The simple “AI in, people out” story is not what the Korean data shows yet
The most familiar AI workforce story is straightforward: automate work, need fewer people, shrink the organization. A Korean firm-level panel covering 2017 through 2024 points to a more complicated pattern.
The Korea Research Institute for Vocational Education and Training estimated a 2.9% increase in employment for firms newly using AI in a given year. Firms that had already been using AI since the previous year showed a 3.8% point estimate, although the statistical significance of the difference between the two estimates was not separately established.
What companies used AI for mattered
When AI was used for product and service development, employment was estimated to be about 4.0% higher. For production-process use, the 1.5% estimate was not statistically significant. That does not mean production AI reduced jobs; it means this analysis did not identify a statistically clear increase or decrease.
In 2024, 59.3% of AI-using firms in the dataset said product and service development was their primary use. That mix matters when interpreting the positive employment result.
Creating new AI products can create work beyond AI engineering
A firm building AI-enabled products may need model specialists, but it can also need product managers, data operations, security, sales, customer support, governance and people who validate AI output.
One technology can reduce some tasks while creating demand for other tasks around the new capability. For workforce strategy, “Does this company use AI?” is therefore less informative than “What is this company trying to do with AI?”
AI adoption did not produce an immediate productivity jump either
A separate KRIVET analysis found no statistically significant same-year increase in sales or sales per employee when comparing a firm’s AI-use years with its non-use years. Clear labor-productivity improvement was not established.
However, event-time estimates showed a pattern of roughly 3–4% positive sales estimates from the second year after initial AI adoption. Researchers explicitly cautioned that this should not be read as a definitive causal effect, but the lag is consistent with the possibility that converting AI into business performance takes time.
AI also appears alongside a broader digital foundation
Another 2024 establishment analysis found AI use at 82.2% among workplaces using cloud, data analytics and IoT together, compared with 4.5% among those using none of the three. This is a contemporaneous association, not proof that the digital foundation caused AI adoption.
Operationally, though, the connection is intuitive: AI needs usable data, connected systems, people who can work with the tools and processes that can absorb the output. Buying an AI product and being organizationally ready for AI are different things.
The workforce question may be less about “How many jobs disappear?” and more about “What work gets created?”
As AI improves, some repetitive and standardized work will almost certainly face stronger automation pressure. But firms using AI to create new products and services may simultaneously require new roles and capabilities.
Before asking whether AI removes people, it may be more useful to ask whether a company is using AI mainly to compress existing work or to create work that did not exist before. Headcount can be an outcome of that choice rather than the starting point.
BANSEOG VIEW
AI may not determine employment direction by itself; the use case changes the role structure
The Korean panel evidence argues against treating AI adoption and workforce reduction as synonyms. Product/service development showed a positive employment estimate, while production-process use showed no statistically significant employment change.
For talent strategy, the useful unit of analysis is Task → new or reduced Role → required Capability, not simply whether a company has adopted AI.
SOURCES
Primary sources and references
- 한국직업능력연구원 — “AI(인공지능) 활용 시, 기업 고용 2.9% 증가한 것으로 추정”
2026-08-06. 2017~2024 기업활동조사 패널 분석. AI 활용과 고용 변화, 활용 목적별 추정치.
- 한국직업능력연구원 — “AI 활용의 기업 성과, 매출은 2년 차부터 증가 양상...노동생산성 향상은 아직 불확실”
2026-08-11. 도입 2년 차 이후 매출 3~4% 수준의 양(+) 추정치와 노동생산성 결과.
- 한국직업능력연구원 — “AI는 디지털 토대 위에 선다…준비도 수준별 이용률 최대 18배 차이”
2026-08-10. 클라우드·데이터 분석·IoT 활용과 AI 이용률의 동시 관찰 관계.
Based on public KRIVET releases available as of Aug. 30, 2026. The +2.9%, +4.0% and 3–4% figures are estimates from panel analyses, not universal causal effects for every firm. No statistically significant change for production-process AI does not mean job losses were established. The 82.2% versus 4.5% comparison is an association observed at the same point in time and should not be treated as proof of causality.