top_banner.png
Why Are Young Workers Losing Jobs in the Early Stage of AI Diffusion?
Since the launch of ChatGPT in November 2022, a wave of generative AI services has rapidly emerged, each pushing the boundaries of what AI can do in ever-shorter innovation cycles. Performance has advanced not only in back-office tasks such as coding, document writing, and data analysis, but also in complex reasoning and multimodal processing. Yet behind the convenience of these technological advances, a new generational divide is taking shape: employment among junior workers(20s) is falling, while jobs for senior workers(50s) continue to grow.
The Labor Market Research Team at the Bank of Korea analyzed changes in the labor market following the spread of AI, using administrative data from the National Pension Service (NPS) on the number of subscribers by age group and industry.
Youth Employment Fell More in highly AI-Exposed Industries
Over the past three years(July 2022 to July 2025), the number of jobs held by young workers (aged 15–29) decreased by 211,000, of which 98.6% occurred in industries with high AI exposure (the top two quartiles of exposure). In contrast, employment among workers in their 50s increased by 209,000 over the same period, with 69.9% of the growth concentrated in highly AI-exposed sectors[Figure 1]. In short, AI diffusion has created opportunities for senior workers but posed challenges for the young.
This seniority-biased employment effect of AI diffusion becomes even clearer when looking across industries. For young workers, employment has declined sharply in the top quartile of AI-exposed industries, widening the gap with the least-exposed sectors. In contrast, employment among those in their 50s increased in the top quartile of AI-exposed industries[Figure 2].
Youth(aged 15–29)
50s
Even within the most AI-exposed 4th quartile, youth employment declined more sharply in fields where entry-level tasks are easily automated. Since November 2022, youth employment has declined by 11.2% in computer programming, systems integration, and management; 8.8% in professional services such as law, accounting, taxation, advertising, and consulting; 20.4% in publishing, and 23.8% in informaton services[Figure 3]. Yet even in these sectors, employment among core-age workers(including those in their 50s) continued to grow.
Computer programming, systems integration, and management
Professional services
Why Are Young Workers More Easily Replaced by AI?
Junior employees typically perform codified and book-learning-based tasks—precisely the types of work AI can execute quickly. In contrast, senior workers tend to possess strengths in tacit knowledge and social skills—such as organizational management, contextual understanding, decision-making, and interpersonal coordination—that AI cannot yet replicate. Indeed, the reduction in working hours due to AI use was greatest among employees with less than five years of experience, indicating that entry-level tasks are the most automatable[Figure 4].
How Should We Respond?
It remains uncertain whether the contraction in youth employment observed in the early phase of AI diffusion will persist. Over the long run, firms are likely to pursue more sustainable strategies—focusing not on downsizing but on training AI-collaborative talent, building systems for AI-human cooperation, and redesigning job structures—to prevent the erosion of their talent pipelines. Also, even the productivity gains from AI may eventually expand labor demand, offering young workers new opportunities as well.
Young people, in turn, can counter limited career experience with new kinds of experience. While 50s initially benefited from the depth of their past experience, AI-savvy youth can explore new industrial opportunities emerging alongside AI, using the technology as a complement to actively create their own opportunities. To this end, sustained social discussions are needed on policy directions that support AI education and training, greater access to public data, and the creation of an inclusive startup ecosystem that tolerates risk and failure.
- [1] Based on monthly administrative data on National Pension Service (NPS) subscribers by age group and industry, provided under an MOU between the Bank of Korea and the NPS. The dataset covers about 16 million subscribers, encompassing most regular workers in Korea.
- [2] For details, see BOK Issue Note No. 2025-30, “AI Diffusion and Youth Employment”.
- [3] Industry-level AI exposure was calculated by taking the employment-weighted average of occupation-specific AI exposure indices from Felten et al. (2021) using occupational distributions from the Regional Employment Survey. Industry-level AI complementarity indices were constructed in the same way, using Pizzinelli et al. (2023). For detailed distributions of AI exposure and complementarity by industry, see Appendices 1 and 4 of the Issue Note.
- [4] The seniority-biased employment effect remains significant even when using an alternative measure, AI utilization rates derived from household survey data in BOK Issue Note No. 2025-22(Suh et al, 2025) and controlling for industry and time fixed effects. For details, see Appendices 2 and 3 of BOK Issue Note No. 2025-30. Meanwhile, youth employment declined less in industries with higher AI complementarity, and the wage effects of AI diffusion have so far remained unclear due to wage rigidity.
- [5] Among junior workers (less than five years of experience), the reduction in working hours due to AI use was largest for those with bachelor’s or master’s degrees, while the effect was smaller for PhD holders, college graduates, and those with a high-school education or below. This suggests that even within the youth group, individuals with mid-to-upper-level educational attainment may be more easily displaced by AI, forming a U-shaped pattern of substitution risk.