Generative AI and the SME Workforce

4. Are SMEs prepared for generative AI?

Main findings

The most common reason for not using generative AI is that it is not suited to the work the SME does (57%). More than half of SMEs (54%) report concerns about copyright, legal or regulatory issues and about what happens to the information fed into generative AI models. Concerns about quality of the output (35%) and value for money (21%) are less common.

Most SMEs (86%) hold neutral or positive attitudes toward generative AI. Only 2% prohibit its use. Attitudes towards generative AI do not therefore appear to be a major barrier.

A third or fewer of SMEs using generative AI are taking measures to train staff, set internal guidelines, or research copyright, legal, and regulatory issues. To close digital and skills gaps between SMEs and larger firms and to ensure that any gains from generative AI are broadly shared across the economy and the workforce, more structured support for SMEs may be needed. SMEs particularly welcome government support in the form of training, financial assistance, information campaigns and business mentoring.

This section examines the preparedness of SMEs to seize the benefits of generative AI while addressing and minimising the risks. It discusses the barriers faced by SMEs to using generative AI, attitudes within SMEs towards generative AI, and measures to prepare workers to use generative AI such as training, research and providing guidelines.

The most common barrier to using generative AI is that it is not suited to the work the SME does

Among SMEs not using generative AI, the most common explanation is that generative AI is not suited to the type of work the company does (57.3%). This is the top barrier for SMEs in sectors such as agriculture, mining, manufacturing and construction, which rely on physical abilities outside the capabilities of generative AI, but this is also the top barrier in some service sectors, such as finance and professional, scientific and technical activities. In fact, 70.4% of SMEs in financial and insurance activities report this as a reason why they do not use generative AI.

The next most reported barrier is concerns about copyright, legal or regulatory issues (54.1%), which are the top barrier among SMEs in sectors such as public administration and defence, human health and social work and administrative and support services. 88.1% of SMEs in public administration and defence gave this as a reason why they do not use generative AI. 52.5% of SMEs say that they do not use generative AI because of concerns about what happens to the information fed into generative AI models.1 Half of non-users (49.8%) report that employees do not have the right skills to use generative AI. This is the main barrier for SMEs in the real estate sector. Concerns about client approval, quality of the output and about value for money2 are less common among SMEs, although client approval is the top barrier reported by SMEs in the education sector.

Figure 4.1. The most common barrier to using generative AI is that it is not suited to the work the company does

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 use generative AI to address skill and labour needs, 2024.

The most reported barrier does not vary by company size, but it does by country (Table 4.1). Lack of skills is the most reported barrier in Japan (82.9%) and Korea (53.0%). In Germany, the most reported barrier is that clients would not approve of the use of generative AI (67.2%).

Table 4.1. The main barrier to using generative AI varies by country

Country	Main barrier	Main support
Austria	Not suited to the type of work (53.0%)	Financial assistance (62.2%)
Canada	Not suited to the type of work (65.0%)	Training (61.4%)
Germany	Company’s clients would not approve (67.2%)	Information campaigns (66.6%)
Ireland	Concerns about copyright, legal or regulatory issues (56.3%)	Training (76.9%)
Japan	Employees do not have the right skills (82.9%)	Financial assistance (69.5%)
Korea	Employees do not have the right skills (53.0%)	Training (77.4%)
United Kingdom	Not suited to the type of work (66.9%)	Training (79.3%)

Note: For the main barrier, 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?”. For the main support, all SMEs surveyed were asked:

“In your opinion, would it be very helpful, somewhat helpful or not helpful if your national government offered the following support to help your company use generative AI?”. To assess which support is the most helpful, a score was created giving an arbitrary weight of one to a “somewhat helpful” response and a weight of two to a “very helpful” response.

Source: OECD survey on how SMEs use generative AI to address skill and labour needs, 2024.

Governments can provide support to SMEs to overcome barriers to use, but also to help companies already using generative AI in harnessing its full potential. Across all SMEs, including those that use generative AI and those that do not, training is considered the most helpful support the government could offer, followed by financial assistance, information campaigns and business mentoring. This result varies by country: training is the most helpful support in the United Kingdom, Ireland, Canada and Korea; in Japan and Austria, financial assistance is the most helpful support; while in Germany, it is information campaigns. The OECD 2025 D4SME Survey shows scope to improve the awareness of government digitalisation supports among SMEs, to streamline application processes and to improve the targeting of supports (Bianchini and Lasheras Sancho, 2025[1]).

A small minority of SMEs report an unfavourable attitude towards generative AI

In most SMEs, attitudes towards generative AI are not a barrier to use. 85.9% of SMEs report either a neutral or favourable attitude towards generative AI within the company (Figure 4.2). Nonetheless, 11.2% report an unfavourable attitude and a further 2.5% prohibit its use entirely.

Attitudes are more likely to be favourable among SMEs using generative AI compared to non-users in Japan (60.9% vs. 22.5%), Korea (42.9% vs. 19.1%) and Germany (39.2% vs. 27.8%), but not in other countries. Non-users are more likely to be either unfavourable or prohibit generative AI use in Germany (27.0% vs. 16.4%), United Kingdom (17.4% vs. 10.6%), and Japan (8.0% vs. 1.5%), but not significantly in other countries. In all countries surveyed, non-users who indicate that they are likely to use generative AI within the next year more often report favourable attitudes compared to non-users that consider this unlikely.

Attitudes are most polarised in Germany. It presents the highest share of SMEs favourable to generative AI (32.7%), the second highest share unfavourable to generative AI (14.8%), and the highest share prohibiting its use (7.7%). The large share of favourable attitudes is consistent with Germany’s high reported use of generative AI, while the unfavourable opinions and prohibitions could be partly driven by concerns about clients’ reactions, which were the biggest barrier to use in Germany. Despite use of generative AI being lower in Japan and Korea, SMEs in these countries report the lowest shares of both unfavourable attitudes and prohibition (Figure 4.2).

Figure 4.2. Most SMEs have either a favourable or neutral attitude towards generative AI

Percentage of SMEs that report a favourable, neutral, unfavourable attitude towards generative AI, or its prohibition, by country

Note: Respondents not using generative AI, excluding one person businesses, were asked: “Is the use of generative AI prohibited within your company?”. Respondents using generative AI and those not using generative AI and not prohibiting it were asked: “Within your company, would you say that the attitude towards generative AI is favourable, unfavourable or neutral?”

Source: OECD survey on how SMEs use generative AI to address skill and labour needs, 2024.

Some SMEs are taking measures to prepare workers to use generative AI in a trustworthy manner

Some SMEs are taking measures to prepare workers to use generative AI in a trustworthy manner, thereby enabling the company to seize the benefits of generative AI while addressing and minimising the risks (some of which are discussed in Box 4.1). These measures could include: training workers, researching potential risks, and providing workers with guidelines. The Danish study described in Box 2.3 (Humlum and Vestergaard, 2025[2]) found that firm-provided training and employer encouragement significantly boosts workers’ use of generative AI and reduces demographic gaps in use. Additionally, the benefits from generative AI – time savings, quality improvements, creativity, task expansion, and job satisfaction – are all 10% to 40% greater when employers encourage its use. The authors highlight the importance of firm-led complementary investments in unlocking the productivity potential of new technologies.

Box 4.1. What risks related to copyright, data protection, and legal and regulatory issues are associated with generative AI?

The use of generative AI in the workplace presents opportunities, but also risks related to copyright, legal and regulatory issues, and to data protection. In the G7 Hiroshima Process on Generative AI, five of the seven G7 members considered intellectual property right infringement and threats to privacy as major risks (OECD, 2023[3]).

Generative AI outputs are not created spontaneously; they result from algorithms trained on large datasets, which enable the model to associate user inputs with learned patterns. These datasets are often compiled from publicly available sources such as third-party websites, databases and social media platforms. If the training data includes protected works (e.g. copyrighted material, image rights, or personal data) without proper authorisation or licensing, the generated outputs may infringe third-party rights (European Innovation Council and SMEs Executive Agency, 2024[4]; OECD, 2025[5]; OECD, 2024[6]). Where personal data are used to train machine learning models, data subjects may face challenges exercising their rights to have the data deleted or corrected, for example because retraining models is resource intensive and because of the difficulty of identifying the data point associated with an individual in an unstructured dataset (OECD, 2024[6]).

In addition, users may inadvertently disclose personal, confidential or proprietary information when interacting with generative AI tools. This information may be retained and used by service providers to improve model performance. For example, under OpenAI’s terms of use, input data and outputs may be used for model development purposes unless users actively opt out.

To mitigate these risks and to ensure that generative AI is used in a trustworthy manner in line with the company objectives and values, SMEs may conduct research on copyright, legal or regulatory issues, establish internal guidelines on the responsible use of generative AI, and offer targeted training to raise awareness among staff about the capabilities and limitations of these technologies. The OECD AI principles [OECD/LEGAL/0449] encourage businesses deploying AI to apply a systematic risk management approach and adopt responsible business conduct to address risks related to security, privacy and intellectual property rights, while the OECD Recommendation of the Council on SME and Entrepreneurship Policy [OECD/LEGAL/0473] calls on countries to support all SMEs and entrepreneurs in adopting digital technologies in line with their needs, digital maturity and aspirations by enhancing access to digital infrastructure; strengthening digital skills, data literacy and management of digital security risk.

Note: And where relevant, in line with the OECD AI principles [OECD/LEGAL/0449], which guides AI actors in their efforts to develop trustworthy AI and provide policymakers with recommendations for effective AI policies.

AI-related training is not common

23.6% of SMEs using generative AI report that their employees participate in training related to AI, ranging 11.3% for SMEs in Japan to 29.4% for SMEs in Canada (Figure 4.3). This share drops to 2.7% for non-users, despite lack of skills being a commonly reported barrier to using generative AI. AI-related training could help workers acquire new skills complementary to AI use and raise awareness among staff about the capabilities, limitations, and risks associated with these technologies. In addition to the risks discussed in Box 4.1, overreliance on generative AI has been shown to lead to reduced performance due to difficulty in detecting errors, particularly for complex tasks among less experienced users (Dell’Acqua et al., 2023[7]; Kabir et al., 2024[8]; Kreitmeir and Raschky, 2024[9]). In the OECD AI surveys, worker outcomes of AI were generally better where workers received training to work with AI (Lane, Williams and Broecke, 2023[10]).

Figure 4.3. Where SMEs use generative AI, employees’ participation in AI-related training is not common

Percentage of SMEs that report their employees participate in AI-related training

Note: Respondents were asked: “Do employees in your company currently participate in training related to AI?“ Results are limited to SMEs using generative AI.

Source: OECD survey on how SMEs use generative AI to address skill and labour needs, 2024.

There are several possible explanations for why participation in AI-related training is not more common, even among SMEs where generative AI is in use. One is that SMEs may not perceive a need to offer training, since workers can interact with generative AI in conversational dialogue, on their own individual initiative and without needing to know how to code. Another possibility is that generative AI has advanced so rapidly that SMEs have not yet had the opportunity to organise training.

Additionally, labour shortages can make it difficult for SMEs to allocate time for training3. The low participation in AI-related training in Japan might be partly explained by Japanese SMEs being the most likely to report experiencing a lack of skills or experience among staff and second most likely to report a worker shortage (Chapter 3). Across OECD countries, time constraints due to work are the most cited barrier to participation in job-related non-formal learning, with the share of adults citing this reason in Japan exceeding the OECD average (OECD, 2025[11]).

In general, larger companies are more likely to offer training to their workers (OECD, 2024[12]).4 Compared to larger companies, SMEs face higher unit costs per worker and may lack the resources to finance training opportunities. Moreover, with fewer employees, SMEs have less flexibility to release staff from revenue-generating activities to undertake training. Concerns around worker poaching, where workers leave for better opportunities after receiving training, further discourage investment in training. SMEs typically offer lower wages, less attractive working conditions, and fewer career advancement opportunities than larger companies. As a result, trained employees may be more inclined to seek employment elsewhere, reducing the returns to training investments for small businesses (OECD, 2023[13]).

Government intervention may be needed to promote AI literacy in the general population and to address the specific challenges that SMEs face in investing in training, including by implementing good practices identified by the OECD (2021[14]). Financial incentives – such as subsidies and vouchers – can help lower the costs of training provision. Pay-back clauses, which make employees liable for some or all the costs of their training if they leave their employer within a specified period would strengthen companies’ incentives to offer training by decreasing the risk of poaching.

One third of SMEs using generative AI have done research on copyright, legal or regulatory issues associated with generative AI

Some SMEs have taken action to mitigate the possible risks related to copyright and data protection. 35.6% of SMEs using generative AI have done research on copyright, legal, and regulatory issues, either by seeking expert advice or on their own. This share drops to 4.8% among non-users, despite concerns about these issues being frequently cited by non-users as a barrier. Providing guidance to SMEs on these issues could help overcome some barriers to use. Austrian SMEs using generative AI are the most likely to have conducted research on these topics (44.7%), while SMEs in Korea are the least likely (17.5%) (Figure 4.4). SMEs using generative AI in the arts, entertainment and recreation sector are the most likely to have done research (53.0%), possibly due to the sector’s exposure to copyright-related risks.

Figure 4.4. Austrian SMEs using generative AI are the most likely to have conducted research on copyright, legal or regulatory issues

% of SMEs that report having done research on copyright, legal or regulatory issues

Note: Respondents were asked: “Has your company ever sought expert advice or done its own research on copyright, legal or regulatory issues associated with generative AI?” Results are limited to SMEs using generative AI.

Source: OECD survey on how SMEs use generative AI to address skill and labour needs, 2024.

A quarter of SMEs using generative AI have implemented guidelines for staff

Employers can provide guidelines to staff to help them understand how to use generative AI in a trustworthy manner in line with company objectives and values, when it should be used and when it should not be used. Among SMEs using generative AI, 28.6% have implemented guidelines for staff on the use of generative AI, while 8.6% of non-users have done so. The limited provision of internal guidelines may indicate a lack of perceived necessity or awareness among companies. It could also reflect that the use of generative AI tools may be driven by individual initiative rather than top-down strategies. The prevalence of such guidelines varies significantly by country. SMEs using generative AI in Germany are the most likely to have guidelines in place (45.4%), while those in Japan are the least likely (5.3%) (Figure 4.5). Among SMEs using generative AI and providing guidelines, 30.7% are concerned about staff not following them.

Figure 4.5. SMEs using generative AI in Germany are the most likely to have guidelines in place

Percentage of SMEs that report having guidelines in place

Note: Respondents were asked: “Does your company have guidelines for staff concerning the use of generative AI?” Results are limited to SMEs using generative AI.

Source: OECD survey on how SMEs use generative AI to address skill and labour needs, 2024.

In addition, 10.5% of SMEs report concerns about staff using generative AI in secret. These concerns are marginally more prevalent among SMEs not using generative AI (11.8%) than among SMEs using it (8.6%).5 Among SMEs already using generative AI, concerns about secret use may reflect fears that staff are not disclosing their use of such tools internally. In both cases, employer guidelines and training may help staff understand when generative AI should be used and when it should not be used, and how to use it in a trustworthy manner in line with company objectives and values.

References

[1] Bianchini, M. and M. Lasheras Sancho (2025), “SME digitalisation for competitiveness: The 2025 OECD D4SME Survey”, OECD SME and Entrepreneurship Papers, No. 68, OECD Publishing, Paris, https://doi.org/10.1787/197e3077-en.

[15] Calvino, F. and L. Fontanelli (2023), “A portrait of AI adopters across countries: Firm characteristics, assets’ complementarities and productivity”, OECD Science, Technology and Industry Working Papers, No. 2023/02, OECD Publishing, Paris, https://doi.org/10.1787/0fb79bb9-en.

[7] Dell’Acqua, F. et al. (2023), “Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of AI on Knowledge Worker Productivity and Quality”, SSRN Electronic Journal, https://doi.org/10.2139/SSRN.4573321.

[4] European Innovation Council and SMEs Executive Agency (2024), Artificial intelligence and copyright: use of generative AI tools to develop new content - European Commission, https://intellectual-property-helpdesk.ec.europa.eu/news-events/news/artificial-intelligence-and-copyright-use-generative-ai-tools-develop-new-content-2024-07-16-0_en (accessed on 4 April 2025).

[2] Humlum, A. and E. Vestergaard (2025), “Large Language Models, Small Labor Market Effects”, https://doi.org/10.3386/W33777.

[8] Kabir, S. et al. (2024), “Is Stack Overflow Obsolete? An Empirical Study of the Characteristics of ChatGPT Answers to Stack Overflow Questions”, Conference on Human Factors in Computing Systems - Proceedings, https://doi.org/10.1145/3613904.3642596.

[9] Kreitmeir, D. and P. Raschky (2024), “The Heterogeneous Productivity Effects of Generative AI”, SSRN Electronic Journal, https://doi.org/10.2139/ssrn.4745624.

[10] Lane, M., M. Williams and S. Broecke (2023), “The impact of AI on the workplace: Main findings from the OECD AI surveys of employers and workers”, OECD Social, Employment and Migration Working Papers, No. 288, OECD Publishing, Paris, https://doi.org/10.1787/ea0a0fe1-en.

[5] OECD (2025), “Intellectual property issues in artificial intelligence trained on scraped data”, OECD Artificial Intelligence Papers, No. 33, OECD Publishing, Paris, https://doi.org/10.1787/d5241a23-en.

[11] OECD (2025), Trends in Adult Learning: New Data from the 2023 Survey of Adult Skills, OECD Publishing, Paris, https://doi.org/10.1787/ec0624a6-en.

[6] OECD (2024), “AI, data governance and privacy: Synergies and areas of international co-operation”, OECD Artificial Intelligence Papers, No. 22, OECD Publishing, Paris, https://doi.org/10.1787/2476b1a4-en.

[12] OECD (2024), Understanding Skill Gaps in Firms: Results of the PIAAC Employer Module, OECD Skills Studies, OECD Publishing, Paris, https://doi.org/10.1787/b388d1da-en.

[3] OECD (2023), G7 Hiroshima Process on Generative Artificial Intelligence (AI): Towards a G7 Common Understanding on Generative AI, OECD Publishing, Paris, https://doi.org/10.1787/bf3c0c60-en.

[13] OECD (2023), OECD SME and Entrepreneurship Outlook 2023, OECD Publishing, Paris, https://doi.org/10.1787/342b8564-en.

[14] OECD (2021), Incentives for SMEs to Invest in Skills: Lessons from European Good Practices, Getting Skills Right, OECD Publishing, Paris, https://doi.org/10.1787/1eb16dc7-en.

Notes

← 1.  Concerns about legal issues and data privacy were also identified as key barriers to generative AI use in the OECD 2025 D4SME Survey (Bianchini and Lasheras Sancho, 2025[1]).

← 2.  While cost was the most reported barrier to AI adoption in the OECD AI survey of employers (Lane, Williams and Broecke, 2023[10]) and is often cited as a key factor explaining the gap in AI adoption between large companies and SMEs (Calvino and Fontanelli, 2023[15]), relatively few SMEs surveyed report that generative AI is bad value for money. Many generative AI tools are available at no cost, although they might entail fees to unlock features or access more advanced models.

← 3.  While Chapter 3 shows that generative AI can help address labour shortages, not all companies using generative AI report experiencing this benefit. Even among those that do, labour shortages may persist, albeit to a lesser extent.

← 4.  Results from the OECD Programme for the International Assessment of Adult Competencies (PIAAC) Employer Module (2022).

← 5.  If staff are using generative AI in secret, then use of generative AI among SMEs could be even higher than reported in this survey.