This chapter explores the characteristics of workers using AI in Japanese workplaces. The proportion of AI users in Japan is lower than in other countries, highlighting significant potential for future growth and the need to promote the safe and trustworthy integration of AI in the workplace. Ensuring inclusive access to AI in the workplace is also essential. Men and older workers, those without a university degree, employees of smaller companies, non-regular workers, those in the Medical, healthcare and welfare sector, and workers in low-skilled occupations (Plant and machine operators, assemblers, and Elementary occupations) are less likely to use AI at work. In contrast, employees with disabilities, those managing childcare and/or long-term care responsibilities, high-income workers, those in the Information and communications sector, the Finance and insurance sector, the Scientific research, professional and technical services sector, workers in high-skilled occupations (Managers and Professionals), and those in companies experiencing mild labour shortages are more likely to use AI at work. In parallel, Japanese companies struggle with a lack of talent with the necessary workplace experience and basic AI knowledge to adopt AI in the workplace. Japan must strengthen its efforts to develop and secure such talent. Furthermore, from the perspective of employees, ensuring the accuracy, safety, and reliability of AI remains a crucial issue for promoting broader adoption.
Artificial Intelligence and the Labour Market in Japan
Abstract
This chapter explores the characteristics of AI users in Japanese workplaces. The share of Japanese workers who say they use AI at work is lower than in other countries, and there are disparities in its use. For Japan to fully harness the benefits of AI, it should ensure that more Japanese workers can access safe and trustworthy AI, while also addressing disparities in use of AI.
8.4% of employees in Japan say they use AI at work, and 6.4% use GEAI. Focusing on the finance and insurance and manufacturing sectors only, the proportion of AI users in Japan is the lowest among the countries for which comparable data is available. This suggests that Japan has significant room growing the use of AI in the workplace.
Many employees already view AI as a megatrend shaping the future of work, although perceptions of the extent to which its use will grow vary among workers. 93.0% of Japanese AI users, 82.4% of Japanese AI non-users (among AI adopters), and 50.7% of Japanese AI non-adopters anticipate that the use of AI in the workplace will increase over the next ten years.
There are significant disparities in the proportion of Japanese employees using AI at work by industry sector. The difference between the sectors with the highest and lowest use is 18.8 percentage points (p.p.): 22.9% in the Information and communications sector and 4.1% in the Accommodations, eating, and drinking services sector. Econometric analysis reveals that, compared to workers in Accommodations, eating and drinking services, those in Agriculture, forestry and fisheries, Manufacturing, Information and communications, Transport and postal services, Wholesale and retail trade, Finance and insurance, and Scientific research, professional and technical services, are more likely to use AI at work. In contrast, workers in Medical, healthcare and welfare are less likely to report using AI in their jobs.
Significant disparities also emerge in the proportion of Japanese employees using AI at work by company size, similar to those observed in other countries surveyed. The difference in AI use at work between companies with up to 19 workers and those with 10 000 or more workers is 16.9 p.p.
According to the Bank of Japan’s Tankan for June 2025, Japanese companies are most likely to report labour shortages in the Construction and Accommodation and food services sector. AI has the potential to address labour shortages in various ways, such as through the automation of routine tasks or the standardisation of complex tasks. Employees in Japanese companies experiencing moderate labour shortages make greater use of AI than those in companies with an appropriate level of staffing. However, this trend is not observed among employees in companies facing severe labour shortages. It is not possible to determine causality, however. It may be the case that companies experiencing moderate shortages had previously experienced severe shortages, but that AI has helped them address these shortages. Alternatively, it could indicate that once companies face severe labour shortages, introducing AI becomes more challenging.
Disparities also emerge in the proportion of Japanese employees using AI at work by region. The difference between the highest proportion in Tokyo (13.8%) and the lowest in Shimane (2.5%) is 11.3 p.p. While these differences are largely thought to arise from variations in the characteristics of companies across regions, an econometric analysis controlling for several individual attributes also shows that the rate of AI use is significantly higher in South Kanto (including Tokyo) compared to other regions.
Japanese AI users in the finance and insurance and manufacturing sectors are more likely to be young or middle‑aged, male and more highly educated than non-users, similar to AI users in other countries. Based on an econometric analysis covering all industries and controlling for several individual attributes, Japanese AI users are more likely to be regular employees, have higher incomes, and work longer hours.
Employees with disabilities or those balancing childcare and/or long-term care responsibilities are more likely to use AI at work than their counterparts without such responsibilities. Since the employees with disabilities referred to in these findings are limited to those who participated in the JILPT web survey, caution is needed in generalising this finding to all employees with disabilities. However, these it suggests that AI could be used more extensively to support individuals with disabilities and those seeking to balance their work with caregiving responsibilities.
AI usage is higher among employees engaged in tasks such as analysing data and information, solving problems using creativity, and managing and motivating teams, while it is lower for those doing routine and repetitive tasks as well as physical tasks. This suggests a higher use among workers who do cognitive tasks and is in line with the finding that physical and repetitive tasks have not so much been automated as delegated to non-regular workers.
Managers, Professionals, Technicians, and associate professionals are among the most frequent AI users, while Plant and machine operators, assemblers and workers in Elementary occupations are less likely to use AI at work. The difference between the occupations with the highest and lowest use is 13.5 p.p. A more detailed classification based on ISCO’s DIGIT3 shows that occupations with the highest AI usage include: ICT service managers, Administration professionals, and Finance professionals, while those with the lowest usage include Domestic, hotel and office cleaners and helpers, Personal care workers in health services, and Manufacturing labourers.
This chapter also explores the barriers to implementing AI in the workplace, reflecting both the challenges faced by companies attempting to adopt AI and the hesitation of workers to use it.
The Japanese Government has outlined plans to develop a total of 2.3 million “Human resources for advancing digital technology implementation” between FY2 022 and FY2 026. However, Japanese companies are struggling more than their U.S. counterparts with a shortage of AI-related talent, with the most pressing challenge being a lack of employees capable of promoting AI adoption using their workplace experience and basic AI knowledge. Other issues companies face include: a lack of information about the benefits of AI use, insufficient examples from other companies, difficulties related to the data required for AI training, concerns around costs, and a shortage of AI products and services that are easy for companies to adopt. A recent OECD survey of SMEs reveals that Japanese SMEs are significantly more likely than their counterparts in other countries to cite a lack of skills to use GEAI as a barrier to its adoption. Japanese SMEs are also concerned about AI-related risks, such as copyright, legal or regulatory issues associated with generative AI.
27.6% of Japanese employees do not have specific reasons for not using GEAI at work. Therefore, if effective methods for using GEAI in the workplace are established, there is a possibility that AI use could expand significantly in Japan. On the other hand, 84.4% of those involved in the introduction of GEAI in the workplace at large companies expressed concerns about using GEAI, including: the accuracy, safety, and reliability of GEAI technologies, as well as the technical difficulty of integrating AI with existing systems.
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[1])”
AI can be seen as a General-Purpose Technology (GPT) (Brynjolfsson, Rock and Syverson, 2017[2]). GPTs are characterised by their pervasiveness, inherent potential for technical improvements and innovational complementarities (Bresnahan and Trajtenberg, 1992[3]). 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.
From the perspective of labour policy, the introduction of AI in the workplace is a key issue for the future of work. Japan, grappling with a severe labour shortage caused by a declining birthrate and an ageing population, has implemented measures to support workers who wish to remain in the workforce, change jobs, or find employment. In addition, Japan has aimed to foster diverse working environments and improve job quality by promoting work style reforms, such as addressing the culture of long working hours and eliminating unfair treatment disparities among workers. These initiatives are also intended to drive sustainable economic and wage growth by enhancing labour productivity through streamlined work tasks, optimised work processes, and increased capital investment. AI has the potential to further advance these policy efforts while aligning with the overall direction of Japan’s labour policy so far. In other words, AI holds significant potential to enhance job performance and working conditions, assist workers in transitioning to more desirable jobs, and improve access to the labour market.
At the same time, maximising the benefits of AI in the workplace requires comprehensive measures and preparations. First and foremost, it is critical to ensure that AI adoption does not exacerbate inequality, contribute to discriminatory practices, facilitate the misuse of surveillance technologies, or infringe upon workers’ privacy. To achieve this, it is essential not only to enhance the safety and reliability of AI technology itself but also to conduct worker consultations during its introduction. Additionally, establishing internal guidelines for employees on AI use can be an effective approach. These guidelines can serve as an ongoing communication tool between labour and management, facilitating discussions on the appropriate use of AI after implementation. They also enable a prompt response to any new issues that may arise post-implementation.
Furthermore, AI integration in the workplace has the potential to reshape job quantity, work tasks, and the skill requirements for various occupations. Therefore, governments must take proactive steps to support workers in reskilling and upskilling to work with AI, as well as in making smooth transitions to new jobs that leverage their existing skills and experience, while making continuous efforts to appropriately understand the actual and potential changes occurring in the world of work.
This report aims to evaluate the use of AI in Japanese workplaces from multiple perspectives and sets out recommendations for policymakers in Japan to support the development of human-centred workplaces empowered by ethical and trustworthy AI. This report primarily consists of findings from thorough analyses of microdata from two worker surveys: one conducted by the OECD (Lane, Williams and Broecke, 2023[4]) and the other by the Japanese Institute for Labour Policy and Training (JILPT) (JILPT, 2025[5]). Particularly, the latter survey is valuable as it was the first large‑scale, all-industry survey conducted by JILPT, an independent administrative agency under the Ministry of Health, Labour and Welfare (MHLW) in Japan, focussing specifically on the use of AI in the workplace.
The OECD surveyed a total of 5 334 workers in the manufacturing and the finance and insurance sectors across Austria, Canada, France, Germany, Ireland, the United Kingdom, and the United States to examine how and why AI is being implemented in the workplace, its impact on job performance, working conditions, wages, management and skill needs, and the concerns and attitudes surrounding AI. The main fieldwork phase for the worker survey took place between mid-January and mid-February 2022.
The JILPT surveyed a total of 22 000 workers across all sectors in Japan. In the comparable manufacturing and finance and insurance sectors, responses were obtained from 5 440 workers. The questionnaire included several of the same questions as in the previous OECD survey, as well as original items addressing the effects of generative AI (GEAI) and the impact of AI on individuals with disabilities or those balancing childcare and/or long-term care responsibilities. The survey employed data from the 2020 national census to allocate the sample in a way that ensured various attributes aligned with the demographic distribution of the Japanese population. The main fieldwork phase for the worker survey took place between the end of May and the end of June 2024.
In the worker surveys, respondents were asked to keep the following definition in mind when answering the questions, regardless of their familiarity with AI: “Artificial intelligence – or 'AI' for short – is what enables smart computer programmes and machines to carry out tasks that would typically require human intelligence.” To reinforce this definition, respondents were also provided with examples of where AI might be found in everyday life and within their own sector. In the JILPT worker survey, GEAI was defined as: “The technology of Generative AI (e.g. ChatGPT, Bing, Bard) has attracted much attention in recent years. Generative artificial intelligence – or Generative AI for short – is a technology that uses artificial intelligence to generate text, images, video or audio. For example, when used to generate text, it can assist with drafting emails or documents, writing computer code, or summarising meeting minutes.” To ensure clarity, respondents in the JILPT survey were encouraged to carefully distinguish between questions about AI in general and those specifically related to GEAI. Additionally, respondents were required to click and confirm that they had reviewed the provided definitions and examples before proceeding to the main questions, thereby encouraging them to read these materials.
This report begins each section of the analysis by examining Japan’s position through international comparisons of the manufacturing sector and the finance and insurance sector. Building on these findings, it delves deeper into Japan’s unique challenges across all sectors, using microdata from the JILPT survey.
The following points should be considered when interpreting the findings in this report based on international comparative analysis.
Given global trends in AI development and adoption, the proportion of employees using AI at work in the countries surveyed in 2022 is likely to have increased further by 2024.
The definition of AI in the JILPT survey was based on that of the OECD’s previous survey but was expanded to include elements specific to GEAI. In contrast, the OECD survey was conducted before GEAI became widespread globally, following its emergence in November 2022.1 While the wording in both questionnaires is generally consistent, a significant technological innovation occurred in the real world between the two surveys. In the JILPT survey, GEAI users are more likely to report improvements in job performance and working conditions compared to AI users excluding GEAI. If similar patterns are observed in other countries, the impact of AI reported in the OECD survey across the seven countries may be understated relative to 2024, as it does not fully capture the effects of GEAI. However, determining the extent to which this factor influences the results of the international comparison in this report is difficult, as the pace and scope of GEAI adoption in workplaces likely vary across countries.
While such caution is needed in interpreting these results, the international comparative analysis aimed at understanding Japan’s overall position offers important insights that outweigh these limitations. Moreover, the results of the large‑scale Japanese worker survey – designed to closely reflect the structure of the population and offering unique insights into various aspects of AI adoption, including the use and effectiveness of GEAI will be of interest to other OECD countries.
In addition, Japan participated in two other surveys conducted by the OECD Secretariat to examine the impact of AI adoption in the workplace. The first is a survey of 6 047 mid-level managers across six countries (France, Germany, Italy, Japan, Spain, and the United States), which investigated the use of algorithmic management tools,2 including those powered by AI, during the period between June and August 2024 (Milanez, Lemmens and Ruggiu, 2025[6]). The second survey focussed on generative AI and was conducted at the end of 2024 among 5 232 small and medium-sized enterprises (SMEs) with fewer than 250 employees across seven countries (Austria, Canada, Germany, Ireland, Japan, Korea, and the United Kingdom) (OECD, 2025[7]). The survey targeted the individual within each company who had the most comprehensive understanding of the technologies used in the organisation. This report also introduces several insights drawn from these valuable studies.
1.2.1. AI users and non-AI users
In the OECD’s 2022 survey, the average proportion of employees whose companies use AI (AI adopters) across the seven countries covered was 60.4% in the finance and insurance sector and 44.1% in the manufacturing sector. The corresponding figure for Japan in 2024 was 26.5% in the finance and insurance sector and 14.5% in the manufacturing sector.
According to Figure 1.1, AI adopters who responded, “I have no interaction with AI at work” or “Don’t know”, as well as employees who answered that “Their companies didn’t use AI” or “Don’t know”, are classified as “AI non-users”, while all others are classified as “AI users”. Using this definition, in the OECD’s 2022 survey, the average proportion of AI users across the seven countries surveys was 42.4% in the finance and insurance sector and 29.3% in the manufacturing sector. The corresponding figures in Japan were 17.8% and 8.9%, respectively.
When asked about their interactions with AI at work, Japanese AI adopters are more likely than their counterparts in other countries surveyed to report being managed by AI3 (8.9% in the manufacturing sector in Japan compared to 5.9% in the other countries) or having no interaction with AI at work (34.9% vs. 30.7%), and less likely to report managing workers who work with AI (9.5% vs. 15.4%) (Figure 1.1). In particular, the situation in Japan, where many AI adopters reported having no interaction with AI at work, suggests that disparities in AI usage among employees within companies may be emerging.
When aggregated across all sectors, 8.4% of employees in Japan use AI at work (including GEAI), while 6.4% use just GEAI. Although direct comparisons are limited due to differences in survey timing, methodology, and target populations, the findings of the JILPT survey, which was carefully designed to closely reflect the demographic composition and employment characteristics of the Japanese workforce, suggest that the proportion of AI and GEAI users in Japan is broadly consistent with the results observed in previous major surveys conducted by other institutions in Japan.
A panel survey conducted by Keio University in January 2022, which included 2 139 participants, found that 5.1% of employees reported that AI had been introduced and was being used in their workplace.4 According to a survey conducted by INTAGE Inc. between October and November 2024 targeting 20 498 businesspersons aged 20 to 65, 11.6% of respondents reported that GEAI had been introduced into their own work (INTAGE Inc., 2025[8]). A survey conducted by the Recruit Works Institute in October 2023, involving around 10 000 participants, reported that 8.0% of Japanese workers were using GEAI (Recruitment Works Institute, 2024[9]). Similarly, the Tankan survey of workers, conducted by the JTUC Research Institute for the Advancement of Living Standards in October 2023, targeting 2 000 individuals aged 20 to 64 living in the Tokyo and Kansai metropolitan areas, found that 5.1% of Japanese employees were using GEAI frequently (Suzuki, 2024[10]). Although it is important to note that this reflects usage among the general public rather than just workers, a survey conducted by the MIC between January and February 2024, targeting 1 030 individuals, found that 9.1% of Japanese people, including those who had experience using GEAI in the past, had used GEAI (MIC, 2024[11]). According to a survey conducted by AI inside Inc. among 1 161 men and women aged 20 to 59 who are full-time employees, employers, or executives, 7.8% of respondents reported regularly using GEAI at work (AI inside Inc., 2023[12]).
Percentage of AI adopters
Note: AI adopters were asked: “Which of these statements best describes your interaction with AI at work? (I work with AI; I manage workers who work with AI; I develop/maintain AI; I am managed by AI; I interact with AI in another way; I have no interaction with AI at work) Respondents could select multiple answers.
Source: OECD worker survey on the impact of AI on the workplace (2022), JILPT worker survey on the impact of the introduction of AI into the workplace on working practices (2024).
Among the countries surveyed, the proportion of AI users is lowest in Japan (Figure 1.2). In the finance and insurance sector, the highest share of AI users is in the United States (53.5%). Similarly, in the manufacturing sector, Ireland leads with 42.9%. The difference in proportions between the United States and Japan in the finance and insurance sector is 35.7 percentage points. Similarly, in the manufacturing sector, the difference in proportions between Ireland and Japan is 34.1 percentage points. Given that the OECD's previous survey was conducted in 2022, the results from Japan's 2024 survey suggest that the use of AI in the workplace in Japan is likely to be even further behind than the figures suggest.
% of all employees, by country
Note: All employees were asked: "To the best of your knowledge, does your company use AI? (Yes; No; Don’t know)", AI Adopters were asked: "Which of these statements best describes your interaction with AI at work? (I work with AI; I manage workers who work with AI; I develop/maintain AI; I am managed by AI; I interact with AI in another way; I have no interaction with AI at work; Don’t know)” AI adopters who responded “I have no interaction with AI at work” or “Don’t know,” as well as employees who answered that “Their companies didn’t use AI” or “Don’t know,” are classified as “AI non-users,” while all others are classified as “AI users”.
Source: OECD worker survey on the impact of AI on the workplace (2022), JILPT worker survey on the impact of the introduction of AI into the workplace on working practices (2024).
The possibility that Japan lags behind other countries in the use of AI in the workplace is supported by other surveys. According to Milanez, Lemmens and Ruggiu (2025[6]), the percentage of managers whose companies provide at least some form of algorithmic management software was 40% in Japan – the lowest among the countries surveyed – compared to 90% in the United States, which reported the highest rate. Similarly, according to OECD (2025[7]) the percentage of SMEs in Japan that report using generative AI was 23.5%, the lowest among the surveyed countries, compared to 38.7% in Germany, which had the highest rate. Furthermore, according to a survey conducted by Veritas Technologies LLC. in December 2023 among 11 500 employees across 11 countries (Australia, Brazil, China, France, Germany, Japan, Singapore, South Korea, the United Arab Emirates, the United Kingdom, and the United States), the use of GEAI tools in Japanese workplaces ranked the lowest (Veritas Technologies LLC., 2024[13]). Although it is important to note that this reflects usage among the general public rather than just workers, a survey conducted by the MIC between January and February 2024 across five countries (China, Germany, Japan, the United Kingdom, and the United States) also found that Japan had the lowest use (MIC, 2024[11]). Considering these survey results together, it can be said that there is a high likelihood that AI usage in Japanese workplaces lags behind that of other countries.
One possible reason5 for the low proportion of AI use in Japan could be that Japanese companies have prioritised hiring non-regular workers to handle routine tasks, rather than investing in AI to automate them. Since the 1990s, with the spread of IT technologies in Japan, non-routine tasks previously performed by regular employees had increasingly been standardised into routine tasks. For example, the widespread adoption of personal computers and software for financial and accounting processes in companies had standardised company-specific, specialised tasks, potentially shifting labour demand towards workers capable of performing routine tasks using such tools. It has therefore been suggested that IT contributed to the increase in non-regular employment by standardising tasks previously performed by regular employees and expanding the volume of routine work (MHLW, 2001[14]). Subsequently, DeLaRica & Gortazar (2016[15]) pointed out that Japan has a relatively high proportion of routine tasks in the workplace compared to other countries. In addition, the share of non-regular workers among all employees has continued to rise, from 24.2% in 20006 to 34.8% in 2024. These suggest that the increase in routine tasks may have been addressed not through further investment in new technologies such as IT, but rather by hiring non-regular workers. One underlying factor is the relatively low labour cost of non-regular workers in Japan, including adjustment costs (Yamamoto, 2019[16]). Japan’s experience with IT so far appears to be applicable to AI as well. In addition, the limited experience in investing in IT capital may also have resulted in a shortage of human resources necessary to introduce AI into companies, along with challenges related to implementation costs and data preparation. The barriers to AI adoption are discussed in more detail later in this chapter.
Japan’s current lag in AI adoption in the workplace may translate into fewer negative impacts from AI than observed in other countries, although it also means Japan may not be reaping the benefits of AI to the same extent as other nations. However, it is notable that many Japanese employees believe that AI use will increase in the future (Figure 1.3). Specifically, 42.8% (58.1%) of AI users in Japan expect that the use of AI in workplace to increase considerably or even dramatically within the next two years (10 years). In addition, 51.3% (55.7%) of employees in Japan expect that the use of AI in workplace to increase within the next two years (10 years).
Percentage of all employees, by whether they and their company use AI
Note: All employees were asked: “To what extent do you think the use of AI in your workplace will increase in the next 2 years/ in the next 10 years?"
Source: JILPT worker survey on the impact of the introduction of AI into the workplace on working practices (2024).
1.2.2. Characteristics of companies that use AI
AI use and industry characteristics
There are significant disparities by sectors in the proportion of Japanese employees using AI at work (Figure 1.4). As a results, some sectors are less likely to face the negative impacts of AI, but may also find it more difficult to benefit from its advantages. The difference between the highest proportion in the Information and communications sector (22.9%) and the lowest in the Accommodations, eating, and drinking services sector (4.1%) is 18.8 percentage points. The econometric analysis reveals that, compared to workers in the Accommodations, eating and drinking services, those in Agriculture, forestry and fisheries, Manufacturing, Information and communications, Transport and postal services, Wholesale and retail trade, Finance and insurance, and Scientific research, professional and technical services are more likely to report using AI at work. In contrast, workers in the Medical, health care and welfare are less likely to report using AI in their jobs.7
% of all employees, by sector
Note: All employees were asked: "What sector (Japanese Standard Industrial Classification) do you work in? If your company runs more than one business, please answer based on the nature of main business." The figure shows the proportion of AI users among all employees in each sector. Some sectors with small samples are omitted.
Source: JILPT worker survey on the impact of the introduction of AI into the workplace on working practices (2024).
Information and communications is the sector with the highest proportion of AI adopters reporting that AI usage had increased since May 2022, at 67.5%. Conversely, the living-related and personal services and amusement services sector shows the lowest proportion, with 47.3% of AI adopters reporting that AI usage in their companies had expanded (Figure 1.5). While Japanese AI adopters across various sectors report that AI usage in their companies had expanded compared to two years earlier, the extent of these expansions also shows disparities between sectors.
% of AI adopters, by sector
Note: AI adopters were asked: "To the best of your knowledge, how has the use of AI in your company changed compared to two years ago (May 2022)?" Some sectors with small samples are omitted.
Source: JILPT worker survey on the impact of the introduction of AI into the workplace on working practices (2024).
AI use and company size
There are disparities in AI use by company size, similar to those in other countries surveyed (Figure 1.6). Specifically, in the finance sectors in Japan, there is a 17.1 percentage point difference in AI use between employees in companies with "500 or more workers" and those in companies with "up to 19 workers" compared to 12.1 percentage points in the other seven OECD countries. Similarly, in the manufacturing sector, the gap is 14.3 percentage points in Japan compared to 20.9 percentage points in the other OECD countries. In addition, A probit analysis controlling for several individual characteristics of Japanese workers reveals a clear trend: the likelihood of AI use in workplace increases with company size (Annex Table 1.A.1). Using workers employed in companies with up to 19 employees as the reference group, those working in companies with over 10 000 employees are 10.6 percentage points more likely to report using AI at work.
% of all employees, by company size
Note: All employees were asked: In the OECD survey, “How many persons work for your company, in the country where you are working? Please include both full-time and part-time staff.” In the JILPT survey, “How many persons work for your company in Japan? Please answer both full-time and part-time employees for your entire company, not just for your establishment.” The figure shows the proportion of AI users among all employees in each company size category.
Source: OECD worker survey on the impact of AI on the workplace (2022), JILPT worker survey on the impact of the introduction of AI into the workplace on working practices (2024).
AI use and labour shortages characteristics
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[17]). In Japan, the Employment Conditions Diffusion Index (D.I.),8 based on the perceptions of Japanese companies, reveals that, as of the June 2025 survey, the D.I. level reflecting labour shortages have surpassed pre‑COVID‑19 levels, with a continuous trend of worsening shortages (Bank of Japan, 2025[18]).
The Basic Policy on Economic and Fiscal Management and Reform, approved by the Japanese Government’s Cabinet in June 2025, commits to addressing labour shortages through the full utilisation of AI and digital technologies (Cabinet Office, 2025[19]). The economic measures approved by the Japanese Government’s Cabinet on 22 November 2024 indicate a direction9 toward accelerating the implementation of AI and robots as part of efforts to address labour shortages (Cabinet Office, 2024[20]). Additionally, the integrated innovation strategy 2025 approved by the Japanese Government’s Cabinet on 4 June 2024 indicates that with the growing labour shortage in Japan, there is an urgent need to improve productivity through automation and labour-saving through AI and robotics (Cabinet Office, 2025[21]). Many companies that have already implemented AI in their workplaces – or are considering doing so in the future – expect AI to help address labour shortages (Annex Figure 1.A.1, (JILPT, 2023[22])).
AI has the potential to alleviate labour shortages in many ways. AI robots are expected not only to serve as a part of the workforce themselves, but also to enhance the employability of older people and women by taking over physically demanding tasks. For example, in Japanese convenience stores, restocking heavy items such as beverages in cold environments is physically demanding work. As a result, there are efforts underway to pilot the use of AI-equipped shelving robots. If these robots can take over such tasks, it may encourage older people and women – who have tended to avoid convenience store jobs due to the physical burden – to enter or re‑enter the workforce.
In addition, AI can improve employees’ mental health and physical safety in the workplace, helping to reduce involuntary turnover and fostering a supportive environment where individuals can more easily continue working for longer periods. For example, the Japanese company DENSO CORPORATION has implemented a service called “Mente for Biz” in certain departments, provided by Medi Face Ltd. This service features an interactive mental health check-up where employees respond to questions from an AI doctor using a PC or smartphone. To enhance the accuracy of the assessment, the AI analyses non-verbal cues such as facial expressions, voice, and speech patterns captured through the device’s camera and microphone. According to the results shown in the check-up, qualified professionals such as industrial physicians conduct one‑on-one interviews with employees, providing comprehensive mental health support. Early detection and intervention for mental health disorders are expected to help prevent declines in company productivity and reduce the risk of employee turnover.10
Employees in Japanese companies experiencing moderate labour shortages make greater use of AI than those in companies with an appropriate level of staffing (Figure 1.7). Using workers who report that their company has an appropriate level of staffing as the reference group, those employed in companies experiencing mild labour shortages are 1.5 p.p. more likely to report using AI at work. However, no statistically significant effect was found for workers in companies facing severe labour shortages (Annex Table 1.A.1). It is not possible to determine causality, however. It may be the case that companies experiencing moderate shortages had previously experienced severe shortages, but that AI has helped them address these shortages. Alternatively, it could indicate that once companies face severe labour shortages, introducing AI becomes more challenging.
Percentage of all employees, by manpower status in company
Note: All employees were asked: “To the best of your knowledge, how would you rate the manpower status in your company, based on the situation of your workplace and relevant departments? (Severe labour shortage; Mild labour shortage; Neither shortage nor excess of employees (Appropriate); Mild labour excess; Severe labour excess; I don't know)”
Source: JILPT worker survey on the impact of the introduction of AI into the workplace on working practices (2024).
AI use and residential areas
The Basic Policy on Economic and Fiscal Management and Reform, approved by the Cabinet in June 2025, emphasises the need to accelerate the social implementation of emerging technologies – such as autonomous driving, drones, and AI – in rural areas in particular (Cabinet Office, 2025[19]). In addition, the economic measures approved by the Japanese Government’s Cabinet on 22 November 2024 indicate a direction11 towards accelerating the implementation of AI and robots as part of efforts to address social problems in rural areas (Cabinet Office, 2024[20]). In this context, it is desirable to have smaller regional disparities in the use of AI in the workplace. However, considerable disparities emerge in the proportion of Japanese employees using AI at work by residential area (Figure 1.8). The difference between the highest proportion of AI use in South Kanto (11.4%) and the lowest in Hokuriku (4.9%) is 6.5 p.p. At the more detailed prefectural level, Tokyo (13.8%) in the South Kanto region has the highest proportion of AI use, with Kanagawa (11.0%) in the same area also showing a relatively high proportion. On the other hand, Shimane (2.5%) in the Chugoku region has the lowest proportion of AI use, while Fukui (3.8%), Toyama (4.6%), and Ishikawa (4.9%) in the Hokuriku region also have a relatively low proportion. A probit analysis controlling for several individual characteristics of workers across all industries in Japan shows that, when using the Tokai region as the reference group, only the South Kanto region is associated with a significantly higher likelihood of AI use in the workplace (marginal effect: 10.9%, Z-value: 2.3).
Percentage of all employees, by residential area
Note: All employees were asked: “Which prefecture do you currently live in?” The 47 prefectures of Japan are divided into 10 blocks based on Japan’s regional classification in the OECD regional database. The figure shows the proportion of AI users among all employees in each region.
Source: JILPT worker survey on the impact of the introduction of AI into the workplace on working practices (2024).
1.2.3. Characteristics of employees who use AI
AI use by age, gender, and education
Japanese AI users in the finance and insurance and manufacturing sectors are more likely to be younger or middle‑aged, male, and better educated than non-users. These gaps between socio-demographic groups are consistent with the patterns observed in other OECD countries, although they appear to be less pronounced in Japan (Figure 1.9).
Age‑based disparities in AI usage are evident in both Japan and other countries, although they appear to be less pronounced in Japan. In the finance and insurance sector in Japan, the share of 15‑49 year‑olds is 9.9 p.p. higher among AI users than among AI non-users, against a difference of 17.7 p.p. in the other seven OECD countries on average. In the manufacturing sector, the respective figures are 16.6 and 22.3 p.p.
Gender-based disparities in AI usage are also evident in both Japan and other countries. While the gap is less pronounced in Japan’s finance and insurance sector than in the seven other OECD countries, it is slightly more pronounced in Japan’s manufacturing sector. In the finance and insurance sectors in Japan, the proportion of male AI users is 7 p.p. higher than that of male AI non-users, compared to 18.1 p.p. in other countries. The respective figures in the manufacturing sector are 6.3 and 5.4 p.p.
Finally, there are also disparities in AI use by educational attainment. Compared to the other seven OECD countries on average, the gap is less pronounced in Japan’s finance and insurance sector but more pronounced in Japan’s manufacturing sector. In the finance and insurance sector, the proportion of AI users with a university degree is 6.6 p.p. higher than among AI non-users in Japan, compared to a 29.2 p.p. gap internationally. In the manufacturing sector, the gaps are 26.4 and 20.3, respectively. A probit analysis controlling for several individual characteristics of workers across all industries in Japan confirms the results, except for gender (Annex Table 1.A.1). In the case of gender, the results change in favour of higher use for women once both industry and individual-level characteristics are controlled for.12
Percentage of AI users or AI non-users, by age, gender and education
Note: All employees were asked on education: In the OECD survey, “Have you completed at least a bachelor's degree or equivalent? (Yes; No; No answer)” In the JILPT survey, “Please answer the question about your last level of education.(Junior high school; High school; Vocational training school; Technical college or junior college; Four-year university; Graduate school; Other than above)” In Japan, the figure for “University degree” is the sum of four‑year university and graduate school.
Source: OECD worker survey on the impact of AI on the workplace (2022), JILPT worker survey on the impact of the introduction of AI into the workplace on working practices (2024).
AI use by employment status, annual income, weekly working hours
Japanese AI users – particularly GEAI users – are more likely to be regular employees than both AI non-users and AI non-adopters (Figure 1.10). While Japanese legislation does not explicitly define “regular” and “non-regular” employees, the Labour Force Survey in Japan classifies employees based on their employment status. Specifically, employees excluding executives are classified into seven categories: “Regular employee”, “Part-time worker”, “Arbeit (temporary worker)”, “Dispatched worker from temporary labour agency”, “Contract employee”, “Entrusted employee” and “Others”. Non-regular employees are defined as those in the six categories excluding regular employees. In other words, non-regular employees are generally characterised as those who do not meet one or more of the following conditions: open-ended employment, full-time work, and direct employment. By contrast, regular employees satisfy all three conditions (Asao, 2011[23]). Following this definition, the proportion of non-regular employees was around 14.8% among AI users, 27.9% among AI non-users, and 33.7% among AI non-adopters. Among GEAI users, non-regular employees made up only 12.2%, while regular employees accounted for 87.8%.
A probit analysis controlling for several individual characteristics of Japanese workers shows that the likelihood of AI use in the workplace is lower among non-regular workers, and higher among those with higher incomes and those working longer hours (Annex Table 1.A.1). That said, using workers with an average weekly working time (including overtime) of 34 hours or less as the reference group, no statistically significant difference was observed in AI use among those working 60 hours or more per week. This may suggest that excessive work demands reduce workers’ capacity to engage with AI use.
Percentage of all employees, by employment status
Note: All employees were asked: “What is your current employment status called at your workplace? (Executive of company; Regular employee; Non-regular employee (e.g. Part-time worker, Dispatched worker, Contract worker))”.
Source: JILPT worker survey on the impact of the introduction of AI into the workplace on working practices (2024).
AI use by disability status or caregiving responsibility status
Employees with disabilities and those balancing childcare and/or long-term care responsibilities are more likely to say they use AI at work compared to their counterparts without disabilities or caregiving responsibilities (Figure 1.11). AI use at work among people with disabilities is 16.2 p.p. higher than among people without disabilities. However, it is important to interpret these results carefully, as only a selection of employees with disabilities may have participated in the web survey conducted by JILPT. Although the survey could be completed using either a smartphone or a PC, differences in PC usage likely vary depending on the type of disability (IICP, 2012[24]). Consequently, careful consideration is necessary to assess whether the findings can be generalised to all employees with disabilities. This caveat applies to all analyses of employees with disabilities presented in this report.
That being said, two surveys conducted in 2018 and 2022 found that around 70% to 80% of companies employing people with disabilities had introduced digital technologies into their work processes (NRI and NRI Mirai., 2018[25]; JEED, 2024[26]). Moreover, the COVID‑19 pandemic likely accelerated the adoption of digital technologies in the workplace for many workers, including those with disabilities (JILPT, 2024[27]; JILPT, 2023[28]; MHLW, 2021[29]). These findings suggest a growing trend in AI use by individuals with disabilities (Box 1.1), particularly focussing on GEAI. Indeed, with appropriate management, AI has the potential to create a more inclusive and accommodating environment for workers with disabilities, helping to remove barriers they face in the labour market (Touzet, 2023[30]).
Percentage of all employees, by disability status, childcare status, long-term care status
Note: All employees were asked: “Do you currently have Disability certificates? 'Disability certificates' here refer to the Physical disability certificate, the Mental disability certificate, the Rehabilitation certificate, etc.” Employees with disabilities are those who answer “Yes, I have” and “No, I don’t, but I have been diagnosed as disabled’. Employees with children and who don’t want to answer about whether they have children were asked: “Are you currently engaged in caring for your own child?” All employees were asked: “Are you currently engaged in long-term care for your family member?”
Source: JILPT worker survey on the impact of the introduction of AI into the workplace on working practices (2024).
Building on interviews with more than 70 stakeholders, Touzet (2023[30]) explores the potential of AI to foster employment for people with disability, accounting for both the transformative possibilities of AI-powered solutions and the risks attached to the increased use of AI for people with disability. It also identifies obstacles hindering the use of AI and discusses what governments could do to avoid the risks and seize the opportunities of using AI to support people with disability in the labour market.
This report identifies 142 examples of AI-powered solutions that could support people with disabilities in the labour market. Over 75% of these would not exist without AI. This box highlights some of the initiatives, particularly those undertaken by the companies that participated in the study.
Next Generation Mobility Assistive Technology Development Consortium (AI-powered suitcase)
This Japanese consortium is developing the AI-powered suitcase. That is an autonomous navigation robot designed to assist visually impaired individuals with mobility. Resembling a standard suitcase, it is equipped with built-in computers, sensors, and motors, enabling it to safely guide users to their destination while avoiding obstacles and other pedestrians. As of the time of writing this report, the AI-powered suitcase has not yet been commercialised but has been undergoing continuous trial operations since April 2024. If commercialised, it could significantly enhance the ease and safety of work-related and everyday mobility for visually impaired individuals.
ULU-RU BPO (eas next)
This Japanese company offers a comprehensive service aimed at promoting the employment of people with disabilities. As part of this, they provide the AI-OCR service “eas”, which enhances job performance by enabling highly accurate data conversion through the collaboration of AI-OCR and human effort.
As other examples, SureTalk (developed by the Japanese company SoftBank Corp.), which facilitates communication by using AI to analyse images of hand sign language movements and convert them into text, is used by some local authorities. Eye Navi (developed by the Japanese company Computer Science Institute Co., Ltd.) is an app that uses a smartphone’s camera function and AI to provide navigation assistance for the visually impaired. CoeFont CLOUD (developed by the Japanese company CoeFont Co., Ltd.) uses AI to learn a person’s voice in advance. This allows individuals with speech impairments to communicate using their natural voice through smartphone text input, even after they lose the ability to speak.
Envision (Envision Glasses)
This Dutch company has developed Envision Glasses, building on Google Glasses as hardware support. Together with a companion phone‑based application, Envision Glasses allow blind users to read typed and handwritten text, recognise humans in front of them, or hear descriptions of objects and landscapes in their vicinity.
Zammo.ai (AI-powered chatbots)
The chatbot developed by the American company Zammo.ai is designed to make job boards, such as LinkedIn or Indeed, accessible for people with vision impairments and neurodiverse individuals. Zammo’s chatbot is designed to sit on top of traditional job boards and provides an alternative and accessible interface for users, who can access information about job postings in a conversational manner, via text or speech queries and answers.
Similarly, the proportion of employees balancing childcare responsibilities using AI at work is 7.2 p.p. higher than those without such caregiving responsibilities. The proportion of employees balancing long-term care responsibilities using AI at work is 16.5 p.p. higher than those without such caregiving responsibilities. Employees with caregiving responsibilities may be increasingly turning to AI in their work to enhance efficiency and better balance their work and caregiving responsibilities.
AI use and workers’ main job tasks
The use of AI is more common among Japanese workers who do cognitive tasks than among workers who do routine and physical tasks (Figure 1.12). Japanese employees who spend “over half” or “almost all” of their working hours on tasks such as “Analysing data and information to make decisions based on the results” are among the most frequent AI users. Similarly, the proportion of Japanese AI users is higher among employees engaged in “Tasks identifying and solving problems using creativity”, “Managing and motivating team members or subordinates”, or “Tasks in hazardous places (Box 1.2)”. Conversely, the proportion of Japanese AI users is lower among employees engaged in “Routine and repetitive tasks” or “Physical tasks”.
These findings contrast with previous technologies such as computers and robots, that have traditionally automated routine tasks and displaced low- and medium-skilled workers (Autor, Levy and Murnane, 2003[31]). Some identify AI as a force to enable the automation of non-routine and/or high-skilled tasks, because of the specific capabilities of AI (Aghion, P. et al., 2017[32]). Additionally, the occupations judged to be most exposed to AI include high-skilled occupations – including some traditionally “white‑collar professions” requiring non-routine cognitive tasks (Lane and Saint-Martin, 2021[33]). AI has made the most progress in its ability to perform non-routine, cognitive tasks, such as information ordering, memorisation and perceptual speed (OECD, 2023[34]).
Percentage of all employees, by tasks in typical working day
Note: All employees were asked: “How much time do you spend on the following tasks in a typical working day? (Almost always during working hours; Over half of working hours; Almost half of working hours; Under half of working hours; Nothing at all)” This figure shows the proportion of workers who answered that they spend “over half of working hours” or “almost all of working hours” for each task in a typical working day. “Physical tasks” include activities such as standing work, operating machinery or vehicles, and carrying loads. “Tasks in hazardous place” include activities at high places, in extremely hot or cold places, in places with a lot of machinery.
Source: JILPT worker survey on the impact of the introduction of AI into the workplace on working practices (2024).
According to JILPT (2025[5]), employees who spend “over half” or “almost all” of their working hours on tasks in hazardous places are most prevalent in the manufacturing sector (29.0%). The proportion is also relatively high in the Construction sector (10.5%) and the Transport and postal services sector (9.4%). Statistics from the MHLW indicate that these sectors also have high rates of occupational accidents. This box highlights examples of AI technologies used for tasks in hazardous places.
VRAIN Solution
The Japanese company VRAIN Solution provides AI services tailored to the manufacturing sector. AI can monitor workers’ posture, safety harness usage, and behaviour, and if it detects a hazard, it will trigger an alarm for the employee and notify the supervisor. Additionally, the AI can perform an emergency stop of an industrial robot if it detects a person entering the robot’s operating area.
Obayashi Corporation, JOHAS, and MetaMoJi (Safety AI Solution)
The Japanese company MetaMoJi, in collaboration with JOHAS, developed the Safety AI Solution, a system that leverages AI to conduct risk assessments based on past disaster data and can identify potential hazards in various operations. This system is currently used by Obayashi Corporation in the construction sector. The Safety AI Solution can predict hazards based on workplace conditions, task details, equipment used, and worker information. These potential risks, along with recommended countermeasures, are integrated into work schedule sheets. The system also can provide access to detailed case‑specific information and personalise safety management insights based on factors such as the worker’s age. By reviewing this information before working, potential risks can be visualised, leading to more advanced and proactive safety management. This approach is particularly beneficial for younger, less experienced employees, as it facilitates the visualisation and sharing of veteran workers’ expertise and tacit knowledge.
Yazaki Corporation (TRUE SAFE)
The Japanese company Yazaki Corporation provides a system aimed at reducing vehicle accidents by leveraging AI to analyse real-time data from sensors installed in transport company vehicles. The AI can monitor driver behaviour, alert drivers about dangerous driving, and evaluate their overall driving performance, facilitating proactive safety management. The system also incorporates specialised AI diagnostic features specifically designed to address the high incidence of accidents during vehicle reversing. In collaboration with user companies, Yazaki Corporation has reported that implementing the system led to a verified reduction in accidents per operation compared to the same period in the previous year.
Suntory Logistics Ltd. and Fujitsu Limited
The Japanese companies Suntory Logistics Ltd. and Fujitsu Limited have jointly developed and implemented a drive recorder capable of capturing 360° images of forklift operations, which are then analysed and evaluated by AI. Forklift operations are central to cargo handling in warehouses and are often linked to occupational accidents. This system enhances efficiency by optimising forklift operations while improving the safety of employees involved in them.
AI use and occupation
Differences in AI use by task translate into differences in AI use by occupation. “Managers” are among the most frequent AI users, as well as “Professionals” and “Technicians and associate professionals”. By contrast, employees in “Plant and machine operators, and assemblers” and “Elementary occupations” are less likely to use AI at work. The difference between the occupations with the highest and lowest use is 13.5 p.p. (Figure 1.13, Annex Table 1.A.1).
A more detailed occupational classification using ISCO’s DIGIT3 reveals that “Information and communications technology service managers” have the highest proportion of AI users (38.0%). Other occupations with relatively high proportions of AI users include “Administration professionals” (36.1%), “Finance professionals” (30.6%), “Database and network professionals” (27.0%), “Software and applications developers and analysts” (27.0%). Conversely, “Domestic, hotel and office cleaners and helpers” have the lowest proportion of AI users (0.8%). Other occupations with relatively lower proportions of AI users include “Personal care workers in health services” (1.2%), “Manufacturing labourers” (1.6%), “Mining and construction labourers” (2.1%), “Other craft and related workers” (2.3%), “Other elementary workers” (2.3%), “Metal processing and finishing plant operators” (2.5%), “Agricultural, forestry and fishery labourers” (2.7%), “Heavy truck and bus drivers” (2.7%), and “Transport and storage labourers” (3.4%).
There are also measures of occupational AI exposure, which typically assess the overlap between the abilities required in an occupation and the technical capabilities of AI (Felten, Raj and Seamans, 2021[35]). According to analyses using this measure, the occupations most exposed to AI are white‑collar occupations involving non-routine cognitive tasks, such as “IT professionals”, “Business professionals”, “Managers”, and “Science and engineering professionals”. Occupations requiring manual skills and physical strength, such as “Cleaners”, “Agricultural, forestry and fishery labourers”, “Food preparation assistants”, and “Labourers”, are the least exposed to AI (Georgieff and Hyee, 2021[36]; Lane, 2024[37]). These results are based on data from 22 OECD countries (excluding Japan), linked to the Survey of Adult Skills (PIAAC), which allows the indicator to vary at the country-occupation level.
Although the results based on occupational AI exposure measures and those from the JILPT survey are not directly comparable due to differences in timing and statistical methods, the characteristics of occupations with high and low AI usage (or exposure) in Japan and the 22 OECD countries show many similarities. Unlike traditional technologies such as computers and robots, which primarily affected blue‑collar jobs involving routine tasks, AI has the potential to influence white‑collar and high-skilled jobs. In other words, AI has the potential not only to improve the job quality of workers in white‑collar and high-skilled occupations, but also to alter their job quantity and task composition, including through increased employment opportunities and a shift towards more complex tasks.
Percentage of all employees, by occupation
Note: All employees were asked: “Which of the following categories (ISCO) best describe your job in current company?” Occupations with fewer than 50 samples were omitted.
Source: JILPT worker survey on the impact of the introduction of AI into the workplace on working practices (2024).
Changes in AI use by occupation
Workers in high-skilled occupations are also the ones who report the greatest expansion in AI use between 2022 and 2024. “Managers”, “Professionals”, and “Technicians and associate professionals” among AI adopters are the most likely to report these changes, whereas those in “Elementary occupations” are less likely to do so (Figure 1.14). The difference in the share of AI adopters reporting an expansion in workplace AI use over the past two years between the occupation with the greatest increase, “Managers”, and that with the smallest increase, “Elementary occupations”, is 30.4 p.p. This suggest that disparities in AI use between high- and low-skilled occupations may be widening. Therefore, to promote a more inclusive and equitable digital transformation,13 efforts should also aim to address occupational differences in AI use.
A more detailed occupational classification using ISCO’s DIGIT3 reveals that “Finance professionals” have the highest proportion of AI adopters reporting an expansion in AI usage within their companies (86.7%). Other occupations with relatively high levels of expansion in AI usage within their companies include “Business services and administration managers” (79.1%), “Process control technicians” (76.0%), “Physical and earth science professionals” (75.0%), and “Engineering professionals (excluding electrotechnology)” (74.4%).
Percentage of AI adopters, by occupation
Note: All employees were asked: “Which of the following categories (ISCO) best describe your job in current company?” AI adopters were asked: “To the best of your knowledge, how has the use of AI in your company changed compared to two years ago (May 2022)?” Occupations with fewer than 20 samples were omitted.
Source: JILPT worker survey on the impact of the introduction of AI into the workplace on working practices (2024).
1.3.1. Perceptions of companies
The economic measures approved by the Japanese Government’s Cabinet on 22 November 2024 indicate a direction14 toward accelerating initiatives to develop human capital for digitalisation (Cabinet Office, 2024[20]). Furthermore, the comprehensive strategy for the Vision for a Digital Garden City Nation (DIGIDEN, 2023 Revision), approved by the Japanese Government’s Cabinet on 26 December 2023, outlined plans to develop a total of 2.3 million “digital promotion personnel” between FY2 022 and FY2 026 (Cabinet secretariat, 2023[38]). These digital professionals are envisioned to include business architects, data scientists, software engineers, cybersecurity specialists, and UI/UX designers, along with experts possessing specialised digital knowledge and skills essential for implementing digital solutions across various regions and sectors. AI-related talent is included as part of the human resources for digital promotion. Under this strategy, the relevant ministries – MHLW, MEXT, and METI – are expected to collaborate systematically to advance the initiative as a unified government effort.
Compared to their counterparts in the United States, Japanese companies are struggling more with a shortage of AI-related talent – a problem which has become more pronounced between 2022 and 2023, and continues to hinder AI adoption (Figure 1.15). In 2023, “Shortage of AI-related talent” was identified as the most critical issue in AI adoption by Japanese companies (62.4%, up 12.7 p.p. since 2022). Other frequently cited issues included “AI is not well understood within the company” (47.0%), “We are concerned about whether we can realise the benefits of the AI implementation” (41.9%, up 13.1 p.p. since 2022). In particular, a higher proportion of Japanese companies than their U.S. counterparts cited concerns about a shortage of AI-related talent and uncertainty over whether the benefits of AI implementation can be realised.
Across all data points, the shortage of AI-related talent stands out as a major issue. As shown in Figure 1.2, the share of workers using AI at work is lower in Japan than in the seven other OECD countries surveyed – a pattern that may be partly explained by this shortage of personnel capable of driving AI adoption within Japanese companies. Other challenges can be broadly categorised as a lack of information about the benefits of AI implementation, insufficient examples from other companies, difficulties related to the data required for AI training, cost concerns, and limitations in the range of available services.
The present OECD report may contribute to addressing the lack of information about the benefits of AI implementation in the workplace. In addition, while the Japanese Government has already compiled and shared cases of municipalities and companies that have enhanced productivity and operations through AI implementation (Cabinet Office, 2018[39]; MIC, 2022[40]; MIC, 2024[41]; The Conference toward AI Network Society, 2022[42]), it would be beneficial to also collect and share cases highlighting strategies and innovations in the process of introducing AI into the workplace, as well as examples of how various job quality improvements have been achieved. Making these cases easily accessible to diverse types of companies could offer valuable guidance for those considering AI implementation in the future. Furthermore, METI provides several subsidies to support the costs of AI introduction for SMEs, while MHLW offers grants to SMEs that aim to raise employees’ minimum wages alongside AI adoption. As highlighted in this chapter, the proportion of AI users in Japanese SMEs lags behind that of larger companies. However, the severe labour shortages faced by SMEs may encourage more of them to consider adopting AI in the workplace in the future. To address these needs, Japan should regularly review the use of subsidies and grants and evaluate whether revisions are necessary to ensure that SMEs and employees can fully benefit from AI adoption.
Percentage of companies excluding those that don’t intend to adopt AI in the future
Note: The results of the survey on the challenges faced when introducing AI. The survey was conducted from 9 February 2024 to 2 May 2024, collecting a total of 1 013 responses from 7 564 companies. Some items with low response rates have been omitted. Respondents could select multiple answers.
Source: Information-technology Promotion Agency, Japan (IPA), DX Trend 2024.
A more detailed analysis of AI-related talent shortages reveals that, in 2022, Japanese companies were notably more likely than their U.S. counterparts to report shortages of “Employees who can promote AI adoption in the company, using their workplace experience and basic AI knowledge”. Moreover, between 2022 and 2023, the proportion of Japanese companies citing a shortage of such employees increased significantly, underscoring the particularly high demand for this type of talent (Figure 1.16). There is also strong demand for “Employees who can design AI-driven products and services” and “Employees who can analyse data using AI tools and apply it to their company’s business”. This indicates that Japanese companies are increasingly seeking employees who can translate AI capabilities into tangible business value through their day-to-day tasks. Other areas where demand in Japanese companies surpasses that in U.S. companies include managers with a strong understanding of AI and AI developers. In the future, as barriers to AI introduction continue to diminish and implementation progresses, the demand for such personnel in Japanese companies is expected to grow further. There is also strong demand for “Employees who can design AI-driven products and services” and “Employees who can analyse data using AI tools and apply it to their company’s business”. This indicates that Japanese companies are increasingly seeking employees who can translate AI capabilities into tangible business value through their day-to-day tasks. Other areas where demand in Japanese companies surpasses that in U.S. companies include managers with a strong understanding of AI and AI developers. These findings indicate that the “shortage of AI‑related talent” – a key barrier to workplace AI adoption – refers not only to a lack of AI developers, but also to the insufficient skills among existing employees to effectively adopt and use AI technologies. OECD (2025[7]) also reveals that, in contrast to trends in other countries, Japan stands out among SMEs as the country where the lack of the right skills among employees is the most significant barrier to the use of GEAI in the workplace (Annex Figure 1.A.2). The importance of companies providing training and financial support to help employees work with AI, as well as supporting their reskilling and upskilling efforts, will be discussed in greater detail in Chapters 2 and beyond.
Percentage of companies excluding those that don’t intend to adopt AI in the future
Note: The results of the survey regarding the adequacy of AI-related human resources. The survey was conducted from 9 February 2024 to 2 May 2024, collecting a total of 1 013 responses from 7 564 companies. “No shortage” combines responses indicating “there are enough talents” and “there are enough talents to some extent”.
Source: Information-technology Promotion Agency, Japan (IPA), DX Trend 2024.
1.3.2. Why do people not use GEAI?
Unless companies prohibit the use of GEAI in the workplace, the decision to use it will remain at the discretion of individual employees. Within this scope of personal discretion, why might employees choose not to use GEAI? To explore this question, this section examines the reasons for not using GEAI, drawing on findings from three different surveys that reflect varying levels of awareness, experiences, and perspectives among respondents.
First, it is important to note that the data referenced here reflects reasons cited by the general public, rather than workers specifically; however, this data provides the advantage of enabling international comparisons. The survey allowed multiple responses, enabling respondents to select more than one reason for not using GEAI. According to the findings, a higher proportion of Japanese respondents, compared to those in the four other countries (China, Germany, the United Kingdom, and the United States), cited “I don’t know how to use it” (41.2%) and “I don’t need it in my life” (39.9%) as reasons for not using GEAI. Conversely, “Concerns about information leaks, security, and safety” (2.4%) were cited least frequently by Japanese respondents and less often than in the other four countries (Figure 1.17). From the perspective of the general public, Japan may have greater potential for increased GEAI usage compared to other countries, provided there is progress in both understanding how to use these tools safely and how they can be useful. Targeted awareness-raising and improved usability could help unlock this latent demand.
Percentage of respondents
Note: The results of the survey on the reasons for not using GEAI. The survey was conducted from January to February 2024. The four-country average includes data from China, Germany, the United Kingdom, and the United States and is calculated as a simple average of the figures from each country. Respondents could select multiple answers.
Source: MIC (2024[11]), Information and Communication Technology White Paper, https://www.soumu.go.jp/johotsusintokei/whitepaper/ja/r06/pdf/00zentai.pdf.
Second, regarding the use of GEAI in the Japanese workplace, while there are similarities to the general public’s views, there are also more cautious opinions. According to a survey conducted by AI inside Inc. among 1 161 Japanese men and women aged 20 to 59 who are full-time employees, employers, or executives, “No specific reasons” (27.6%) and “I don’t have an idea of how to use it” (21.6%) ranked high amongst the reasons given by the 464 GEAI non-users (Figure 1.18), indicating that if effective methods for utilising GEAI in the workplace were to be established, GEAI usage in Japan might increase. On the other hand, “Concerns about the accuracy of the answers’ content” (28.7%) was the most frequently cited issue among non-users, with “Concerns about insufficient legal framework” (20.5%), “Concerns about ethical issues, such as bias in AI” (19.4%), and “Concerns about security” (19.0%) also ranking high.
Finally, individuals involved in the introduction of GEAI in the Japanese workplace have expressed greater concerns about the accuracy, safety, and reliability of GEAI technology, with more specialised issues also being raised. A multiple‑response survey targeting 218 individuals employed at large Japanese companies who have been involved in introducing GEAI in the workplace revealed that 184 respondents expressed concerns about using GEAI. The most commonly cited technical issue was “The generation of inaccurate information” (59.2%), followed by “Risks related to security, such as the leakage of confidential information” (54.9%), “Lack of consistency in response quality” (54.3%), “Uneven accuracy and quality of responses” (48.9%), “The opacity of the reasoning and processes behind AI-generated outputs” (44.0%), “The difficulty of integration with existing internal systems” (35.3%), and “Lack of capability to handle the latest or specialised information” (34.2%) (AI inside Inc., 2024[43]). Those involved in the introduction of GEAI have provided deeper insights into its accuracy, safety and reliability of GEAI, highlighting challenges such as the difficulty of integrating with existing systems and the lack of capacity to handle the latest or specialised information. Furthermore, from the perspective of SMEs, there is additional evidence that concerns about copyright, legal, or regulatory issues associated with GEAI may be more pronounced in Japan than in other countries as a reason for not using it in the workplace (Annex Figure 1.A.2).
Percentage of full-time employees, employers, or executives
Note: The results of the survey on the reasons for not using GEAI. This survey, conducted in July 2023, involved 1 161 men and women aged 20 to 59 who are full-time employees, employers, and executives. This refers to the results of a survey conducted among 464 GEAI non-users, asking about the issues they perceive in using GEAI in the workplace. Respondents could select multiple answers.
Source: AI inside Inc. (2023[12]), Survey to identify the actual use of and intention to use generative AI, https://inside.ai/news/2023/09/26/ai-researchresults-4/.
In practice, the most common use of GEAI in Japan is “Drafting documents or texts”, followed by “Organising and analysing data”, “Revising, editing, proofreading, and summarising texts”, “Translation into other languages”, “Coding or debugging code”, and “Testing and refining ideas” (Annex Figure 1.A.3). Although it is essential to avoid relying entirely on GEAI outputs in the workplace without human judgement, current usage patterns in Japan highlight how GEAI is being used to assist with various human tasks and support the refinement of ideas. In this context, collecting and widely sharing concrete success stories of how GEAI has been effectively utilised in the workplace could play a key role in improving access to the benefits of AI for a broader range of workers across Japan. On the other hand, when using GEAI in the workplace for job-related purposes, a number of concerns have been raised regarding the accuracy, safety, and reliability of the technology. Successful examples of GEAI adoption in the workplace are likely characterised by the careful management of such risks as a foundational element. In particular, from the broader perspective of AI technologies in general, building trust that only safe and trustworthy AI is used in the workplace may significantly influence the extent to which employees can benefit from its use. This issue will be explored further in Chapter 2, which focusses on the impact of AI adoption in the workplace on job quality.
Percentage of companies excluding those that have not adopted AI-based digital technology and have no plans to do so in the future
Note: Companies excluding those that have not adopted AI-based technology and have no plans to do so in the future: “Please select the option that best describes your company’s view on the future (or current) benefits of introducing (or having already introduced) AI-based technology. AI addresses labour shortages (I agree; I somewhat agree; I somewhat disagree; I disagree)” The figure shows the proportion of companies that answered “agree” or “somewhat agree”.
Source: JILPT panel survey on investment in human resources and corporate strategy (2022).
Marginal effects after probit regression
Note: *** Significant at the 1% level, ** 5% level, * 10% level.
Source: JILPT worker survey on the impact of the introduction of AI into the workplace on working practices (2024).
Percentage of SMEs that report each barrier to use
Note: Non-users of generative AI were asked: “I will read you a few possible reasons why a company might not use generative AI. Can you tell me whether you agree or disagree with these reasons?”.
Source: OECD survey on how SMEs are using generative AI to address skill and labour needs, 2024.
Percentage of GEAI users, by age
Note: GEAI users were asked: “How do you use generative AI in your current work?” Respondents could select multiple answers.
Source: JILPT worker survey on the impact of the introduction of AI into the workplace on working practices (2024).
References
[32] Aghion, P. et al. (2017), “Artificial Intelligence and Economic Growth”, NBER Working Paper, Vol. No. 23928, https://www.nber.org/papers/w23928.
[43] AI inside Inc. (2024), Fact-finding survey on the business adaptation and challenges of generative AI, https://inside.ai/news/2024/10/17/gen-ai-hallucination-issue/ (accessed on 13 December 2024).
[12] AI inside Inc. (2023), Survey to identify the actual use of and intention to use generative AI, https://inside.ai/news/2023/09/26/ai-researchresults-4/ (accessed on 16 December 2024).
[23] Asao, Y. (2011), Overview of non-regular employment in Japan, https://www.jil.go.jp/english/reports/documents/jilpt-reports/no.10_japan.pdf (accessed on 2 December 2024).
[31] Autor, D., F. Levy and R. Murnane (2003), “The Skill Content of Recent Technological Change: An Empirical Exploration”, The Quarterly Journal of Economics, https://doi.org/10.1162/003355303322552801.
[18] Bank of Japan (2025), Tankan (Short-Term Economic Survey of Enterprises in Japan), https://www.boj.or.jp/en/statistics/tk/gaiyo/2021/tka2409.pdf.
[3] Bresnahan, T. and M. Trajtenberg (1992), “General Purpose Technologies “Engines of Growth?””, https://doi.org/10.3386/W4148.
[2] 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, https://doi.org/10.3386/w24001 (accessed on 25 July 2024).
[19] Cabinet Office (2025), Basic Policy on Economic and Fiscal Management and Reform 2025: Toward a Society Where People Can Truly Feel That “Tomorrow Will Be Better Than Today”, https://www5.cao.go.jp/keizai-shimon/kaigi/cabinet/honebuto/2025/2025_basicpolicies_ja.pdf.
[21] Cabinet Office (2025), Integrated Innovation Strategy 2025, https://www8.cao.go.jp/cstp/tougosenryaku/togo2025_zentai.pdf (accessed on 21 July 2025).
[20] Cabinet Office (2024), Comprehensive economic measures for the security and safety of the people and sustainable growth - Increasing current and future wages and incomes for all generations., https://www5.cao.go.jp/keizai1/keizaitaisaku/2024/1122_taisaku.pdf.
[39] Cabinet Office (2018), List of advanced and good practices for operational reforms using AI, ICT, etc., https://www5.cao.go.jp/keizai-shimon/kaigi/special/reform/koukyou/06_aiict/index.html (accessed on 22 January 2025).
[38] Cabinet secretariat (2023), Comprehensive Strategy for the Vision for a Digital Garden City Nation (DIGIDEN) (2023 Revision), https://www.cas.go.jp/jp/seisaku/digital_denen/pdf/20231226honbun.pdf.
[15] DeLaRica & Gortazar (2016), Differences in Job De-Routinization in OECD Countries: Evidence from PIAAC, https://docs.iza.org/dp9736.pdf.
[35] Felten, E., M. Raj and R. Seamans (2021), “Occupational, industry, and geographic exposure to artificial intelligence: A novel dataset and its potential uses”, Strategic Management Journal, Vol. 42/12, pp. 2195-2217, https://doi.org/10.1002/SMJ.3286.
[36] Georgieff, A. and R. Hyee (2021), “Artificial intelligence and employment: New cross-country evidence”, OECD Social, Employment and Migration Working Papers, https://doi.org/10.1787/c2c1d276-en.
[24] IICP (2012), Research and study on the use of the internet and other services by people with disabilities, https://www.soumu.go.jp/iicp/chousakenkyu/data/research/survey/telecom/2012/disabilities2012.pdf.
[8] INTAGE Inc. (2025), Intage conducts composite survey on the actual use of generative AI, https://www.intage.co.jp/news/5169/ (accessed on 2 April 2025).
[26] JEED (2024), Research on changes in job areas for people with disabilities due to technological developments such as AI, etc., https://www.nivr.jeed.go.jp/research/report/houkoku/absstu0000000o8d-att/houkoku177.pdf (accessed on 10 February 2025).
[5] JILPT (2025), Worker survey on the impact of the introduction of AI into the workplace on working practices, JILPT research series, https://www.jil.go.jp/institute/research/2025/documents/0256.pdf (accessed on 4 June 2025).
[27] JILPT (2024), Survey on the Impact of COVID-19 on Corporate Management (1st–6th) – JILPT continuous panel survey of Companies on COVID-19 –, JILPT research series No. 237, https://www.jil.go.jp/institute/research/2024/documents/0237.pdf.
[22] JILPT (2023), Panel survey on investment in human resources and corporate strategies (JILPT Corporate Panel Survey), JILPT research series No.232, https://www.jil.go.jp/institute/research/2023/documents/0232.pdf.
[28] JILPT (2023), Survey on the use of digital technology in the manufacturing sector and the securing and development of human resources, JILPT Research series No.233, https://www.jil.go.jp/institute/research/2023/documents/0233.pdf.
[37] Lane, M. (2024), “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”, OECD Artificial Intelligence Papers, No. 26, OECD Publishing, Paris, https://doi.org/10.1787/14dc6f89-en.
[33] Lane, M. and A. Saint-Martin (2021), “The impact of Artificial Intelligence on the labour market: What do we know so far?”, OECD Social, Employment and Migration Working Papers, No. No. 256, OECD Publishing, Paris, https://doi.org/10.1787/7c895724-en.
[4] 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, https://doi.org/10.1787/ea0a0fe1-en.
[29] MHLW (2021), Material (Discussion Paper for future consideration) from the 110th meeting of the Employment of Persons with Disabilities Subcommittee of the Labour Policy Council, https://www.mhlw.go.jp/content/11704000/000841111.pdf.
[14] MHLW (2001), The White Paper on Labour Economy: Innovation in Information and Technology (IT) and Employment.
[41] MIC (2024), Guidebook for the use and introduction of AI in municipalities <Separate Appendix> List of cases of generative AI implementation in leading organisations, https://www.soumu.go.jp/main_content/000956981.pdf (accessed on 22 January 2025).
[11] MIC (2024), Information and Communication Technology White Paper, https://www.soumu.go.jp/johotsusintokei/whitepaper/ja/r06/pdf/00zentai.pdf.
[40] MIC (2022), Guidebook for the use and introduction of AI in municipalities, https://www.soumu.go.jp/main_content/000820109.pdf (accessed on 22 January 2025).
[6] Milanez, A., A. Lemmens and C. Ruggiu (2025), “Algorithmic management in the workplace: New evidence from an OECD employer survey”, OECD Artificial Intelligence Papers, No. 31, OECD Publishing, Paris, https://doi.org/10.1787/287c13c4-en.
[25] NRI and NRI Mirai. (2018), Survey on the employment of persons with disabilities, https://www.nri.com/-/media/Corporate/jp/Files/PDF/news/newsrelease/cc/2018/181130_1.pdf.
[7] OECD (2025), Generative AI and the SME Workforce: New Survey Evidence, OECD Publishing, Paris, https://doi.org/10.1787/2d08b99d-en.
[1] OECD (2024), “Explanatory memorandum on the updated OECD definition of an AI system”, OECD Artificial Intelligence Papers, No. 8, OECD Publishing, Paris, https://doi.org/10.1787/623da898-en.
[17] OECD (2024), OECD Employment Outlook 2024: The Net-Zero Transition and the Labour Market, OECD Publishing, Paris, https://doi.org/10.1787/ac8b3538-en.
[34] OECD (2023), OECD Employment Outlook 2023: Artificial Intelligence and the Labour Market, OECD Publishing, Paris, https://doi.org/10.1787/08785bba-en.
[9] Recruitment Works Institute (2024), It’s not too late to touch generative AI now - through a survey of the ‘current’ use of generative AI and its characteristics, https://www.works-i.com/research/project/newcareer/issue/detail10.html (accessed on 29 November 2024).
[10] Suzuki, T. (2024), “The impact of AI on employment - in light of the Tankan survey of workers”, JTUC Rengo Research Institute DIO, Vol. JTUC Rengo Research Institute DIO No.393, pp. 39-43, https://www.rengo-soken.or.jp/dio/dio393-h.pdf.
[42] The Conference toward AI Network Society (2022), Compilation of Case Studies for “Safe, Secure, and Trustworthy Implementation of AI in Society”- Collection of Good Practices -, https://www.soumu.go.jp/main_content/000888398.pdf (accessed on 22 January 2025).
[30] Touzet, C. (2023), “Using AI to support people with disability in the labour market: Opportunities and challenges”, OECD Artificial Intelligence Papers, No. 7, OECD Publishing, Paris, https://doi.org/10.1787/008b32b7-en.
[13] Veritas Technologies LLC. (2024), Survey on the use of generative AI tools in the workplace, https://www.veritas.com/ja/jp/news-releases/2024-02-22-ai-research-japan (accessed on 16 December 2024).
[16] Yamamoto, I. (ed.) (2019), Artificial Intelligence and the Economy, Keiso syobo.
← 1. In November 2022, the American company OpenAI released “ChatGPT (Generative Pre‑trained Transformer)”.
← 2. Algorithmic management is the use of a diverse set of technological tools to automate tasks traditionally carried out by human manager. It started to receive attention over the last decade, initially due to legal action against firms that have relied on algorithmic decision making in employment contexts. More recently, interest has grown with recognition that the use of such tools has spread from the confines of the platform economy to more traditional work settings in what has been dubbed “the platformisation of work”
← 3. (Milanez, Lemmens and Ruggiu, 2025[6]) reveals that the use of algorithmic management software among mid-level managers in Japan is the lowest among the surveyed countries. In contrast, (JILPT, 2025[5]) shows that the proportion of workers managed by AI in Japan is higher than the average across the seven countries surveyed. These findings indicate the possibility that AI is being utilised by Japanese workers as a tool for self-management. For example, in Japan, individuals make use of AI-powered services for purposes such as scheduling meetings, managing calendars, automatically organising tasks, prioritising them, and optimising the timing of reminders.
← 4. The aggregated data was provided by Professor Isamu Yamamoto, Director of the Panel Data Research Center at the Keio University in Japan.
← 5. Another possible reason for the low proportion of AI usage in Japan could be that Japanese respondents applied stricter criteria when determining whether their companies use AI. While the JILPT survey used the same question wording as the previous OECD survey, it included additional supplementary explanations to help respondents answer accurately. For instance, it clarified: “Please answer “Yes (the company uses AI)” if AI is used in any way in work, regardless of how frequently it is used.” Despite these efforts, Japanese respondents may have been less likely to answer “Yes (the company uses AI)” unless AI was used on a relatively consistent basis.
← 6. The figure for the year 2000 is based on the results of the Special Survey of the Labour Force Survey conducted in February.
← 7. The results for the industry variable in the estimation equation Annex Table 1.A.1 are as follows. Using the Accommodation and food services sector as the reference group, statistically significant effects were observed for the following industries: Agriculture, forestry, and fisheries (marginal effect: 6.3%, Z-value: 2.3), Manufacturing (2.3%, 1.9), Information and communications (6.7%, 5.3), Transport and postal services (2.4%, 1.8), Wholesale and retail trade (2.9%, 2.5), Finance and insurance (5.0%, 3.7), Scientific research, professional and technical services (5.6%, 3.7), Medical, healthcare and welfare (−4.3%, −3.4), and Compound services (5.9%, 2.7).
← 8. The Employment Conditions Diffusion Index (D.I.) is calculated by subtracting the proportion of companies reporting “Insufficient employment” from the proportion of those reporting “Excessive employment”. In the September 2024 survey, the D.I. for all sectors indicated ‑36, compared to an average of ‑33 in 2019. During the COVID‑19 period, the D.I. narrowed to ‑6 in the June and September 2020 surveys, but it has since shown a continuous trend of widening negative values.
← 9. For example, it states: “To address labour shortages, enhance productivity, and advance digital transformation (DX), Japan will accelerate the implementation of AI and robots in the manufacturing and service sectors by developing and upgrading the necessary infrastructure.”
← 10. Employees may provide sensitive personal information, so it is important that AI service providers take appropriate measures to address the associated risks and handle such data responsibly.
← 11. For example, it states: “Japan will promote social implementation of AI to address social issues such as the ageing of the population in rural areas and the education gap with urban areas.”
← 12. Among AI users, when limiting the analysis to those who use GEAI based on their own initiative rather than company instruction and conducting the same type of regression analysis as Annex Table 1.A.1, the gender difference in AI usage was not statistically significant. These findings suggest that more women than men use GEAI based on company instruction, and that the results in the Table 1.A.1 may reflect this pattern.
← 13. Japan’s Digital Agency, in its “Priority Plan for the Realisation of a Digital Society”, identifies “a digital society where no one is left behind” as one of six key goals. The plan emphasises the importance of inclusiveness and equity in digital access and use.
← 14. For example, it states: “Japan will accelerate initiatives to develop human resources for digitalisation by fostering and securing young and female professionals in rural areas and supporting their matching with diverse job opportunities.”