Generative AI and the SME Workforce

1. Introduction

Generative AI: a transformative technology for SMEs?

Small- and Medium-sized Enterprises (SMEs) are a cornerstone of labour markets, accounting for over 99% of companies and 60% of business sector employment in OECD economies (OECD, 2023[1]). Yet they consistently lag larger companies in digitalisation and advanced technologies and are especially vulnerable to skill gaps (Bianchini and Lasheras Sancho, 2025[2]; OECD, 2024[3]). These challenges not only affect individual SMEs and their employees, but also constrain overall productivity and economic growth. The questions of how to close these digital and skill gaps, how to enable SMEs to use AI in a trustworthy manner, and how to help them comply with regulations and standards, are of prime importance to the OECD and its member countries.

To better understand the potential for generative AI to address these challenges, the OECD conducted a survey at the end of 2024 with over 5 000 SMEs across seven countries. The survey examines: to what extent and how SMEs use generative AI; the perceived benefits (including the potential to improve productivity); the impact on labour and skills needs (including where the SME is experiencing shortages); barriers to use; and what measures SMEs are taking to prepare employees to use generative AI.

Insights from the survey are intended to help policymakers identify the impacts of generative AI already being experienced within SMEs in their countries and to develop policy to close digital and skill divides between companies of different sizes.

Survey methodology

The survey covered 5 232 SMEs in Austria, Canada, Germany, Ireland, Japan, Korea and the United Kingdom. The OECD commissioned Ipsos NV to undertake the survey of SMEs, defined for the purpose of the survey as companies with up to 249 employees (i.e. including one‑person companies). Interviewers asked to speak to the person with the best overview of the technologies used in the company, often the owner or manager, as described in Box 1.1.

The main fieldwork was conducted between 14 October and 6 December 2024, in English, French, German, Japanese and Korean as needed. It was conducted as a telephone survey (CATI or Computer Assisted Telephone Interviewing), with sampling frames provided by TSR Data for Japan, SMTp for Korea, and Dun & Bradstreet for all other countries, all recent high-quality and high-coverage business registers.1 A stratified random sampling approach by country, company size and sector was used to draw the sample.

Box 1.1. .How respondents were recruited for the survey

Initial communications with prospective respondents purposefully omitted the terms “generative Artificial Intelligence” and “Artificial Intelligence” when describing the survey, to avoid biasing participation towards SMEs where AI or generative AI is in use. Instead, initial communications referred to a survey on the topic of “technology”. Interviewers asked to speak to the person with the best overview of the technologies used in the company. They explained that usually this would be the owner or manager of the company, but that it could also be a specific manager for technology if there was one.

Only SMEs where this individual had heard of “generative AI” were screened into the survey. The following definition was used to establish the concept as clearly and consistently as possible:

In this interview, I’m going to ask you about technologies that use artificial intelligence to generate text, images, video or audio. These technologies are sometimes called “generative artificial intelligence” or “generative AI”. ChatGPT is an example of generative AI used to generate text [in Korea, HyperCLOVA X was also given as an example]. For instance, ChatGPT could be used to write an email or create content for a website. Microsoft Copilot is an example of generative AI used to write code for a computer program. Stable Diffusion is an example of generative AI used to generate images.

Where the individual most knowledgeable about technology within the SME had not heard of “generative AI”, it was considered that they would be unable to respond knowledgeably to the interview questions. These SMEs were therefore screened out of the survey, although they were taken into account later when weighting the data and calculating the proportion of SMEs using generative AI.

Table 1.1 and Table 1.2 provide a breakdown of the survey sample (raw numbers and weighted percentages) by country, sector and company size. Throughout this report, results are weighted (by the number of firms rather than by number of employees) to ensure that the analysis is representative of the underlying population of SMEs in the countries studied, in terms of these characteristics, as reflected in official statistics.

Table 1.1. Descriptive statistics – number of surveys completed

Number of surveys completed by country, NACE sector grouping and size

	Country
	United Kingdom	Germany	Japan	Korea	Austria	Canada	Ireland	Total
	Number of observations
NACE Sector (The Statistical Classification of Economic Activities in the European Community)
Manufacturing, mining, and agriculture (A, B, C)	91	109	42	76	74	83	58	533
Construction, engineering, and waste collection and treatment (D, E, F)	97	102	53	73	85	122	123	655
Trade, transport, food / accommodation, and arts / entertainment activities (G, H, I, R)	243	247	179	376	294	274	276	1 889
Information and communication, financial / insurance, real estate, professional / administrative activities, other services (J, K, L, M, N, S)	293	244	89	250	248	214	258	1 596
Public administration and defence; compulsory social security, education, human health and social work (O, P, Q)	77	104	37	36	104	111	90	559
Size
1 employee	300	174	135	299	271	231	325	1 735
2 to 9 employees	187	290	102	206	246	130	196	1 357
10 to 49 employees	176	205	96	183	170	258	157	1 245
50 to 249 employees	138	137	67	123	118	185	127	895
Total	801	806	400	811	805	804	805	5 232

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

Table 1.2. Descriptive statistics – composition of the weighted sample

Percentage of total surveys completed (weighted) by country, NACE sector and size

	Country
	United Kingdom	Germany	Japan	Korea	Austria	Canada	Ireland	Total
	% of total (weighted)
NACE Sector (The Statistical Classification of Economic Activities in the European Community)
Manufacturing, mining, and agriculture (A, B, C)	7.1	6.5	10.7	9.7	4.8	5.2	3.8	6.8
Construction, engineering, and waste collection and treatment (D, E, F)	15.0	14.6	11.6	8.8	8.1	15.0	17.3	12.9
Trade, transport, food / accommodation, and arts / entertainment activities (G, H, I, R)	30.7	30.4	43.1	53.3	30.6	30.9	27.1	35.1
Information and communication, financial / insurance, real estate, professional / administrative activities, other services (J, K, L, M, N, S)	42.2	39.3	24.6	21.1	38.9	35.5	41.2	34.7
Public administration and defence; compulsory social security, education, human health and social work (O, P, Q)	4.9	9.2	10.2	7.1	17.7	13.5	10.6	10.5
Size
1 employee	56.0	31.5	47.9	53.7	50.1	51.4	58.8	49.9
2 to 9 employees	34.8	53.4	37.3	40.8	42.6	29.0	34.4	38.9
10 to 49 employees	7.8	12.9	12.1	4.7	6.1	16.2	5.6	9.3
50 to 249 employees	1.5	2.2	2.7	0.8	1.1	3.4	1.2	1.8
Total	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0

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

The full questionnaire in English, as well as more information on the sampling approach, fieldwork, data cleaning and weighting is provided in the technical report available on the OECD website.

Brief literature review

The purpose of this literature review is to provide context for the survey, to identify the main gaps that this report seeks to address, and to establish a basis for comparing survey results with existing evidence.

Large companies are more than twice as likely to use AI than SMEs

According to national statistics aggregated by the OECD, AI (including generative and non-generative AI) is used in 40.1% of large companies on average across OECD Member countries, while only in 20.4% and 11.9% of medium and small companies respectively. Large companies are more likely to use AI than SMEs in all countries included in Figure 1.1. This finding is common to many major studies of companies’ use of AI, including the OECD employer survey on the impact of AI on the workplace (Lane, Williams and Broecke, 2023[4]) and the US Annual Business Survey (Zolas et al., 2020[5]). Analysing official surveys across 11 OECD countries with a harmonised methodology, OECD analysis has established that this gap remains when comparing firms operating in the same sectors and even after accounting for the role of other confounding factors such as firm age or assets composition (Calvino and Fontanelli, 2023[6]).

Figure 1.1. AI use increases with company size

Percentage of companies reporting use of AI, by country and company size

Note: Latest data available ranges from 2020 to 2024 depending on the country. Eurostat data collection uses standardised questions, facilitating cross-country comparisons. Surveys from other regions differ in design, sectoral coverage and definitions, affecting comparability.

Source: OECD, ICT Access and Usage by Businesses, Businesses using artificial intelligence (AI), http://data-explorer.oecd.org/s/1r1, using survey data collected by National Statistical Offices.

Lower use of AI is just one aspect of a widening digital gap between SMEs and larger companies. A recent OECD report (2024[3]) describes digitalisation as an important source of resilience for SMEs against economic shocks, as a tool for experimenting with and adopting new, innovative and sustainable business practices, and as a lever to increase productivity, scale up and compete with larger companies. The report shows that although digitalisation among SMEs has increased, SMEs have not kept pace with larger firms. SMEs fall behind due to various factors, including insufficient access to finance and limited organisational capabilities, which can stymie investment in technical equipment as well as efforts to upskill and drive organisational change.

Evidence on the use of AI within microbusinesses (i.e. companies with fewer than 10 employees) is rare. Nonetheless, data collected in five European countries, Belgium, Spain, the Netherlands, Portugal and Sweden, (Eurostat, 2024[7]) suggests that AI use by microbusinesses was on average approximately half that of companies with 10 to 249 employees when captured in 2024.

Generative AI use is growing among SMEs

The use of generative AI has grown rapidly since its inception among SMEs, larger companies and private individuals. On 30 November 2022, OpenAI publicly launched a ChatGPT demo, allowing users to interact with it for free. Users shared examples of its output across social media and, within five days, ChatGPT had reached one million users. Image generation tools such as DALL·E 2, Stable Diffusion and Midjourney, which preceded ChatGPT, also attracted online attention. Microsoft, Anthropic and Google quickly followed by publicly launching their own text generation tools within the following five months. In 2023, global venture capital investments in generative AI firms totalled approximately USD 18 billion, followed by nearly USD 20 billion in 2024 (OECD.AI, 2025[8]).

Today, generative AI is widely used for many different tasks. A recent snapshot of Anthropic’s text generation tool, Claude.ai, showed that 37% of four million analysed Claude.ai interactions were related to computer and mathematical tasks (such as coding and debugging), 10% to arts and media tasks (possibly reflecting its use for marketing, writing and content generation) and 9% to education and library tasks (Handa et al., 2025[9]).

Generative AI holds significant potential for enhancing labour productivity in the workplace. However, there is less evidence on its adoption across companies than there is for AI more broadly. Eurostat provides figures on enterprise use of AI technologies that generate written or spoken language (Natural Language Generation), although it excludes other forms of content generation. According to this indicator, use of AI for language generation in 2024 among SMEs remains modest, with figures ranging from 1.3% in Romania to 17.8% in Denmark (Figure 1.2). Nonetheless, uptake is increasing considerably over time: in 2023 the highest reported use of generative AI among SMEs was just 4.7%, observed in Slovenia. (Eurostat, 2023[10]) In the United Kingdom, a 2025 survey of companies of all sizes reveals that 8.2% use Large Language Models (LLMs) for text generation and 7.0% use visual content creation. (Office for National Statistics, 2025[11]) In Canada, 8.5% of businesses with one to four employees were already using generative AI in the first quarter of 2024, compared with 14.7% of businesses with 100 or more employees. (Statistics Canada, 2024[12]).

Figure 1.2. Natural language generation use among SMEs

Percentage of SMEs using AI technologies generating written or spoken language (natural language generation)

Note: Companies with 10 to 249 persons employed.

Source: Eurostat (2024[7]), [isoc_eb_ai] Artificial intelligence by size class of enterprise, 2024, https://ec.europa.eu/eurostat/databrowser/view/isoc_eb_ai__custom_16529522/default/table?lang=en&page=time:2024.

Can generative AI enable SMEs to achieve more with fewer resources?

Generative AI tools have a few features which could make them particularly attractive to SMEs, as outlined in a recent OECD report on “SME digitalisation to manage shocks and transitions” (2024[3]). Many generative AI tools can be accessed for free or for a low monthly subscription fee, with more advanced features unlocked with a higher fee. Entrepreneurs and employees can interact with generative AI tools in conversational dialogue, without needing to know how to code. Since generative AI tools have been trained on massive amounts of online content, an SME can use the tool to generate basic content regardless of the maturity or scale of its own internal data infrastructure. The report explains that the main implication of this is that generative AI enables SMEs to achieve more with fewer resources, which could be particularly useful for SMEs grappling with labour shortages and skill gaps.

Difficulty finding employees with the right skills is the top-ranking challenge for SMEs in the EU (European Commission, 2023[13]), above other challenges such as access to finance and regulatory obstacles. The main consequences are increased workload among staff (reported by 48% of those experiencing this challenge), followed by loss of sales (31%) and reduced profitability and productivity (each 25%). While these shortages affect firms of all sizes, SMEs and entrepreneurs tend to be more vulnerable to short-term skills imbalances compared to large firms, because of their limited capacity to reorganise or outsource activities, automate and attract specialised workers (OECD, 2023[14]).

The idea that generative AI could help SMEs meet this challenge is the main research question explored by this survey, which collects new data on how SMEs use generative AI across seven countries. The survey contributes to the existing evidence (including the 2025 OECD D4SME Survey on digitalisation of SMEs, described in Box 1.2.) by delivering:

• data based on a comprehensive definition of generative AI covering text, audio, video and image;

• comparable and representative data for each country;

• a systematic analysis of the reported labour market impacts of generative AI;

• insights to inform policy to close digital and skill divides between companies of different sizes.

Box 1.2. Connections between this survey and the 2025 OECD D4SME Survey

The 2025 OECD Digital for SMEs (D4SME) Survey examines the digitalisation of small and medium-sized enterprises (SMEs) across ten OECD countries: Australia, Canada, France, Germany, Italy, Japan, Korea, Spain, the United Kingdom, and the United States. The survey was conducted in Q4 2024, contemporaneously to this survey, and gathered answers from 1 009 SMEs using large digital platforms and service providers (Amazon, Intuit, Kakao, Rakuten, Sage). While the scope of the OECD D4SME Survey is broader (covering AI technologies, digital tools, fintech solutions and digital security) than this survey, it also captures valuable insights into SMEs’ use of generative AI, complementing the findings in this report. Key insights include:

• Generative AI is on the rise: the share of SMEs using generative AI rose by 40% since the previous year. Key barriers to generative AI include concerns about legal issues and data privacy, as this study also shows in Chapter 4. The D4SME report calls on governments to provide tools and resources specifically dedicated to unlocking more advanced, productivity-enhancing uses of large language models (LLMs) by SMEs.

• Enhancing digital security remains central: two out of three SMEs do not have “robust” digital security practices in place, suggesting a gap in the knowledge, means and skills needed to ensure safe and responsible digital uptake by SMEs. While this study did not capture information on digital security skills specifically, the D4SME findings suggest scope for developing these skills, alongside the skills mentioned in Chapter 3 (e.g. data analysis and interpretation).

• Governments can support SME digitalisation: only 21% of SMEs are aware of government digitalisation support, with only half of this number accessing it. The D4SME report highlights the need for more streamlined and better-targeted initiatives to improve SME access to government digitalisation programmes, a point that also applies to the government supports (training, financial assistance, information campaigns and business mentoring) described in Chapter 4.

Source: (Bianchini and Lasheras Sancho, 2025[2]), “SME digitalisation for competitiveness: The 2025 OECD D4SME Survey”.

References

[2] 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.

[6] 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.

[13] European Commission (2023), Eurobarometer - SMEs and skills shortages, https://europa.eu/eurobarometer/surveys/detail/2961.

[7] Eurostat (2024), Artificial intelligence by size class of enterprise, https://ec.europa.eu/eurostat/databrowser/view/isoc_eb_ai__custom_16529522/default/table?lang=en&page=time:2024.

[10] Eurostat (2023), Artificial intelligence by size class of enterprise, https://ec.europa.eu/eurostat/databrowser/view/isoc_eb_ai__custom_12093291/default/table?lang=en&page=time:2023.

[9] Handa, K. et al. (2025), “Which Economic Tasks are Performed with AI? Evidence from Millions of Claude Conversations.”, arXiv preprint, https://arxiv.org/abs/2503.04761.

[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.

[3] OECD (2024), “SME Digitalisation to manage shocks and transitions: 2024 OECD D4SME survey”, OECD SME and Entrepreneurship Papers, No. 62, OECD Publishing, Paris, https://doi.org/10.1787/eb4ec9ac-en.

[1] OECD (2023), Future-Proofing SME and Entrepreneurship Policies: Key Issues Paper, https://www.oecd.org/content/dam/oecd/en/topics/policy-issues/smes-and-entrepreneurship/2023-ministerial-documents/Ministerial-key-issues-paper-2023.pdf.

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

[8] OECD.AI (2025), Worldwide VC Investments in generative AI visualisations powered by JSI using data from Prequin, http://www.oecd.ai (accessed on 8 July 2025).

[11] Office for National Statistics (2025), Business Insights and Conditions Survey (BICS) Wave 129, 17.03-30.03.2025, https://www.ons.gov.uk/economy/economicoutputandproductivity/output/datasets/businessinsightsandimpactontheukeconomy.

[12] Statistics Canada (2024), Canadian Survey on Business Conditions, first quarter of 2024, Business’s use of Generative AI, first quarter of 2024, https://www150.statcan.gc.ca/t1/tbl1/en/tv.action?pid=3310078401.

[5] Zolas, N. et al. (2020), “Advanced Technologies Adoption and Use by U.S. Firms: Evidence from the Annual Business Survey”, https://www.census.gov/fsrdc. (accessed on 7 May 2025).

Note

← 1.  For instance, the Dun and Bradstreet sampling frames were also used for the EU-OSHA Fourth European Survey of Enterprises on New and Emerging Risks (ESENER 2024), and for the European Commission DG ECFIN Financial services Sector Survey (FINA, 2012-2023).