, while younger entrants are being squeezed out before they can set foot on the first rung.
Sector-by-Sector Toll Shows No Marginal Decline
The BOK report, which draws on National Pension Service subscriber records and the economically active population survey, offers granular breakdowns by industry that make clear just how unevenly generative AI’s employment effects have landed.
Youth employment in IT services fell 31.4 percent over the four-year study window. Publishing — a category that includes software and web design professionals — declined 27.4 percent. Computer programming dropped 16.6 percent, and professional services fell 11.6 percent.
These are not marginal statistical fluctuations. In industries that once served as a reliable on-ramp for South Korean university graduates entering tech and creative-sector careers, the entry-level floor has effectively collapsed. The BOK set 2022 as the study’s baseline explicitly because that was the year OpenAI released ChatGPT, treating the chatbot’s public launch as the demarcation point for the generative AI era.
Higher Degrees No Longer Protect Against AI Displacement
Perhaps counterintuitively, the employment data reveals that higher educational attainment is no longer a reliable shield against AI-driven displacement — at least for the young.
From 2019 to 2022, the average unemployment rate among young people with undergraduate or graduate degrees was 8.2 percent, compared with 8 percent for those who graduated from secondary school or junior college. After 2022, those rates shifted to 7 percent and 5.4 percent respectively — gap widened to 1.6 points, according to the BOK report.
The implication is uncomfortable: degree-holders entering AI-exposed white-collar fields are faring worse in the post-ChatGPT labor market than their peers with vocational or secondary credentials. That gap had been relatively stable before 2022; its widening since then tracks closely with AI adoption in knowledge-work sectors — precisely the industries where four-year degrees were supposed to open the door.
What AI Is Actually Doing: Replacing Codified Knowledge, Sparing Tacit Judgment
BOK researcher Oh Sam-il, writing in the report’s research department, offered an interpretation that resists attributing all the damage to AI alone.
“AI can increase the productivity of young people to a large extent. This also means they can be replaced by AI,” Oh Sam-il wrote. “However, we cannot say that AI is entirely behind the reduction in youth employment. Instead, AI is accelerating the trend of the diminishing career ladder for young people.”
The BOK also noted that the employment outcome depended not on AI adoption itself but on whether businesses choose to replace or assist human work.
This distinction — automation versus augmentation — has emerged as the key mechanistic insight in the economics of AI and labor. The Stanford Digital Economy Lab’s August 12, 2026 revision of its “Canaries in the Coal Mine” study, which draws on ADP payroll data covering 4.6 million workers across more than 730 occupations, provides the clearest articulation of why the pattern falls so sharply on the young.
The Stanford team found that young workers in codified-knowledge jobs — roles drawing on formal, standardized, documented knowledge teachable through education and written procedures — saw employment fall, while employment rose among experienced workers in roles relying on tacit knowledge acquired through practice, mentorship, and repeated exposure to real situations.
Generative AI is fundamentally trained on text and digitally encoded information — which means it can replicate codified knowledge with high fidelity. The gap it cannot close is tacit: the judgment that comes from a decade of navigating ambiguity in real situations, from mentorship that exists outside any document, from the accumulated pattern recognition that no training corpus can capture. That is what experienced workers possess and young workers have not yet had the opportunity to develop.
This explains why workers in their 50s are gaining in the same sectors where young people are losing: AI can replicate the outputs of a first-year analyst far more easily than it can replicate the judgment of a decade-long veteran. The entry-level “career ladder” — the traditional bargain in which new graduates trade rote work for mentorship and upward mobility — is eroding precisely because the rote work is now cheaper to automate than to hire.
Generational Inversion, Confirmed at Scale
The Stanford team’s August 2026 update found that the employment gap for young workers in AI-exposed occupations has now reached 19 percent below where it would be if those workers had kept pace with similarly aged peers in less-exposed occupations — up from 15 percent at the July 2025 data vintage. In absolute terms, employment of workers ages 22 to 25 in the two most AI-exposed quintiles fell about 11 percent between November 2022 and June 2026; employment of the same age group in less-exposed quintiles grew about 10 percent.
Crucially, the Stanford data shows reduced hiring, not layoffs or wage cuts, as the primary adjustment mechanism. Companies are pausing entry-level hiring under uncertainty and AI cost-saving calculations — not eliminating experienced staff at the same rate. The crisis is invisible in aggregate unemployment statistics because the workers who were never hired are not counted in those figures.
Is AI the Only Driver? South Korea’s Demographics Complicate the Picture
The BOK is careful to note that South Korea’s severe demographic decline complicates a clean causal reading of the data. The country’s population decline also contributed to changes in youth employment over the period, while the advent of AI may have accelerated the trend.
South Korea’s total population has been falling for years, and its youth cohort has shrunk alongside it. Fewer young people in the workforce means fewer young workers available to fill positions — which could explain some of the headline employment drop independently of AI.
But the sectoral concentration of losses undercuts a purely demographic explanation. Youth employment tumbled most severely in IT services, publishing, computer programming, and professional services — sectors with highest AI exposure per the BOK report. If demographics alone were responsible, job losses would be more evenly distributed across the economy, not clustered overwhelmingly in sectors where AI tools have been most aggressively deployed.
South Korean youth unemployment fell for a 45th consecutive month as of July 2026, when the youth employment rate stood at 44.2 percent — down 1.6 percentage points from a year earlier. The youth unemployment rate was 6.8 percent in July, marking the largest increase in five and a half years.
What Governments and Employers Are Doing
South Korea’s response to the BOK findings is arriving in layers. The country’s Ministry of Employment and Labor announced plans for a “National Responsibility System for Youth First Careers,” targeting AI-affected entry-level workers, with a goal of creating more than 200,000 youth jobs in advanced and AI-adjacent fields by 2030. On August 14 — four days before the BOK report’s publication — two South Korean lawmakers filed a three-bill package that would impose a mandatory levy on companies whose AI use leads to significant workforce reductions, the first legislation anywhere to tie an automation tax to employer headcount outcomes rather than AI developer revenue.
The government’s approach recognizes what the BOK data makes clear: the policy challenge is not simply about counting lost jobs but about preserving the entry-level training grounds that produce tomorrow’s experienced workforce.
How Long Can the Experience Premium Hold?
The report arrives as South Korea grapples with multiple overlapping pressures: a shrinking workforce, an aging population, historically low birth rates, and now evidence of technology-accelerated entry-level displacement.
The deeper structural risk embedded in the BOK data goes beyond any single generation’s job losses. Entry-level positions exist not merely to produce output — they exist to form skills. The junior analyst who spends two years writing reports that now get drafted by AI, the programmer who debugs code that now gets generated automatically, the customer service representative who resolves edge-case complaints that now get filtered by chatbots: these workers were not only doing jobs. They were developing the tacit knowledge and judgment that would eventually make them the experienced workers commanding the experience premium that currently protects workers in their 50s.
If generative AI continues to automate the entry-level tasks that once served as training grounds for tomorrow’s senior workers, the experience premium now benefiting older workers today may eventually thin the pipeline that creates them. Some analyses have estimated that if early-career learning opportunities continue to contract, the resulting gap in future senior-worker supply could reduce productivity growth by 0.05 to 0.35 percentage points as the pipeline of future experienced workers thins. The career ladder does not simply disappear for one generation — it reshapes the workforce for every generation that follows.
Frequently Asked QuestionsWhat did the Bank of Korea’s report actually find?
The Bank of Korea published a study on August 18, 2026 drawing on National Pension Service administrative data for the four-year period from June 2022 to June 2026. It found that 285,000 net jobs held by workers aged 15 to 29 were lost over that period, of which 268,000 — or 94 percent — were concentrated in sectors with high AI exposure: IT services, publishing, computer programming, and professional services. Over the same period, workers in their 50s gained a net 230,000 jobs, with 173,000 of those gains coming from those same AI-exposed industries. The report does not claim AI alone caused these losses, but identifies AI as a major accelerant of a trend the BOK describes as “the diminishing career ladder.”
Why are degree-holders faring worse than workers without degrees?
The BOK data shows that after ChatGPT’s emergence, the gap between unemployment rates for college and graduate degree-holders and secondary school graduates among young workers widened, rather than narrowed. The most plausible explanation is that degree-holders disproportionately enter the knowledge-work sectors — IT services, professional services, publishing, programming — that generative AI is affecting most directly. Workers with vocational or secondary credentials are more likely to be in sectors where AI automation is less advanced, making the degree a less reliable shield than it once was in exactly the fields it was designed to open.
What is the difference between AI automation and AI augmentation, and why does it matter for employment?
When companies deploy AI to automate tasks — writing first drafts, processing routine queries, generating code templates — those tasks no longer require a junior human to perform them, and entry-level hiring falls. When companies deploy AI to augment experienced workers — supporting complex judgment, surfacing relevant data, flagging potential errors — employment holds stable or rises. Stanford Digital Economy Lab research published August 12, 2026 found that the employment decline is concentrated specifically in occupations where AI is used more to substitute for human tasks; where AI is used more to complement workers, employment is flat or rising, especially among more experienced staff. The distinction has major implications for career development: skills built on codifiable, documented procedures are far more replicable by AI than the judgment built through years of practice and mentorship.
Will the career ladder eventually recover as AI matures?
The honest answer is that it is too early to know. The Bank of Korea and Stanford researchers both note significant uncertainty about whether the pattern will accelerate, stabilize, or reverse. One view holds that AI will eventually expand labor demand as it enables new industries and productivity gains, with the primary benefits accruing to younger workers who grow up as “AI natives.” The countervailing risk is that if entry-level on-the-job learning continues to contract, the supply pipeline for future experienced workers thins — meaning the experience premium that currently protects older workers could erode over time as fewer workers cycle through the formative early-career stages that build it. South Korea’s government appears to be betting on training interventions to bridge the gap, but the structural dynamics documented in the BOK study suggest the challenge runs deeper than any single retraining program can address.