Artificial Intelligence and the Labour
Market in Korea
Artifi
cial Intelligence and the Labour M
arket in Korea
Artificial Intelligence
and the Labour Market
in Korea
This work is published under the responsibility of the Secretary-General of the OECD and the President of the Korea
Labor Institute. The opinions expressed and arguments employed herein do not necessarily reflect the official views of
the Member countries of the OECD or of the Korea Labor Institute.
The names and representation of countries and territories used in this joint publication follow the practice of the OECD.
This document, as well as any data and map included herein, are without prejudice to the status of or sovereignty over
any territory, to the delimitation of international frontiers and boundaries and to the name of any territory, city or area.
Please cite this publication as:
OECD/Korea Labor Institute (2025), Artificial Intelligence and the Labour Market in Korea, OECD Publishing, Paris,
ISBN 978-92-64-85645-5 (print)
ISBN 978-92-64-41590-4 (PDF)
ISBN 978-92-64-93169-5 (HTML)
Photo credits: Cover © Gorodenkoff/.
Corrigenda to OECD publications may be found at:
© OECD/Korea Labor Institute 2025
Attribution International (CC BY )
This work is made available under the Creative Commons Attribution International licence. By using this work, you accept to be bound by the terms of this licence
(
Attribution – you must cite the work.
Translations – you must cite the original work, identify changes to the original and add the following text: In the event of any discrepancy between the original work and the
translation, only the text of original work should be considered valid.
Adaptations – you must cite the original work and add the following text: This is an adaptation of an original work by the OECD and the Korea Labor Institute. The opinions expressed
and arguments employed in this adaptation should not be reported as representing the official views of the OECD or of its Member countries or of the Korea Labor Institute.
Third-party material – the licence does not apply to third-party material in the work. If using such material, you are responsible for obtaining permission from the third party and for
any claims of infringement.
You must not use the OECD’s or the Korea Labor Institute’s respective logo, visual identity or cover image without express permission or suggest the OECD or the Korea Labor
Institute endorse your use of the work.
Any dispute arising under this licence shall be settled by arbitration in accordance with the Permanent Court of Arbitration (PCA) Arbitration Rules 2012. The seat of arbitration shall
be Paris (France). The number of arbitrators shall be one.
3
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
Foreword
As the general purpose technology of our time, Artificial Intelligence (AI) is expected to profoundly change
all aspects of our life, including work. The technology is rapidly evolving and is increasingly making its way
into the workplace, bringing promises of increased productivity and improvements in job quality, amongst
others. The question is not so much whether AI should be used at work, but rather how, so that its benefits
can be maximised, while managing some of the risks such as: job automation, invasions of privacy, bias
and discrimination, and increased work pressure and stress, to name just a few. The evidence suggests
that policies and institutions matter to making a success of AI, including: training and social dialogue, but
also clear and proportionate regulation.
In this series of country reviews, the OECD analyses the impact AI is having on a country’s labour market
from an internationally comparative perspective, and also takes stock of that country’s policies and
institutions, against the backdrop of the OECD AI Principles for trustworthy AI. These country reviews aim
to help policymakers better understand the risks and opportunities, and offer them a menu of options to
help workers and employers make a success of AI, drawing on examples and best practice from across
the OECD. In addition, by providing an in-depth analysis of a particular country, these reviews allow
policymakers from across the OECD to draw lessons from the experience of a specific country to inform
their own policies and institutions.
4
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
Acknowledgements
This report was prepared jointly by the OECD (Stijn Broecke and Carla Ruggiu) and the Korea Labor
Institute (Hyeongjun Bang and Seri No), with contributions from Yongjin Nho (Seoul National University of
Science and Technology) and Hwanoong Lee (Konkuk University). Valuable comments were provided by
Glenda Quintini, Head of the Skills and Future Readiness Division at the Employment, Labour and Social
Affairs Directorate of the OECD. The OECD acknowledges the support of the Ministry of Foreign Affairs of
the Republic of Korea and KLI the support of the National Research Council for Economics, Humanities
and Social Sciences. The opinions expressed and arguments employed in this report do not necessarily
reflect the official views of these two organisations nor those of the Member countries of the OECD or of
the Korea Labor Institute.
5
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
Table of contents
Foreword 3
Acknowledgements 4
Abbreviations and acronyms 7
Executive summary 9
1 Overview 11
Korea’s population is ageing, threatening economic growth 11
Artificial Intelligence could be part of the solution, but adoption in Korea is still low 12
To encourage adoption, AI should be safe and trustworthy 13
Social dialogue can facilitate the AI transition 14
Investing in skills will be critical to make a success of AI 16
Special support should be provided to SMEs 17
References 18
2 The impact of AI on the labour market 20
In Brief 21
The impact of AI on job quantity and skills: Evidence from OECD countries 23
The impact of AI on job quantity and skills: Evidence from Korea 30
The impact of AI on job quality: Evidence from OECD countries 46
The Impact of AI on job quality: Evidence from Korea 49
The impact of AI on inclusiveness: Evidence from OECD countries 53
The impact of AI on inclusiveness: Evidence from Korea 59
References 62
Notes 69
3 Seizing the opportunities and managing the risks: The policy response to AI 70
In Brief 71
Review of international developments on AI policy and regulation 73
Recent regulatory and policy developments in Korea 90
References 102
Notes 107
6
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
FIGURES
Figure . Korea is projected to have the largest decline in working age population in the OECD by 2060 11
Figure . AI adoption in SMEs is lower in Korea than in other countries 13
Figure . The majority of firms in Korea do not have guidelines for the use of AI 14
Figure . Few workers in Korea say their employer consults unions or councils on the adoption of AI 15
Figure . SMEs in Korea are positive about the impact of generative AI 18
Figure . White collar occupations are more exposed to AI than occupations requiring manual skills and
strength 24
Figure . Currently, most Korean firms only invest a small share of sales revenue in AI, and even planned
investment in AI by Korean firms is relatively low 30
Figure . Nearly 1 in 3 Korean workers uses AI 1 to 2 times a day 31
Figure . Full automation of tasks by AI is rare in Korea 35
Figure . Few Korean SMEs report an impact of generative AI on staffing needs 37
Figure . The vast majority of firms in Korea say that AI only replaces up to 10% of tasks 38
Figure . Firms in Korea report increases in the kinds and levels of skills required following AI adoption 40
Figure . Generative AI increases the demand for skills in SMEs, including in Korea 41
Figure . AI increases the frequency of communication within firms in Korea 42
Figure . Nearly one in four firms in Korea say AI has helped them address labour shortages 43
Figure . Generative AI has helped some Korean SMEs compensate for worker shortage or lack of skills 43
Figure . Proficiency in problem solving in technology-rich environments among adults 45
Figure . Proficiency in problem solving in technology-rich environments among adults, by educational
attainment and age 45
Figure . Less than half of firms in Korea say they have provided training for workers to work with AI 46
Figure . According to firms and employees, AI improves productivity, performance, and job satisfaction 51
Figure . Most workers disagree that AI reduced physical labour intensity and mental stress 52
Figure . Adoption of generative AI increases with firm size 55
Figure . The most reported barrier among Korean SMEs is a lack of skills among employees 56
Figure . Nearly one in two workers say their employers provide training to work with AI 76
Figure . Employers cite a lack of skills as a major barrier to adopting AI 77
Figure . Workers who say their employer consults them about the adoption of new technologies are more
positive about the impact of AI on their jobs 79
Figure . Two in five workers in Korea say their employers are not being transparent about the use of AI in
the workplace 93
Figure . AI-driven matching in Korea’s Work24 95
TABLES
Table . The impact of AI on full-time, permanent employment in Korea: Regression results 34
Table . The impact of AI on the wages of full-time, permanent employees in Korea: Regression results 50
Table . The impact of AI on full-time, permanent employment growth in Korea: Regression results by
gender, age and skill level 60
Table . The impact of AI on full-time, permanent wage growth in Korea: Regression results by gender, age
and skill level 61
7
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
Abbreviations and acronyms
ADS Automated Decision System
AEDT Automated Employment Decision Tool
AI Artificial Intelligence
AIDA Artificial Intelligence and Data Act (Canada)
AIOE AI Occupational Exposure
EAPS Economically Active Population Survey
EU European Union
FTA Free Trade Agreement
GDPR General Data Protection Regulation
GPT General Purpose Technology
HR Human Resources
ICT Information and Communications Technologies
KLI Korea Labor Institute
KSIC Korean Standard Industrial Classification
LLM Large Language Model
NTC National Training Card
OECD Organisation for Economic Co-operation and Development
PES Public Employment Service
PIAAC Programme for International Assessment of Adult Competencies
PIPA Personal Information Protection Act
8
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
R&D Research and Development
SME Small and Medium-sized Enterprise
STEM Science, Technology, Engineering and Mathematics
TFP Total Factor Productivity
UK United Kingdom
US United States
9
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
Executive summary
The Korean fertility rate has fallen to the lowest in the world and the population is set to halve over the next
six decades. The old-age dependency ratio is projected to surge, putting considerable strain on the labour
supply which, in turn, could threaten long-term productivity and economic growth.
While AI will not solve this challenge by itself, it may help boost productivity and address skills shortages
resulting from an ageing workforce. By automating tasks that humans do, AI could help address skills
shortages. Labour productivity could be improved if workers are augmented by these new technologies,
allowing companies to do more with less. However, adoption of AI in Korea remains low by international
standards: only 31% of SMEs in Korea are using AI, compared to over 50% in Germany – yet the share of
employment in SMEs in Korea is particularly high (over 80%).
A lack of skills is cited by Korean SMEs as the most important barrier to AI adoption. Indeed, 30% of adults
in Korea have no or limited experience with computers or lack confidence in their ability to use them. In
addition, the use of AI in the workplace is increasing demand for high-level skills (including data analysis
and interpretation) and social skills.
Korea already has a range of programmes in place to promote AI-related education and training, however
the brain drain of AI talent continues to be a major challenge. To successfully address skills issues related
to AI, it will be important that Korea: promotes more on-the-job learning; ensures that training programmes
are tailored to the needs of SMEs; and co-ordinates effectively between ministries. To achieve the latter,
Korea may want to consider setting up a specialised, overarching AI agency to align AI education and
training policies with industrial policies.
In order to promote adoption and use, AI will also need to be safe and trustworthy. Korea is only the second
country in the world, after the EU, to adopt comprehensive AI legislation, which aims to simultaneously
boost innovation and trust in AI.
However, specific guidance on the use of AI in the workplace is currently lacking in Korea and there may
be a need to develop more specific guidance for employers on issues such as: data protection and privacy
in the workplace; bias and discrimination; automated decision making and algorithmic management
practices; transparency; explainability; and accountability. In some cases, regulatory changes may need
to be considered too.
While automation through AI could help address skills and labour shortage, it could also result in job losses
for some workers. Across the OECD, there is little evidence so far of a negative impact of AI on aggregate
employment. However, this report presents some new analysis, first-of-its-kind in that it distinguishes
between types of AI, and which shows that in Korea some forms of more “traditional” AI appear to be
associated with lower employment growth in full-time, permanent jobs for youth, low- and medium-
educated workers, as well as in the manufacturing sector. This suggests that the benefits and risks of AI
may not be equally distributed, and Korean policymakers will need to make sure that no groups are left
behind.
10
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
Social dialogue, social protection and re-employment services can play an important role in supporting
workers through the transition. In particular, access to social protection for workers in non-standard forms
of work in Korea could be strengthened. With regards to social dialogue, worker consultation appears less
common in Korea than in other OECD countries. While the Korean Labour Standard Act stipulates that
employers need to consult workers when they intend to alter the rules of employment in a way that is
unfavourable to employees (. leads to a deterioration in working conditions), it is not clear in practice
whether this law would apply in the case of AI adoption in the workplace. Finally, Korea already uses AI in
the provision of employment services and counselling to job seekers (Work24), thereby improving labour
market matching, but services could further tailored to the specific needs of SMEs and job seekers.
11
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
Korea’s population is ageing, threatening economic growth
Fertility has been declining in OECD countries, including in Korea which now has the lowest fertility rate in
the world (OECD, 2024[1]). At the same time, people are living longer: life expectancy at birth exceeds
80 years in more than two-thirds of OECD countries, with Korea once again towards the top of the ranking
( years).
These opposing trends in fertility and longevity mean people are getting older. Across the OECD, the old-
age dependency ratio, defined as the ratio of seniors to the working-age population, has increased from
19% in 1980 to 31% in 2023, and it is projected to increase further to 52% by 2060 (above 75% in Korea).
At the same time, the working-age population in the OECD is projected to decline by 8% between 2023
and 2060–and by up to 46% in Korea (Figure ) ( (OECD, 2025[2])).
Figure . Korea is projected to have the largest decline in working age population in the OECD by
2060
Projected percentage change in the working age population (aged 20-64 years), 2023-2060
Note: The medium scenario of the population projections is used. OECD: Weighted average of OECD countries.
Source: Secretariat’s calculations based on United Nations (2024), World Population Prospects 2024,
-60%
-50%
-40%
-30%
-20%
-10%
0%
10%
20%
30%
40%
1 Overview
12
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
Without further policy action or changes in behaviour these trends will weigh significantly on economic
growth and the capacity of OECD countries to maintain their living standards. In most countries, a
combination of boosting migration, closing the gender gap in employment, and raising the employment
rates of old-age people, would be sufficient to avoid a fall in the employment-to-population ratio (OECD,
2025[2]). Additional interventions will be needed if Korea wants to maintain living standards.
Artificial Intelligence could be part of the solution, but adoption in Korea is still
low
Artificial Intelligence (AI) (see Box ) could help reverse at least part of the drop in GDP growth resulting
from population ageing. AI is a new General Purpose Technology (GPT), comparable to earlier digital
technologies such as the internet and personal computers, or previous breakthrough innovations like the
steam engine and electricity. These past inventions have led to periods of accelerated economic growth.
Similarly, it has been estimated that AI could result in growth in total-factor productivity of between
percentage points (.) per year ( . for labour productivity) (Filippucci, Gal and Schief,
2024[3]). In addition, through its impact on automation, AI could help address skills and labour shortages
(OECD, 2025[2]), and it could help extend working lives by improving the quality of jobs (OECD, 2023[4]).
Box . What is Artificial Intelligence?
The OECD defines an AI system as “a machine-based system that, for explicit or implicit objectives,
infers, from the input it receives, how to generate outputs such as predictions, content,
recommendations, or decisions that can influence physical or virtual environments. Different AI systems
vary in their levels of autonomy and adaptiveness after deployment” (OECD, 2024[5]).
AI can be seen as a General-Purpose Technology (GPT) (Brynjolfsson, Rock and Syverson, 2017[6]).
GPTs are characterised by their pervasiveness, inherent potential for technical improvements and
innovational complementarities (Bresnahan and Trajtenberg, 1992[7]). AI has the potential to be
pervasive, impacting a broad variety of sectors and occupations. Not only does it improve over time
through the expertise of inventors or developers, but also by learning on its own from data and its past
predictions. Furthermore, it has the capability to spawn complementary innovations.
The applications of AI are wide-ranging. Wherever large amounts of data are available, AI has the
potential to improve decision making and reduce costs. AI is used in workplaces for various tasks,
including coding, drafting emails, improving written text, generating summaries, and translating content.
In human resources, AI streamlines recruitment by screening resumes and matching candidates. In
marketing and advertising, AI enables personalised content creation and optimises ad targeting. In
finance, AI enhances fraud detection and improves financial forecasting. In manufacturing, AI-powered
robotics assist production, while predictive maintenance helps prevent equipment failures.
However, adoption of AI in Korea remains low, with estimates ranging from % to % (Han, 2023[8];
NIA, 2025[9]; KOSIS, 2025[10]). Evidence from a new OECD survey shows that adoption of AI by SMEs
(<250 employees) in Korea (31%) is higher than in Japan (27%), but considerably lower than in some other
OECD countries, such as Austria (42%), Ireland (45%) and Germany (51%) (Figure ). Yet the share of
employment in SMEs in Korea is particularly high (over 80%).
13
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
Figure . AI adoption in SMEs is lower in Korea than in other countries
Percentage of SMEs reporting usage of generative AI or other forms of AI
Source: OECD (2025[11]), Microdata from the OECD SME Survey on Generative AI.
To encourage adoption, AI should be safe and trustworthy
While AI has the potential to bring many benefits, including to the workplace, employers will be hesitant to
adopt AI, and workers reluctant to use it, if the risks are not adequately addressed, including risks to:
worker safety and health, data protection and privacy, as well as bias and discrimination.
Korea is only the second country in the world, after the EU, to adopt comprehensive AI legislation. The
Basic Act on the Development of Artificial Intelligence and the Establishment of Trust (“AI Basic Act”
henceforth) was approved in January 2025 and will be enforced from January 2026 onwards. The AI Basic
Act refers to the need to “minimise risks and build trust”. The AI Basic Act covers important principles such
as transparency and explainability, as well as safety and reliability of AI. Firms using “high-impact AI” will
be expected to provide users with advance notice. They will also need to: develop and implement a risk
management plan, a user protection plan, ensure human supervision and oversight, and maintain
documentation on safety and reliability measures – although details on how this should be done are still
left open.
Korea also has comprehensive data protection legislation. The Personal Information Protection Act (PIPA)
offers similar rights and protections as the GDPR in the EU, including in the case of fully automated
decisions where the PIPA, unlike the GDPR, specifically mentions the use of AI. When personal data is
collected and processed, the consent of individuals is required. While the meaningfulness of consent in
the context of an employer-employee relationship has been questioned, given the power imbalance
between the two parties, in Korea, an interesting solution to this challenge appears to have been found: if
there is a trade union (labour union) in the workplace, consent is deemed to be given when the union
agrees, while in workplaces without unions, the condition is met when more than half of all workers express
their agreement.
0%
10%
20%
30%
40%
50%
60%
Germany Ireland Austria Canada UK Korea Japan Total
14
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
Despite these legislative efforts, Korea’s approach to regulating AI is based more on principles than on
detailed prescriptions and regulations. This may encourage innovation, but it also makes enforcement
more difficult. The AI Basic Act is deliberately less restrictive than the EU AI Act and aims to combine
ethical AI and safe practices with domestic innovation. While the EU AI Act is closer to a product safety
regulation, the AI Basic Act contains strong elements of an industrial policy. Other principles, like the
guarantee of human rights, and bias and discrimination, are not included in the AI Basic Act but are covered
by the Korea Guidelines for Ethical Standards of Artificial Intelligence instead.
Going forward, it will be important to develop specific guidance on the use of AI in the workplace. Indeed,
while some companies in Korea (%) have developed their own guidelines for the use of AI, the majority
have not (Figure ). The AI Basic Act defines “high-impact AI”, similar to “high-risk AI” in the EU AI Act.
However, the use of AI in employment (a high-risk case in the EU AI Act) is only alluded to briefly in the AI
Basic Act when it mentions AI that involves “Judgment or evaluation that has a significant impact on an
individual's rights and obligations, such as hiring.” It is not clear, therefore, to what extent some of the AI
Basic Act would apply to situations in the workplace other than hiring, nor what steps employers should
take to ensure the safe and trustworthy use of AI.
Figure . The majority of firms in Korea do not have guidelines for the use of AI
Percentage of firms reporting the existence of guidelines, by sector and firm size
Note: The survey targeted firms that use AI, focussing on industries classified under the Korean Standard Industrial Classification (KSIC),
specifically: Manufacturing, Information and communication, Professional scientific and technical service, Healthcare. Only firms that utilise AI
and have 10 or more employees were included in the survey. HR managers and AI developers provided the survey responses. The survey was
conducted over a two-month period, from 20 October 2024 to 31 December 2024. The survey covered a population of 9 625 establishments,
including 3 292 in manufacturing, 3 118 in information and communication, 1 788 in professional and scientific services, and 790 in healthcare.
The sample was drawn using a random sampling method, with a target sample size of 200. Ultimately, the study achieved valid responses from
145 firms, whose data were incorporated into the final analysis.
Source: Survey on AI Utilisation and Labour Market Changes conducted by the Korea Labor Institute (2024).
Social dialogue can facilitate the AI transition
In translating the above laws and principles into practical guidelines in the workplace, social dialogue will
be critical. OECD research has shown that social dialogue and collective bargaining can play a key role in
supporting workers and firms in the AI transition, and in fostering fair and dynamic labour markets (OECD,
2023[4]), and also that the outcomes for workers when AI is adopted in the workplace are better when
employers consult them about the use of AI (Lane, Williams and Broecke, 2023[12]).
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
Information and
Communication
Professional,
Scientific, and
Technical Services
Manufacturing Health and Social
Work
299 or fewer 300 or more
Industry Firm size
15
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
At the national level, social dialogue in Korea is still in its infancy. Discussions on AI and technology issues
take place in the National Artificial Intelligence Commission, which is composed of ministers, heads of
public institutions, academics, and presidents of technology companies – however there is no specific
focus on the labour market. The latter topic is discussed separately in the Economic, Social and Labour
Council, which recently launched the “AI and Labour Research Association”, composed of worker and
employer representatives, officials from the ministries of labour and Industry, as well as professors and
researchers.
At company level, the Labour Standard Act stipulates that employers in Korea need the consent from
workers (from the union if there is one, if not from the majority of workers) when they intend to alter the
rules of employment in a way that is unfavourable to employees (. leads to a deterioration in working
conditions). However, there is uncertainty surrounding how AI adoption impact working conditions and
what this means for the obligation of employers to consult workers.
In practice, worker consultation on AI adoption in Korea appears limited. A survey of workers in Korea
showed that % were not involved in discussions around AI adoption in their workplace and only %
said their firms engaged in discussions with unions or labour management councils (Figure ). As
expected, these discussions are far more common in large firms. In the few cases where workers were
consulted, the most common topics covered were: AI’s impact on the number of jobs, changes in specific
occupations, emerging training needs, and methods of data collection and use. A case from Korea further
illustrates how the involvement of workers in AI adoption and maintenance appears limited (Box ).
Figure . Few workers in Korea say their employer consults unions or councils on the adoption
of AI
Percentage of employees saying employers consult unions or councils on AI adoption, by firm size
Note: “Yes” indicates that a labour union or works council exists and is consulted. “No” indicates that a labour union or works council exists but
is not consulted. “None” indicates that no labour union or works council exists. The survey targeted individual employees who use AI in firms
operating within four industries classified under the Korean Standard Industrial Classification (KSIC): Manufacturing, Information and
Communication, Professional, Scientific and Technical Services, and Healthcare. Employees using AI provided the survey responses. The
survey was conducted over a two-month period, from 20 October 2024 to 31 December 2024. The sample was drawn using a random sampling
method, with a target sample size of 600. Ultimately, the study achieved valid responses from 426 employees, whose data were incorporated
into the final analysis.
Source: Survey on AI Utilisation and Labour Market Changes conducted by the Korea Labor Institute (2024).
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
299 or fewer
300 or more
Total
Yes No None
16
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
Box . Case study: AI adoption processes in an electronic parts manufacturer in Korea
A firm producing multilayer ceramic capacitors (MLCC), power inductors, chip resistors, tantalum,
camera and IT modules, semiconductor package substrates, and others, applies AI to the quality
inspection of finished products in the MLCC production process. As product quality is a key competitive
factor in the MLCC market, strict quality control is essential. To address yield losses caused by the
limitations of visual inspections conducted by human workers, the firm introduced an AI model.
At the AI development stage, production workers contribute minimally to AI development, while AI
developers and manufacturing technology team members play dominant roles. The production workers
were expected to provide domain knowledge to aid in AI development, but instead, some manufacturing
technology team members, who possess high domain expertise and extensive work experience, took
on this role. As a result, the production workers had little involvement in AI development. At the
improvement stage, AI developers again dominate the process of re-teaching AI. As the data used for
AI training drifts over time, due to changes in product designs or work methods, regular re-learning of
AI is necessary. AI for machinery-operating tasks is updated daily, while AI for product quality inspection
is updated weekly. Interestingly, the re-learning process is automated within the AI algorithm itself. As
a result, production workers play a very limited role in AI improvement.
During the AI use stage, AI tends to replace human labour in this firm, leaving little room for workers to
contribute. The adoption of AI led to the downsizing of hundreds of workers in the quality inspection
processes. Most of the affected workers were from subcontractors, while directly employed workers
were reassigned to other roles. The machinery-operating processes also face downsizing, with just two
workers now handling jobs that previously required dozens of employees. Additionally, two-thirds of
clerical workers who perform simple tasks are expected to be made redundant in the future. However,
the firm has been growing rapidly, and AI adoption has not yet led to overall redundancy but rather
slowed the increase in the workforce. Notably, new plants were recently built at each of the two domestic
factories.
Investing in skills will be critical to make a success of AI
A lack of skills can be a significant barrier to AI adoption and use. Indeed, the greatest barrier to adoption
cited by SMEs that have not adopted generative AI in Korea, is a lack of skills among employees (cited by
53%) (OECD, 2025[11]). These findings are consistent with research for other countries, which showed that
40% of employers in the manufacturing and finance sectors said that skills were a barrier to AI adoption
(Lane, Williams and Broecke, 2023[12]). At the same time, workers in these same sectors who say they
have been trained to work with AI are considerably more positive about the impact of AI on their
performance, enjoyment of work, and mental and physical health than workers who did not receive such
training (Lane, Williams and Broecke, 2023[12]).
In addition, Korea faces a challenge in retaining AI talent – reflecting a wider brain drain issue which can
be traced back to the 1950s, with high-skilled workers seeking better job opportunities and higher living
standards abroad. While the Korean Government has introduced AI-focussed education and vocational
training programmes, the shortage of AI experts has been persistent with intensified competition and a
surge in global demand for AI professionals. This threatens technological progress and productivity growth.
It is therefore encouraging to see the AI Basic Act mention the importance of skills. It stipulates that a
“basic” plan will have to be elaborated every three years by the Minister of Science and ICT which must
cover: “Matters related to training professional manpower for systematic fostering of the artificial
17
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
intelligence industry”. The AI Basic Act also states that “The Minister of Science and ICT shall train and
support professional manpower related to artificial intelligence.” Korea already has several publicly funded
education and training programmes to develop AI skills, including K-Digital Training, a vocational training
initiative that aims to provide high-skilled workers in digital and edge-tech industries and which, so far, has
benefited over 5 000 participants in more than 200 courses. However, going forward, the
Korean Government should: promote more on-the-job learning; ensure that training programmes are
tailored to the needs of SMEs; and co-ordinate effectively between ministries. To achieve the latter, Korea
may want to set up a specialised, overarching AI agency to co-ordinate AI education and training (Ministry
of Labour and Employment) and industrial policies (Ministry of Trade, Industry and Energy).
It will be important to ensure that training does not focus only on AI technical skills required for developing
and maintaining AI, but also on more general skills required to work with AI. Indeed, most workers will be
exposed to AI in the workplace, without necessarily needing such technical skills (Green, 2024[13]). In this
respect, the K-Digital Beginner-level Skill Training, which provides beginner-level digital skills to
participants, is a promising programme, although its coverage is currently limited (22 000+ participants)
compared to the number of workers who will be exposed to AI at work. Article 4 of the EU AI Act, which
requires providers and deployers of AI systems to ensure a sufficient level of AI literacy of their staff and
other persons dealing with AI systems on their behalf, may also serve as inspiration to Korean
policymakers.
Special support should be provided to SMEs
Boosting AI adoption among SMEs in Korea could result in significant economic gains, particularly given
the large share of employment that they represent. The productivity of SMEs in Korea is only about
one-third of that of large companies (compared to around half in other OECD countries) (OECD, 2024[1])
and AI adoption is also significantly lower among SMEs than it is among larger firms: % in firms with
50 to 249 employees, compared to % in firms with more than 250 employees (NIA, 2025[9]). At the
same time, SMEs that use AI in Korea are very positive about its impact: nearly half of the SMEs
experiencing skills shortages believe that generative AI has helped address them, 4 in 5 say it has
increased employee performance, and around 2 in 5 say it has saved money, increased revenue, and
helped them compete against larger companies (Figure ).
18
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
Figure . SMEs in Korea are positive about the impact of generative AI
Percentage of SMEs reporting generative AI helped the company
Note: The total reflects the combined results from all countries participating in the survey (Austria, Canada, Germany, Ireland, Japan, Korea,
the United Kingdom).
Source: OECD (2025[11]), Microdata from the OECD SME Survey on Generative AI.
It is encouraging to see the AI Basic Act mention “education support related to the introduction and
utilisation of artificial intelligence technology for […] employees of small and medium-sized enterprises” as
well as “support for funds used for the introduction and use of artificial intelligence technology by small and
medium-sized enterprises.” At the same time, Korea should ensure that such support goes hand-in-hand
with efforts to consolidate the various other forms of support and protections for SMEs, to reduce the risk
that these policies lock resources into low-productive uses and thereby hold back overall productivity
(OECD, 2024[1]).
References
Bresnahan, T. and M. Trajtenberg (1992), “General Purpose Technologies “Engines of
Growth?””,
[7]
Brynjolfsson, E., D. Rock and C. Syverson (2017), “Artificial Intelligence and the Modern
Productivity Paradox: A Clash of Expectations and Statistics”, National Bureau of Economic
Research Working Paper Series, (accessed on 25 July 2024).
[6]
Filippucci, F., P. Gal and M. Schief (2024), “Miracle or Myth? Assessing the macroeconomic
productivity gains from Artificial Intelligence”, OECD Artificial Intelligence Papers, No. 29,
OECD Publishing, Paris,
[3]
Green, A. (2024), “Artificial intelligence and the changing demand for skills in the labour market”,
OECD Artificial Intelligence Papers, No. 14, OECD Publishing, Paris,
[13]
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
Increase revenue Perform tasks that
could not be
performed before
Offer new products or
services
Save money Improve employee
performance
Compete with larger
companies
Total Korea
19
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
Han, J. (2023), 인공지능으로 인한 노동시장의 변화와 정책방향 (The Impact of Artificial
Intelligence on the Labor Market and Policy Implications), Korea Development Institute (KDI),
[8]
KOSIS (2025), Survey of Business Activities 2023 - Technology Development and Utilization:
Artificial Intelligence,
FstatHtml%%3Flist_id%3DJ2_7_300_009%26obj_var_id%3D%26seqNo%3D
%26tblId%3DDT_1EP1337_1%26vw_cd%3DMT_ZTITLE%26orgId%3D101%26path%3D%2
52FstatisticsList%%26conn_path%3DMT_ZTITLE%26itm_id%3D%
26lang_mode%3Dko%26scrId%3D%26 (accessed on 18 June 2025).
[10]
Lane, M., M. Williams and S. Broecke (2023), “The impact of AI on the workplace: Main findings
from the OECD AI surveys of employers and workers”, OECD Social, Employment and
Migration Working Papers, No. 288, OECD Publishing, Paris,
[12]
NIA (2025), 2024 Yearbook of Enterprise Informatization Statistics, National Information Society
Agency and Ministry of Science and ICT, Republic of Korea,
(accessed on 25 April 2025).
[9]
OECD (2025), “Microdata from the OECD SME Survey on Generative AI”. [11]
OECD (2025), OECD Employment Outlook 2025: Can We Get Through the Demographic
Crunch?, OECD Publishing, Paris,
[2]
OECD (2024), “Explanatory memorandum on the updated OECD definition of an AI system”,
OECD Artificial Intelligence Papers, No. 8, OECD Publishing, Paris,
[5]
OECD (2024), OECD Economic Surveys: Korea 2024, OECD Publishing, Paris,
[1]
OECD (2023), OECD Employment Outlook 2023: Artificial Intelligence and the Labour Market,
OECD Publishing, Paris,
[4]
20
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
While most OECD economies are still in the early phases of AI adoption,
the technology is expected to have a profound impact on the world of work.
This chapter explores the evidence from across the OECD as well as from
Korea on how AI has so far affected: job quantity and skills, job quality, and
inclusiveness in the labour market. So far, there is little evidence of a
negative impact on the number of jobs, although some groups are more
affected than others. AI holds promise to improve job quality, but there are
risks too. Moreover, these risks and benefits of AI are not equally
distributed across population sub-groups.
2 The impact of AI on the labour
market
21
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
In Brief
The impact of AI on job quantity and skills
• AI has made the most progress in non-routine, cognitive tasks. Therefore, the occupations most
exposed to AI tend to be white-collar occupations, such as IT professionals, business
professionals, managers, and science and engineering professionals. However, high exposure
to AI does not necessarily imply workers in these occupations will be displaced. So far, across
OECD countries, there is little evidence of negative aggregate employment outcomes due to AI.
• In Korea, there is some evidence that over the period 2018 to 2023, more “traditional” AI was
associated with lower growth in full-time, permanent jobs, particularly in the manufacturing
sector. However, no such relationship was found for generative AI. These findings need to be
interpreted in a context of % employment growth in full-time, permanent jobs overall during
the same period, as well as % growth in total employment (including non-standard forms of
work). In addition, in a survey of Korean firms, % report no workforce changes so far at the
department- or team-level following the adoption of AI.
• These limited effects on job automation suggest that, by itself, AI will not solve labour shortages
in Korea. However, AI can help mitigate them. For instance, of the 37% Korean SMEs reporting
a worker shortage in the last two years, 27% say that generative AI helps compensate for these
shortages. Similarly, 24% of Korean SMEs report a lack of skills and experience among staff,
and 47% of these say that generative AI helps to address this challenge.
• While AI may not automate jobs at large scale, it does change the tasks workers do and the
skills required of them. % of Korean firms that have adopted AI say it has replaced specific
tasks within existing jobs. Moreover, % say the use of AI has resulted in an increase in the
kinds of skills required to carry out current tasks, and % say that AI has increased the level
of skills required. Firms in Korea that have adopted AI are more likely to report an increase in
communication among team members (% v. %), percentage), with managers (% v.
%), as well as between teams (27% v. %). For Korean SMEs, AI increases the importance
of data analysis and interpretation skills increases the most, followed by programming and
coding skills.
The changing skills needs resulting from the adoption of AI in the workplace call for new training
opportunities. This is particularly important in the context of the continued brain train from Korea,
including of AI talent. However, Participation in adult learning in Korea is the lowest across
OECD countries: 13%, compared to an OECD average of 40%. While firms in Korea do provide
training to employees for working with AI, only 42% of those that have adopted AI have done
so, and the share is higher in large firms than it is in small ones.
The impact of AI on job quality
• Wages are a key dimension of job quality. If AI boosts productivity then it could result in higher
wages for workers. In OECD countries, the wage benefits of AI have so far been concentrated
among high-income and highly skilled workers. Similarly, in Korea, only the occupations most
exposed to generative AI have benefited from higher wage growth.
22
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
• AI could improve job quality in other ways as well. The automation of tedious and repetitive tasks
could improve job enjoyment and allow workers to focus on more complex and interesting tasks.
AI could also improve physical safety by automating dangerous tasks and improving monitoring
systems and safety procedure controls. At the same time, there are some risks too. For example,
case study evidence from OECD countries shows instances of increased work intensity due to
higher performance targets or complexity induced by AI. Ultimately, the effect of AI on the work
environment depends on how thoughtfully and strategically it is integrated into workplace
practices.
• In Korea, AI appears to improve job satisfaction. However, there is a gap between the
perceptions of firms and employees, with the former being more positive than the latter. In
addition, many workers in Korea – particularly those in smaller firms and the manufacturing
sector – report no noticeable reduction in either physical or mental burden following the adoption
of AI. One possible explanation is that AI adoption in Korea is still in its early stages, and its
potential to ease work intensity has not yet fully materialised.
The impact of AI on inclusiveness
• Workers vary in the extent to which they are exposed to AI, but also in their ability to adapt to
and benefit from new technologies. Thus the impact of AI need not be uniform across different
socio-demographic groups. So far, the evidence from OECD countries suggests that high-
income and high-skilled workers benefit the most from AI, while low-skilled workers may lose
out. For example, the impact of AI on employment growth has been found to be significant and
positive for high-income and high-skilled occupations, and for jobs where computer use is high.
Similarly, high-income and high-skilled occupations, as well as jobs with high computer use,
tend to experience positive effects on wage growth associated with AI exposure, while lower
income and lower skilled workers do not seem to benefit in the same way, or less so.
• In Korea, the negative impact of traditional AI on regular, full-time employment growth appears
to be concentrated among younger workers, low- to medium-skilled workers and women –
although for the latter, as well as for high-skilled workers, higher exposure to generative AI is
associated with higher employment growth. Furthermore, generative AI is associated with higher
wage growth for men and high-skilled workers, while traditional AI is associated with higher wage
growth for older workers and high-skilled workers. By contrast, traditional AI appears to reduce
wage growth for low-skilled workers. These findings apply to full-time, permanent employees
only, and the findings for non-standard workers may be different.
23
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
The impact of AI on job quantity and skills: Evidence from OECD countries
AI has made the most progress in non-routine cognitive tasks, therefore affecting mostly
white-collar occupations
Recent advances in AI have extended the types of tasks that can be automated to non-routine, cognitive
tasks, exposing workers who were previously relatively protected from the risk of automation (. the high-
skilled). In the past, computers and robots followed strict rules set by programmers and therefore could
only automate routine tasks, affecting mostly low- and medium-skilled workers (Autor, Levy and Murnane,
2003[1]).
As of 2023, AI has exceeded human performance across various tasks. It outperformed human baselines
in image classification1 as early as 2015, basic and medium-level reading comprehension in 2017 and
2018 respectively, visual reasoning2 in 2020 and natural language inference3 in 2021 (Maslej et al.,
2024[2]). Advancements in Natural Language Processing (NLP), and in particular in Large Language
Models (LLM), enable applications like Generative AI to perform a wide range of language and cognitive
tasks, often at a level comparable to humans and much faster. Generative AI refers to AI systems capable
of creating new content based on patterns learned from existing data. For instance, ChatGPT, Gemini, or
HyperCLOVA X in Korea, can write poems, computer code, and essays, compose music, and explain
complex scientific ideas to a broader audience. When evaluated against answers given by experts on
different questions, ChatGPT performance has been assessed as good as that of a team of experts (Guo
et al., 2023[3]).
AI can now answer around 80% of the literacy and two-thirds of the numeracy questions included in the
OECD Survey of Adult Skills of the Programme for International Assessment of Adult Competencies
(PIAAC). Comparing these results to those of the adults performing the tests highlights the potential for AI
to outperform large portions of the adult population in reading and mathematics. Experts predict that
increasing investments in AI research and development, in particular in NLP, will lead to further significant
advancements of AI in both reading and mathematics over the coming years (OECD, 2023[4]).
Important progress has also been made in AI’s ability to replicate psychomotor abilities, specifically: the
ability to work in cramped workspace, finger dexterity and manual dexterity. Finger dexterity refers to the
ability to make precisely co-ordinated movements of the fingers to grasp, manipulate, or assemble very
small objects. Manual dexterity, by contrast, is the ability to quickly move the hand, the hand together with
the arm, or the two hands to grasp, manipulate, or assemble objects. Older technologies, such as robots,
are being improved through the integration of AI (Lassébie and Quintini, 2022[5]).
AI has made the least progress in physical abilities such as static, dynamic, and trunk These are
tasks more common in non-cognitive, non-routine occupations such as dancers, athletes, bricklayers, and
farm There are also other skills and abilities that humans still have a comparative advantage in,
such as negotiation, social perceptiveness, assisting and caring for others, originality, and persuasion.
Bringing people together and reconciling different views, understanding why people react a certain way,
or providing emotional support, all remain complicated tasks for machines to perform (Georgieff and Hyee,
2021[6]; Lassébie and Quintini, 2022[5]). As of 2023, AI fails to exceed human ability also in some more
complex cognitive tasks, such as visual commonsense reasoning6 and advanced level mathematical
problem solving (Maslej et al., 2024[2]).
Measures of AI exposure, such as the one constructed by Felten, Raj and Seamans (2021[7]), evaluate the
overlap between the abilities required in an occupation and the technical capabilities of AI. The occupations
most exposed to AI are white-collar occupations, which are most likely to involve non-routine cognitive
tasks requiring formal training and/or tertiary education, such as IT professionals, business professionals,
managers, and science and engineering professionals. Occupations requiring manual skills and strength,
such as cleaners, agricultural forestry and fishery labourers, food preparation assistants and labourers,
24
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
are the least exposed to AI (Figure ) (Lane, 2024[8]; Georgieff and Hyee, 2021[6]). Focusing on
generative AI, Eloundou et al. (2023[9]) observe that most occupations exhibit some degree of exposure to
LLMs. Occupations with higher wages and information processing industries exhibit high exposure, while
manufacturing, agriculture, and mining industries demonstrate lower exposure (Eloundou et al., 2023[9]).
Felten, Raj and Seamans (2023[10]) also find that occupations with higher wages are more likely to be
exposed to rapid advances in language modelling, and that education and legal service sectors exhibit
higher exposure. Box presents data on AI adoption based on survey results.
Figure . White collar occupations are more exposed to AI than occupations requiring manual
skills and strength
Average AI exposure by occupations, 2022
Source: Average AI exposure scores retrieved from Lane (2024[8]), “Who will be the workers most affected by AI?: A closer look at the impact
of AI on women, low-skilled workers and other groups”,
0 1
Cleaners, helpers
Agricultural forestry, fishery labourers
Food preparation assistants
Labourers
Refuse workers, other elementary workers
B. Five least exposed occupations
0 1
Science, engineering professionals
Chief executives
Managers
Business professionals
IT technology professionals
A. Five most exposed occupations
25
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
So far, in OECD countries, AI seems to have had no significant negative impact on
overall employment
High exposure to AI does not necessarily imply workers in these occupations will be displaced.
Theoretically, there are various channels through which the introduction of AI in the workplace could impact
labour demand. Firstly, AI can substitute workers by automating tasks previously performed by human
labour (displacement effect). Secondly, as some tasks are automated and AI can complement workers
helping them perform tasks more efficiently, productivity increases and costs are reduced. This leads to
lower quality-adjusted prices, potentially increasing product/service demand and, consequently, the
demand for workers essential in the production process (productivity effect). Lastly, AI can create new
tasks and jobs, particularly in AI development and maintenance (reinstatement effect). Therefore, the
overall effect of AI on labour demand is theoretically ambiguous and depends on which effects dominate
(Acemoglu and Restrepo, 2019[13]). To understand the impact of AI on aggregate employment empirical
research is needed.
So far, across OECD countries, there is little evidence of negative aggregate employment outcomes due
to AI. Instead, there appears to be a slight positive relationship between AI exposure and employment
growth, suggesting that AI may be creating more jobs than it is destroying. At the same time, specific AI
technologies could have different, and in some cases negative, impacts. What most studies highlight, is
that while more jobs may be impacted by AI, very few are at risk of disappearing entirely. Most occupations
involve a combination of skills and abilities that can and cannot be automated. Even highly impacted
occupations are unlikely to be fully replaced by automation. Instead, work may need to be organised
differently, and workers in these roles may require retraining as technology takes over certain tasks
(Lassébie and Quintini, 2022[5]).
Case studies carried out by the OECD in the finance and manufacturing sectors of 8 OECD countries7 in
2022 showed that for 23% of the firms interviewed, AI technologies reduced the number of jobs in the most
affected occupations. However, most firms managed these reductions by reallocating workers within the
company or through attrition, keeping employees until they either left voluntarily or retired. In addition, firms
often opted to slow hiring instead of implementing job cuts, using this approach as a safeguard against the
Box . AI Adoption: Evidence form surveys
AI is increasingly recognised as a transformative technology with the potential to significantly impact
workplaces. As a result, there is growing interest in understanding how widely these technologies are
adopted by companies. According to Information and Communication Technology (ICT) surveys conducted
by National Statistical Offices in 2024, the average AI adoption rate across OECD countries is 14%.
Adoption rates vary by firm size. On average across OECD countries, 40% of large firms use AI, compared
to 20% of medium-sized firms and just 12% of small firms (OECD, 2025[11]). For some countries, the latest
available data predate the release of ChatGPT and other forms of generative AI. If more recent data were
available for these countries, the average OECD AI adoption rate would likely be higher.
While European surveys benefit from standardised questions that facilitate cross-country comparisons,
surveys from other regions often differ in design and definitions, posing challenges to comparability. New
data on AI adoption in SMEs (small and medium-sized enterprises) has emerged from an OECD survey
conducted between October and December 2024 examining the impact of generative AI on SMEs’ labour
and skill needs (OECD, 2025[12]). The results highlight significant cross-country differences in AI adoption
by SMEs, ranging from 27% in Japan to 51% in Germany. Korea is on the lower end of the scale, with 31%
of SMEs saying they’ve adopted AI (Figure ).
26
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
potential failure or underperformance of AI solutions (Milanez, 2023[14]). This is consistent with the finding
by Acemoglu et al. (2022[15]) that firms more exposed to AI reduce their overall hiring. Only a handful of
studies, exploiting variation in AI adoption across US commuting zones, have found a negative effect of AI
exposure on employment overall (Huang, 2024[16]; Bonfiglioli et al., 2025[17]).
The majority (77%) of the firms participating in the aforementioned case studies reported no impact on the
quantity of jobs for workers most affected by AI technologies. Half of these firms implemented AI
technologies to boost production volumes or improve product or service quality, rather than to reduce
labour costs. For the other half, the implementation of AI led to the reorganisation of jobs, with workers
displaced from certain tasks reassigned to other existing or new tasks. In some cases, AI technologies
automated tasks that constituted only a minor share of workers’ jobs, thus not leading to displacement. In
other cases, job reorganisation affected more substantial shares of workers’ tasks. However, these jobs
were not eliminated, and the automation of certain tasks allowed workers to focus on more complex tasks
that could not yet be automated (Milanez, 2023[14]). Additionally, 83% of SMEs report that the use of
generative AI has had no effect on the overall number of staff they need (OECD, 2025[12]). Similarly, several
studies do not find a significant relationship between AI exposure and aggregate employment (Felten, Raj
and Seamans, 2019[18]; Georgieff and Hyee, 2021[6]; Acemoglu et al., 2022[15]). However, it is possible that
significant impacts on aggregate economic data only become detectable once the technology is widely
adopted and the necessary complementary processes and assets are developed (Brynjolfsson, Rock and
Syverson, 2017[19]; Acemoglu et al., 2022[15]; Lane, 2024[8]).
Studies by Albanesi et al. (2023[20]) and Lane (2024[8]) have found a small, positive and statistically
significant effect of AI exposure on aggregate employment, although direct causality is difficult to prove.
The positive association between AI exposure and employment could be due to a productivity effect, or
because AI creates new jobs directly. Green and Lamby (2023[21]) find that employment growth for the AI
workforce, defined as workers with the skills necessary to develop and maintain AI systems, is strong. On
average employment growth was 63% for the AI workforce between 2017 and 2019–although this
workforce is still relatively small, representing less than % of workers overall. Acemoglu et al. (2022[15])
also find a rapid take-off of AI vacancy postings starting in 2010 and accelerating around 2015-2016.
Moreover, in 30% of the OECD case studies, interviewees noted that employment was increasing in
occupations related to the development and maintenance of AI (Milanez, 2023[14]).
AI could help mitigate labour shortages
Labour shortages are becoming a critical concern across many OECD countries. Labour market tightness,
measured as the number of vacancies per unemployed person, has eased in the last quarter of 2023 but
continues to exceed pre-COVID-19 levels in many countries (OECD, 2024[22]). Population ageing is a
significant factor contributing to this challenge. As the workforce shrinks and demand for services like
healthcare grows, innovative solutions are needed to avoid significant skills and labour shortages.
AI could help address these challenges by automating tasks and by enhancing worker productivity,
enabling a more efficient use of resources, and making organisations better equipped to manage with a
reduced workforce. AI could also support healthcare professionals by, for example, serving as a
documentation assistant reducing the time spent on administrative tasks, or by assisting radiologists in
scanning medical images, freeing up time for doctors to spend on care (Anderson and Sutherland,
2024[23]).
Furthermore, AI could help extend working lives and increase the labour market participation of the elderly.
Many physically demanding jobs can lead to muscular-skeletal problems, but AI could be used to protect
workers from injury, enabling them to remain employed longer. For example, the company German
Autowerks invested in AI to analyse videos of mechanics at work, identifying pressure points and potential
problem areas on the body. This information was then used to select specific exoskeletons which make
heavy tasks much easier to perform. The company opted not to automate these tasks, believing that
27
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
humans are more flexible and adaptable to changing job requirements. Instead, they focussed on using AI
to assist workers, helping them work for longer (Machin, 2024[24]).
Despite the potential of AI to address challenges associated with labour shortages, it can only be part of a
wider package of solutions to tackle these issues. Even if AI, particularly since the advent of Large
Language Models, could be applied to a substantial share of tasks done by workers (Eloundou et al.,
2023[9]), there are still tasks AI cannot do (or that society would not find acceptable for AI to do) and it
cannot therefore fully replace workers (Lassébie and Quintini, 2022[5]). In addition, while several
experimental studies show that AI could significantly enhance worker productivity in certain tasks (see
section on Equalisation of performance within occupations below), the extent of this impact on aggregate
productivity remains a topic of debate. This uncertainty is reflected in the ‘’productivity paradox’’, which
refers to the lag in productivity growth over the past decade despite advancements in AI and other
technologies. One possible explanation is that the aggregate productivity gains from AI might be modest
(Acemoglu, 2024[25]). Alternatively, delays in AI implementation and organisational restructuring could
mean that substantial economic gains from AI may take years or even decades to materialise (Lane and
Saint-Martin, 2021[26]).
In OECD countries, AI has increased the need for new skills, including specialised AI
and analytical skills
The integration of AI into the workforce has expanded the demand for specialised AI skills needed for the
development and maintenance of AI systems. However, these positions still only represent a small fraction
of total employment (Green and Lamby, 2023[21]). Most workers will have to interact with AI applications
which often feature user-friendly interfaces, requiring only basic digital skills. The demand for skills
complementary to AI – such as cognitive, management, social, and digital skills – appears to be increasing
overall, while that for routine skills might decrease. Nonetheless, research also suggests these increases
might not be related to AI per se, and AI exposure might be associated with a fall in demand for some of
those skills.
The demand for AI skills in the labour market is increasing. Using data on skill requirements in online
vacancies, Alekseeva et al. (2021[27]) show that the demand for these skills quadrupled over the period
2010 to 2019. The skills most demanded in AI vacancies are machine learning, natural language
processing, deep learning, image processing, programming languages like Python, and big data
management (Alekseeva et al., 2021[27]; Manca, 2023[28]; Squicciarini and Nachtigall, 2021[29]). These skills
will be needed not only to design algorithms, but also to explain their functioning to non-technical
professionals, and to monitor outcomes to make sure that AI systems are operating as intended, detecting
mistakes and potential biases, and addressing any unintended consequences (Wilson, Daugherty and
Morini-Bianzino, 2017[30]). Job postings that require specialised AI skills also tend to ask for high-level
cognitive skills such as creative problem solving, social skills and management skills (project and people
management), suggesting that these skills are complementary to AI. Conversely, these jobs typically do
not require routine skills, like general administrative and clerical skills. As a result, an increase in AI-related
employment is likely to drive demand for high-level cognitive skills while decreasing demand for routine
skills (Alekseeva et al., 2021[27]; Manca, 2023[28]).
Nonetheless, most workers who will interact with AI may not need AI-specific skills or a deep understanding
of AI systems. A survey of AI start-ups found that only 10% required users of their AI products to have
expert coding or data skills, while 59% required only general computer familiarity, and the remainder
required no specialised skills at all (Bessen et al., 2023[31]).
In many cases, AI adoption has not yet significantly changed skill requirements within firms. In 2022, 57%
and 48% of firms that had adopted AI in finance and manufacturing, respectively, reported no change in
skill needs (Lane, Williams and Broecke, 2023[32]). Similarly, in case studies of firms having implemented
AI in finance and manufacturing, 60% of firms said that AI adoption had not yet modified skill requirements
28
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
(Milanez, 2023[14]). This could be partly because AI adoption at the time of those studies was still relatively
low and many firms were only experimenting with the technology, but also because interacting with AI
applications often requires only basic digital skills, such as the ability to use a computer or smartphone,
relying on existing skills.
That being said, 40% of the firms interviewed as part of the above-mentioned case studies reported a need
for new skills, including specialised AI skills and analytical skills. As simple tasks become automated, the
proportion of complex tasks performed by workers rises, necessitating specialised knowledge and
advanced analytical skills, such as the ability to comprehend and apply new ideas (Milanez, 2023[14]).
Managers using algorithmic management software report that the use of such tools mostly increases their
need for the ability to use or interpret data, and for digital skills (Milanez, Lemmens and Ruggiu, 2025[33]).
Employers also say that, while AI has increased the importance of specialised AI skills, it has increased
the importance of human skills, such as creativity and communication, even more, as well as the need for
highly educated workers more generally (Lane, Williams and Broecke, 2023[32]). Green (2024[34]) shows
that occupations with high AI exposure predominantly demand management, business processes, social
and digital skills, with the largest increase in demand for skills related to collaboration, originality, and basic
office tools.
At the same time, there is tentative evidence of a relative decline in the demand for management, business
process, cognitive, digital, emotional and communication skills in workplaces that are highly exposed to AI
(Green, 2024[34]). These skills are the skills of white-collar support occupations: finance, human resources,
legal, communications, administrative assistants and project managers. These effects are modest and
should be viewed as relative changes amidst an overall increase in demand for most of these skills in the
aggregate (Green, 2024[34]). In addition, however, managers using algorithmic management tools are more
likely to report decreases in human interactions than increases. In the European countries surveyed,8
managers were also more likely to believe that the use of such tools was decreasing managers’ need for
empathy rather than increasing it (Milanez, Lemmens and Ruggiu, 2025[33]).
These findings point to an additional concern, which is the potential deskilling of the workforce as a result
of AI adoption. The case studies carried out by the OECD in the manufacturing and finances sectors of
eight OECD countries documented some instances of deskilling, where the machine performed the skilled
tasks, and the worker was only required to operate a very intuitive system, with no judgment involved
(Milanez, 2023[14]).
Most OECD countries will need to ramp up training provision to address AI-induced
skills demand
The changing skills needs resulting from the adoption of AI in the workplace call for new training
opportunities. While initial education plays a crucial role in equipping workers with the skills to work with
AI, upskilling and reskilling the existing workforce will be equally important to help individuals adapt and
prepare for the transition (OECD, 2024[35]). Older adults and lower skilled workers in particular will need to
acquire basic digital skills essential for interacting with AI technologies. Meanwhile, managers and
business leaders require training to efficiently organise the integration of AI into their operations.
More than half of workers who use AI in the manufacturing and finance sectors said that their company
had either provided or funded training so that they could work with AI. Yet, more training would help
address existing barriers to AI adoption, considering that around 40% of employers in those sectors
declared that the lack of relevant skills was a barrier to AI adoption (Lane, Williams and Broecke, 2023[32]).
When the AI technology is simple to use, training can be brief and take the form of webinars, presentations,
or workshops (Milanez, 2023[14]). In a small number of case studies, large firms operated more ambitious
training programmes to help employees transition to other occupations. Some large companies try to grow
AI talent in-house instead of seeking those employees on the external labour market. However, several
29
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
firms call for more government funding for AI education and training, recognising that these specialised AI
skills should also be developed in initial education (Milanez, 2023[14]).
Initial education plays a crucial role in acquiring the skills necessary to develop and maintain AI systems,
with two-thirds of the AI workforce holding a tertiary degree. The share of AI workers who report having
participated in some sort of training in the last four weeks is similar to that of the entire population with a
tertiary degree (16% and 18%, respectively). Most of the training that the AI workforce undertakes is non-
technical in nature. Nonetheless, workers who have skills closely related to AI skills may acquire more
explicit AI skills simply by being part of a research team or the AI development process within their firms
(Green and Lamby, 2023[21]).
Lower-skilled and older workers are less likely to possess the basic digital skills required in a workplace
transformed by the adoption of AI. Based on the Survey of Adult Skills, that tests adults in basic information
processing skills, around one in four adults (aged 16-65) have no or only limited experience with computers
or lack confidence in their ability to use them. Additionally, nearly half of all adults can only use familiar
applications to solve problems involving few steps and explicit criteria, such as sorting emails into
pre-existing folders. Among low-educated adults, 41% lack basic proficiency in using information and
communications technology (ICT) to even take the survey’s test, and those who can undertake the test
perform poorly. The percentage of adults without basic ICT skills decreases to 15% for those with upper
secondary education and 4% for those with tertiary education. Compared to younger adults (aged 25-34)
older adults (aged 55-65) are significantly more likely to have no computer experience and lower scores.
Among young adults, 8% have no computer experience and 43% perform well. In contrast, 34% of older
adults have no computer experience, and only % perform well (OECD, 2019[36]).
Moreover, older and lower-skilled adults, as well as low-wage workers, are less likely to take part in adult
learning in every single country participating in the Survey of Adult Skills. Considering that around half of
all adults neither participate nor want to participate in adult learning, it will be crucial to find effective ways
to address barriers and motivation to training participation (OECD, 2019[37]). For example, the provision of
more flexible learning options (. part-time study or online delivery) would allow learners to better balance
training alongside work or other commitments (OECD, 2024[35]).
Managers also need training to effectively organise the integration of AI into their operations. They need
to understand AI systems to assess where and how innovation can be utilised within the company, identify
the benefits and risks of AI, and determine the best ways to integrate AI systems into existing processes.
Managers would have to decide which tasks are better performed by AI systems and which by humans,
recognising the strengths and weaknesses of each (OECD, 2023[38]). A German insurance provider
reported that managers planning AI projects are expected to have a minimum knowledge of how the
technology works (Milanez, 2023[14]). In the OECD survey on algorithmic management in the workplace
(2025[33]), 75% of managers report their firms offer training on how to use the software. Training for
managers could enhance their proficiency with specific tools, deepen their understanding of the data used
by these tools, and ensure their skills keep pace with the growing demand for analytical capabilities.
Additionally, it could help them use software in a trustworthy manner (Milanez, Lemmens and Ruggiu,
2025[33]).
Encouragingly, several OECD countries have developed dedicated AI training strategies (OECD, 2024[35]).
Many have introduced incentives to support employers in providing training for their employees. Fourteen
governments have invested in publicly funded AI training programmes, with nine focussed on developing
AI professionals and seven aimed at enhancing AI literacy for the general public. More broadly, publicly
funded digital skills training, without an explicit focus on AI, is more common (OECD, 2024[35]).
30
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
The impact of AI on job quantity and skills: Evidence from Korea
The adoption and use of AI in the workplace in Korea is lower than in other countries
The adoption of AI in Korea appears low compared to other OECD countries (% in 2024 from Han
(2023[39]), % in 2023 from 2024 Survey of Business Activities (KOSIS, 2025[40]), % in 2023 from
the 2024 Enterprise Informatization Statistics (NIA, 2025[41]) and see Figure ) and firms appear to take
a cautious stance towards AI, as shown by their investments in AI (Box ).
Box . Firm investment in AI in Korea
The majority of firms that participated in the Survey on AI Utilisation and Labour Market Changes
conducted by the Korea Labor Institute invest 5% or less of their sales revenue in AI development
(Figure , Panel A). Among the companies that indicated they would increase AI investment in the future,
a majority (%) reported that their investment would be limited to 5% or less of their total sales revenue,
suggesting a cautious stance on AI development expenditures in Korea (Figure , Panel B). This stance
toward AI is due to concerns about the maturity of the technology and uncertain returns. Many firms prefer
a gradual and incremental approach.
Figure . Currently, most Korean firms only invest a small share of sales revenue in AI, and even
planned investment in AI by Korean firms is relatively low
Percentage of firms reporting investment of 5% or less (of sales revenue), 6-10%, 11-20% and 21% or more
Note: The survey targeted firms that use AI, focussing on industries classified under the Korean Standard Industrial Classification (KSIC),
specifically: Manufacturing, Information and communication, Professional scientific and technical service, Healthcare. Only firms that utilise AI
and have 10 or more employees were included in the survey. HR managers and AI developers provided the survey responses. The survey was
conducted over a two-month period, from 20 October 2024 to 31 December 2024. The survey covered a population of 9 625 establishments,
including 3 292 in manufacturing, 3 118 in information and communication, 1 788 in professional and scientific services, and 790 in healthcare.
The sample was drawn using a random sampling method, with a target sample size of 200. Ultimately, the study achieved valid responses from
145 firms, whose data were incorporated into the final analysis.
Source: Survey on AI Utilisation and Labour Market Changes conducted by the Korea Labor Institute (2024).
0%
10%
20%
30%
40%
50%
60%
70%
80%
5% or fewer 6-10% 11-20% 21% or more
(A) Current investment
0%
10%
20%
30%
40%
50%
60%
70%
80%
5% or fewer 6-10% 11-20% 21% or more
(B) Planned investment
31
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
Even if AI is adopted by organisations in Korea, that does not mean that employees use it frequently.
% of workers who use AI, use it once or twice a week, while % used it once or twice a day
(Figure ). According to the Survey on AI Utilisation and Labour Market Changes conducted by the Korea
Labor Institute as well as qualitative evidence, employees within the same team or department differ
significantly in how frequently and intensively they use AI. AI users typically begin experimenting with the
technology out of curiosity. As they gain experience, they become more comfortable and gradually expand
their use of AI across various tasks. Users emphasise that AI reduces the time and effort needed to
visualise design concepts, allowing for greater efficiency. In contrast, non-users tend to avoid engaging
with AI altogether. They cite two main reasons: first, they see no meaningful application of AI in their
specific tasks; second, they believe their current manual methods are faster and more accurate. This
variation in AI adoption, even among employees performing identical work, highlights the importance of
understanding individual-level factors that influence technology acceptance in the
provides a concrete example of this in a publishing firm in Korea.
Figure . Nearly 1 in 3 Korean workers uses AI 1 to 2 times a day
Percentage of employees reporting using AI 1-2 times a week, 1-2 times every day, 1-2 times a day, countless times
Note: The survey targeted individual employees who use AI in firms operating within four industries classified under the Korean Standard
Industrial Classification (KSIC): Manufacturing, Information and Communication, Professional, Scientific and Technical Services, and
Healthcare. Employees using AI provided the survey responses. The survey was conducted over a two-month period, from 20 October 2024 to
31 December 2024. The sample was drawn using a random sampling method, with a target sample size of 600. Ultimately, the study achieved
valid responses from 426 employees, whose data were incorporated into the final analysis.
Source: Survey on AI Utilisation and Labour Market Changes conducted by the Korea Labor Institute (2024).
0%
5%
10%
15%
20%
25%
30%
35%
40%
1-2 times a week 1-2 times every 3 days 1-2 times a day Used countless times
32
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
Box . Understanding individual differences in AI adoption within the workplace: Evidence
from the use of image-generation AI in a book cover design team
A Korean publishing firm employing around 100 people promotes the adoption of AI in the workplace
but does not require employees to use it. AI is applied across multiple departments. A notable example
is a team responsible for designing book covers, where designers use image-generation AI tools. Even
within this team, the level of AI use varies. The team is composed of five designers; two actively use
AI, whereas the other three do not. Those who use AI often begin out of curiosity. As they gain
experience, they become more comfortable with the tools and gradually expand their use across
different tasks. Designers who do not use AI rarely attempt to engage with the technology. AI users
emphasise the reduced time and effort required to visualise design concepts. They report that AI
enables them to generate images more quickly and efficiently. In contrast, non-users argue that their
manual design methods offer superior quality and greater control.
While total employment continues to grow in Korea, some forms of more traditional AI
appear associated with less growth in full-time, permanent employment
So far, in Korea, most research either finds no or a small negative impact of AI on employment overall,
and the impact depends on both the type of technology and the industry. Chang et al. (2024[42]) concludes
that AI-adopting firms experience productivity gains without reducing employment while Han (2023[39]) finds
employment increases in high-skilled jobs and losses in low-skilled jobs. New analysis carried out for this
report shows that more “traditional” AI appears to be associated with lower growth in full-time, permanent
jobs, concentrated in the manufacturing sector, while there is no such association with generative AI
(Box ). In particular, a 10-percentile point increase in exposure to “traditional” AI (Webb, 2020[43])
appears to be associated with % lower growth in full-time, permanent employment. This needs to be
interpreted in a context of % employment growth in full-time, permanent jobs overall during the same
period, as well as % growth in total employment (including non-standard forms of work).
33
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
Box . The impact of AI on full-time, permanent employment growth in Korea: New evidence
from employment insurance data
This box presents new analysis of the impact of AI on employment growth at the occupational level in
Korea, spanning the period from 2018 to 2023. The analysis draws on three primary data sources:
employment insurance data, exposure indices from the studies of Webb (2020[43]) and Felten et al.
(2023[10]), and the routinisation index extracted from the 2020 Korea Dictionary of Occupations,
published by the Korea Employment Information Service (2020[44]).
Employment insurance data contain detailed individual-level information, including worker
demographics (. age, gender, career history), occupation, and wages, as well as information
regarding the region, industry, and size of the establishments where individuals are employed. Notably,
the employment insurance data include annual information on individual labour market transitions and
the diverse characteristics of workplaces. However, as this dataset is specifically used for the
administration of employment insurance, it only covers individuals and workplaces enrolled in the
scheme. In Korea, employment insurance typically applies only to permanent employees, resulting in a
potential bias, as the data predominantly reflects regular, full-time workers (Kwon, 2022[45]). According
to the Economically Active Population Survey, approximately 25% of employed individuals in Korea
were temporary workers as of 2023.
Two distinct AI exposure indices were employed: one developed by Webb (2020[43]) and the other by
Felten et al. (2023[10]). Webb’s (2020[43]) index measures AI exposure by extracting job-related
information from AI patents, representing exposure to more “traditional” AI technologies. In contrast,
Felten et al. (2023[10]) assess exposure based on the degree to which occupations are affected by AI
language modelling capabilities, providing a measure of exposure to generative AI To
account for other automation technologies that influence labour demand and supply within occupations,
it was necessary to control for their effects, distinct from those of AI. Accordingly, the analysis
incorporates software exposure and robot exposure at the occupational level, as calculated by Webb
(2020[43]).
While existing studies on AI and automation technologies typically focus on the US labour market,
variations in occupational structures between countries such as Korea, the US, and EU and its high-
income constituent members, can arise due to diverse factors like industrial structure. To control for
these potential differences and to reflect the exposure to pre-AI automation technologies within Korea’s
occupational classification system, a routinisation index was used. This index, first introduced by Kim,
Koh and Cho (2014[46]), categorises occupations according to three primary attributes – data, people,
and “objects” (. the various tangible assets employed during the execution of job duties)–based on
the 2020 Korea Dictionary of Each occupation is assigned a score between 0 and 2 for
each characteristic, with the highest score determining the final routinisation index. A higher value
indicates lower exposure to routinisation, while a lower value suggests greater exposure.
For the analysis, raw values of the AI exposure indices, as well as the routinisation index, were
converted into percentiles. This approach mitigates potential distortions caused by the clustering of
values within the raw indices, facilitating a clearer understanding of the differences in occupational
exposure. Consequently, the results focus on changes in the dependent variables in response to shifts
in percentiles, rather than the raw values of the indices.
Table presents the findings of the analysis, which explores the impact of the two AI exposure indices
on the growth in full-time, permanent employment in Korea’s labour market over the 2018-2023 period.
The year 2018 was chosen as the baseline, as AI adoption in Korea was minimal prior to this period.
34
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
Model (1) controls for various occupational and industry-level characteristics, such as gender ratio,
average monthly wages, and age distribution. Model (2) incorporates the software exposure index,
while Model (3) includes the robot exposure index. Model (4) further controls for the routinisation index,
and Model (5) accounts for trends in the dependent variable prior to the analysis period (2015-2018).
Unless otherwise noted, the results discussed herein refer to the outcomes from Model (5).
Table . The impact of AI on full-time, permanent employment in Korea: Regression results
(1) (2) (3) (4) (5)
Panel A: Impact of Webb’s AI Occupational Exposure on Log Employment
AIOE Percentile * * *
() () () () ()
Panel B: Impact of Felten’s AI Occupational Exposure on Log Employment
AIOE Percentile
() () () () ()
Obs. 4 633 4 633 4 633 4 633 4 633
Control Variables v v v v v
Software Index
v v v v
Robot Index
v v v
Routinisation Index
v v
Pre-Trend(EMP)
v
Note: AIOE stands for AI Occupational Exposure Standard errors are shown in parentheses, and statistical significance is indicated as
follows: *** 1%, ** 5%, * 10%.
Source: Employment insurance data and information provided by Webb (2020[43]), “The Impact of Artificial Intelligence on the Labor Market”,
and Felten et al. (2023[10]), “How will Language Modelers like ChatGPT Affect Occupations and
Industries?”, compiled by the authors.
The results show that a 10 percentile point increase in exposure to traditional AI is associated with a
% decline in full-time, permanent employment growth, whereas no statistically significant relationship
exists for generative AI. Given that there has been a % increase in permanent employment over this
period (2018 to 2023), these findings suggest that there has been employment growth at median levels
of exposure to traditional AI, but that this positive effect reduces as exposure rises and that the
displacement begins to outweigh the productivity gains at higher levels of exposure to traditional AI.
1. Although Felten et al. (2018[47]) also created an index for traditional AI through expert surveys, analyses using this index did not yield
results significantly different from those obtained from Webb’s exposure index. Therefore, for the purpose of this study, only the Webb and
Felten indices are employed to distinguish the impacts of traditional and generative AI technologies on the labour market.
2. The 2020 Korea Dictionary of Occupations breaks down each occupation into three characteristics: the use of data, interaction between
people, and the use of things. In other words, “objects” collectively refers to the various types of tangible assets employed during the
execution of job duties.
35
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
At the same time, a recent survey showed that % of firms in Korea reported no workforce changes at
the department- or team-level following the adoption of AI. Case studies carried out in Korea suggest this
may be because AI adoption still remains relatively low (see Box ). Among the firms that reported
changes in tasks as a result of AI adoption, only % reported full automation of all tasks in a job
(Figure ). In addition, % of firms indicated that AI had been adopted to perform entirely new tasks
that had not previously existed within the scope of traditional roles, suggesting that AI adoption may not
necessarily be geared towards replacing workers.
Figure . Full automation of tasks by AI is rare in Korea
Percentage of firms reporting that AI creates entirely new tasks, partially replaces tasks, or fully replaces tasks in a
job
Note: The survey targeted firms that use AI, focussing on industries classified under the Korean Standard Industrial Classification (KSIC),
specifically: Manufacturing, Information and communication, Professional scientific and technical service, Healthcare. Only firms that utilise AI
and have 10 or more employees were included in the survey. HR managers and AI developers provided the survey responses. The survey was
conducted over a two-month period, from 20 October 2024 to 31 December 2024. The survey covered a population of 9 625 establishments,
including 3 292 in manufacturing, 3 118 in information and communication, 1 788 in professional and scientific services, and 790 in healthcare.
The sample was drawn using a random sampling method, with a target sample size of 200. Ultimately, the study achieved valid responses from
145 firms, whose data were incorporated into the final analysis.
Source: Survey on AI Utilisation and Labour Market Changes conducted by the Korea Labor Institute (2024).
0%
10%
20%
30%
40%
50%
60%
Newly created Partial task replaced Fully replaced
36
ARTIFICIAL INTELLIGENCE AND THE LABOUR MARKET IN KOREA © OECD/KOREA LABOR INSTITUTE 2025
Box . Current and future workforce changes from AI: Evidence from Korean case studies
KLI carried out interviews with HR and technology managers from firms using AI within the industries
covered by the Survey on AI Utilisation and Labour Market Changes, .: manufacturing
(2 interviewees), information and communications (5), healthcare (3), and professional, scientific, and
technical services (5). The interviews were conducted over approximately one month, from 24 February
2025 to 26 March 2025, either in person or online, depending on the interviewees’ preferences. The
interviews paint a nuanced picture of AI’s impact on automation. While job losses may happen in the
future, automation is currently limited partly because of low adoption but also because AI tends to
automate only parts of, and not the entire job.
Case study 1: Intellectual Property Dispute Resolution
One company in Korea has developed an internal AI tool for foreign language translation and utilises
an internal version of ChatGPT. The AI tool assists with translating foreign language documents (with
90% accuracy), searching case law, and drafting legal documents. The implementation of AI has not
led to any changes in the team’s composition. Employees perceive that AI can perform only about 2 out
of 10 tasks in their overall workload. However, as technology advances, they anticipate that fewer team
members will be needed for the same tasks in five to six years.
Case study 2: Radiology Technologist
In the field of pathology, AI adoption is more advanced than in other areas due to the vast amount of
accumulated data, allowing for extensive AI training. As a result, AI is more actively utilised in pathology
than in any other hospital department. However, one major limitation remains: since organ placement
and size vary across patients, achieving consistently accurate results is still challenging. With further
technological advancements, AI is expected to not only assist in imaging procedures but also take over
diagnostic interpretations. Consequently, the demand for radiologic technologists may decline, as fewer
professionals will be required to perform these tasks.
Case study 3: Manufacturing
Most of the AI models adopted in a manufacturing firm in Korea are used as references for the
corresponding production workers. Workers can shut down and inspect the equipment by an alert of an
AI to a potential equipment failure. Previously, workers would intuitively detect such failures based on
noise, temperature, vibration, or other factors. Now, the AI senses the failure by combining relevant
data. If the AI is accurate and reliable, stricter criteria can be set for the failure threshold. If not, looser
criteria are used. The strictness of these criteria varies across AI models, but many models in this firm
are said to have somewhat loose criteria. This suggests that most AI models are used as references
rather than as a judge which completely replaces human labour. It is also worth noting that even in
cases where AI replaces human labour, only part of the job (one or two tasks) is replaced, not the entire
role. In this sense, AI and human labour work together, and the adoption of AI models rarely affects the
overall employment size of the firm.
The limited impact of AI on human labour stems primarily from the lower accuracy and reliability of the
AI models in this manufacturing firm. Therefore, this does not necessarily mean that AI
