Generative AI and the SME Workforce

3. How is generative AI impacting SMEs’ skill and labour needs?

Main findings

Generative AI helps SMEs compensate for skill gaps and labour shortages. Among SMEs that use generative AI and have experienced a skill gap, 39% say that generative AI helped compensate for it. This figure is even higher (46%) where SMEs report that generative AI has improved employee performance, indicating that generative AI is particularly beneficial where a lack of skills is the main constraint for employee performance. SMEs not using generative AI are also optimistic that it could help them address skill gaps (42%) and worker shortages (24%), suggesting some untapped potential.

For a third of SMEs, using generative AI has reduced staff workload, or the entrepreneur’s own workload in the case of a one‑person business. In cases of overwork, technology that reduces workload by automating tasks and generating efficiencies may be very welcome. In addition, 14% of SMEs say that generative AI reduced their reliance on external contractors. It could be that generative AI enables these SMEs to perform tasks that they previously would have outsourced, given limited skills or resources in house.

Yet generative AI does not appear to lead SMEs to cut jobs. The vast majority of SMEs (83%) report that generative AI has had no effect on overall staff need. While AI might be expected to increase staff need by boosting productivity and consumer demand or decrease staff need by automating tasks, there are only modest signs of change, suggesting that SMEs prefer to wait before making internal staffing adjustments. 6% of SMEs report an increase in staff needs while 9% report a decrease.

Skills are essential for SMEs to use generative AI effectively. Although generative AI performs impressively on many cognitive tasks, it does not remove the need for skilled workers. Twice as many SMEs say that generative AI increases skills’ needs (20%) as say that it decreases them (9%). For workers, data analysis and interpretation skills and creativity and innovation skills are perceived as the skills that have increased the most in importance due to generative AI.

This section explores the idea of generative AI helping SMEs compensate for labour shortages and skill gaps through automation and efficiency improvements. It examines the impact of generative AI on SMEs’ labour and skills needs, including workload, reliance on external contractors, and number of staff needed, and investigates whether (and which) skills increase in importance due to generative AI.

Generative AI helps SMEs compensate for labour and skill shortages

One of the main motivations for this survey was to understand whether generative AI helps SMEs compensate for labour shortages and skill gaps through automation and efficiency improvements. Finding employees with the right skills is the top-ranking problem for SMEs in the EU (European Commission, 2023[1]), above other challenges such as access to finance and regulatory obstacles. Generative AI’s potential to automate tasks could be particularly welcome in an SME experiencing shortages, where it can relieve pressure on labour and skills.

Large SMEs are most likely to report a labour shortage or skill gap

In this survey, 36.8% of SMEs overall report experiencing a worker shortage in the past two years while 33.8% report a lack of skills or experience among staff in the same period (i.e. a skills gap). In many cases, SMEs report experiencing one but not the other: 15.3% of SMEs overall report experiencing both a worker shortage and a skill gap while 17.9% report a worker shortage only and 21.0% report a skill gap only.

SMEs in Japan are the most likely to report experiencing a skill gap (41.4%) and the second most likely to report a worker shortage (44.7%), behind Austria where 45.6% of SMEs report this (

Figure 3.1). The largest SMEs are the most likely to report a worker shortage (43.4%) or a skill gap (44.2%), although the prevalence of skill gaps does not increase monotonically with size. Larger SMEs are also more likely to report worker shortages in the European Commission survey than smaller SMEs (European Commission, 2023[1]), while the PIAAC Employer Module shows that medium and large firms are more likely than small firms to experience a skill gap (OECD, 2024[2]). This may reflect real differences in skill gaps by firm size, but also greater awareness of these gaps in large firms, which are more likely to assess skill needs.

Figure 3.1. Large SMEs are most likely to report a labour shortage or skill gap

Percentage of SMEs that have experienced worker shortage or skill gap

Note: SMEs using generative AI were asked: “In the last two years, has a worker shortage (a lack of skills or experience among staff) ever been a challenge for your company?”

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

SMEs in the accommodation and food services sector were most likely to have experienced worker shortages (48.7%), followed by SMEs in education (42.9%). SMEs in the financial sector were the most likely to have experienced a lack of skills or experience among staff (43.1%).

If labour shortages and skill gaps were driving SMEs to use generative AI,1 one might expect to see higher use among SMEs that have experienced these issues – but that is not the case. SMEs that have experienced either a worker shortage or a skill gap were roughly equally likely to report use of generative AI as those that have not experienced them.

SMEs report that using generative AI has helped compensate for labour and skill shortages

Among SMEs using generative AI and experiencing a skill gap in the last two years, 39.1% say that generative AI helped compensate for the gap. Similarly, 25.2% of SMEs facing a worker shortage in the last two years say generative AI helped them compensate for this worker shortage (Figure 3.2). Where generative AI is used in core activities, SMEs are more likely to say that it helped compensate for skill gaps and worker shortages, at 49.3% and 38.2% respectively.

Figure 3.2. Some SMEs report that generative AI helps compensate for labour shortages and skill gaps

Percentage reporting that generative AI helped compensate for worker shortage/skill gap, among SMEs using generative AI and experiencing shortage/gap

Note: SMEs using generative AI were asked: “Did generative AI help your company compensate for this worker shortage (the lack of skills or experience)?”

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

There is an inextricable link between generative AI compensating for skill gaps and improving employee performance.2 Where SMEs report that AI improved employee performance, they are twice as likely (46.3%) to say that it helped address skill gaps as those that do not report that it helped improve employee performance (21.1%). The direction of causality is unclear: whether improving employee performance is how generative AI helps to compensate for skill gaps, whether compensating for skill gaps is how generative AI improves employee performance or whether there is a third factor driving both. In any case, the finding suggests that where the constraint for employee performance is a lack of skills, generative AI is beneficial.

SMEs in Japan are far more likely than SMEs in any other country to report that generative AI helped compensate for a skill gap (63.3%, compared to the average 39.1%). Japan was also the country where SMEs were most likely to report a skill gap. SMEs in Japan and Austria are most likely to say that generative AI helped compensate for a worker shortage (32.6 and 32.2% respectively). These are also the countries where worker shortages were most prevalent, according to the SMEs surveyed. This suggests that where labour shortages and skill gaps are most acute, SMEs are most optimistic that generative AI provides a solution.

Figure 3.3. SMEs in Japan are most likely to say that generative AI helps compensate for labour and skill shortages

Percentage reporting that generative AI helped compensate for worker shortage/skill gap, among SMEs using generative AI and experiencing shortage/gap

Note: SMEs using generative AI were asked: “Did generative AI help your company compensate for this worker shortage (the lack of skills or experience)?”

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

Although the largest SMEs (with between 50 and 249 employees) are the most likely to experience either type of shortage, they are less likely than the average to say that generative AI helped compensate for these shortages. Instead, one‑person SMEs are more likely than larger SMEs to say that generative AI helped compensate for a worker shortage (30.5%). It could be that the compensating effect of generative AI is more noticeable in a workplace with fewer staff. SMEs in the arts, entertainment and recreation sector are most likely to say that generative AI helped compensate for a worker shortage (45.1%) while SMEs in the information and communication sector are most likely to say the same for a skills shortage (56.2%).

There is some untapped potential for generative AI to help address labour shortages and skill gaps

Approximately two‑thirds of SMEs experiencing a skill gap or worker shortage in the last two years do not use generative AI. 41.7% of those experiencing a skill gap and not currently using generative AI said that they thought that generative AI could help their company compensate for the gap, while the figure was 23.8% for those experiencing a worker shortage. While the question was hypothetical, these percentages are similar to the percentages of generative AI users reporting the same actual benefits. This suggests some untapped potential for generative AI to help address labour shortages and skill gaps. Notably, among companies experiencing a skill gap or worker shortage and reporting generative AI could help compensate for these issues, the most common barrier to use is concerns regarding copyright, legal or regulatory issues, followed by lack of skills among employees. Barriers are discussed in more detail in Chapter 4.

Generative AI reduces staff workload

Generative AI helps reduce the staff workload of SMEs. 32.7% report a decreased workload while 11.8% report an increase (Figure 3.5). In cases of overwork, employees (or the entrepreneurs themselves in the case of one‑person businesses) may welcome technology that reduces their workload (e.g. by automating tasks and generating efficiencies), although the survey does not capture these sentiments.3 SMEs that say that generative AI saves them money are more likely to say that workload decreased (36.2%) compared to those that say it does not save them money (30.0%).4 There is no such link with improved employee performance, suggesting that other efficiency gains (unrelated to employee performance) are at play. One potential explanation (which cannot be tested using the data) is that these savings derive from reduced overtime.

50.7% of SMEs in Japan report that generative AI decreased workload, well above the percentages observed in other countries. SMEs in the United Kingdom and Germany report the highest increase in workload, although there is less variation by country compared to the decrease. SMEs in the financial sector are most likely to report a reduction in staff workload due to generative AI (42.1%), while SMEs in agriculture and education are most likely to report an increase. 38.5% of SMEs in agriculture and 29.6% in education say that generative AI increased workload, more than double the percentages in any other sector.

One‑person businesses are most likely to report that generative AI has decreased (their own) workload (36.4%), which could be because they have more control over their working hours compared to employees. Mid-size SMEs are most likely to report an increase (13.6%), although differences by size are not statistically significant.

Figure 3.4. One‑person businesses are most likely to report that generative AI has decreased workload

Percentage of SMEs that report generative AI decreased, had no effect or increased workload, by company size

Note: SMEs using generative AI were asked: “Can you tell me whether the use of generative AI within your company has increased or decreased or had no effect on staff workload/your workload?"

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

Generative AI reduces SMEs’ reliance on external contractors

14.3% of SMEs report that generative AI reduced their reliance on external contractors (Figure 3.5). It could be that generative AI enables SMEs to perform tasks that they previously would have outsourced, given limited skills or resources in house. Indeed, SMEs that say that generative AI enables them to perform new tasks are 50% more likely to say that their reliance on external contractors decreased (18.2%) compared to those that say it does not enable to perform new tasks (12.4%).5 On the other hand, 6.3% of SMEs say that generative AI has increased their reliance on external contractors, which could be because some SMEs contract services to get the most out of this new technology (for instance, to build a custom platform for the company or seeking advice on how to comply with regulations). SMEs using generative AI for complex tasks are more than twice as likely to say that their reliance on external contractors had increased (11.7%), compared to those using it for simple tasks (4.5%), which supports this hypothesis.

Figure 3.5. Generative AI reduces SMEs’ reliance on external contractors and staff workload

% of SMEs that report generative AI decreased, had no effect or increased each aspect of their staffing needs

Note: SMEs using generative AI were asked: “I’m going to list some aspects of a company’s staffing needs. For each of these, can you tell me whether the use of generative AI within your company has increased or decreased or had no effect on this aspect of your company’s staffing needs?”

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

SMEs in Canada, the United Kingdom and Japan are the most likely to say generative AI reduced their reliance on external contractors (all above 17%), and SMEs in Korea, Ireland and the United Kingdom are most likely to report an increase (all above 8%). As before, the divide in the UK results could signal heterogeneous impacts across UK SMEs or a greater readiness among UK SMEs to adjust contracting arrangements. Reliance on external contractors is most likely to increase for SMEs in the agriculture, forestry and fishing sector (15.7%) and is most likely to decrease for SMEs across a handful of service sectors: real estate; arts, entertainment and recreation; education; professional, scientific and technical activities; and information and communication (all above 17%). One possible interpretation is that generative AI enables SMEs in service sectors to do more in-house while SMEs in agriculture need more external support to use generative AI.

Just as one‑person businesses are the most likely to say that generative AI reduced overall staffing needs, they are also the most likely to say that it reduced reliance on external contractors (16.5%). Reliance on external contractors increased the most for mid-size SMEs with 10 to 49 employees (7.4%).

Yet generative AI has little effect on overall staff need

Most SMEs (83.0%) using generative AI report that it had no effect on the overall number of staff their company needed (Figure 3.6). Among those who report an effect, more SMEs report that generative AI decreased the overall staff need (9.1%) than increased it (5.5%). A decrease in staff needs could be interpreted as a reduction in labour need due to generative AI’s capacity to automate tasks and thereby substitute for labour. An increase in staff needs could be due to performance improvements: SMEs want more workers because each worker is more productive when aided by AI and/or because the resulting productivity gains induce greater demand in the product market, in turn increasing demand for labour (i.e. the productivity effect). SMEs were not asked about the size of the effects (positive or negative), making it impossible to calculate a net effect. While more SMEs report a decrease than an increase, it was more common to report an impact on workload or on reliance on external contractors, suggesting that SMEs prefer to wait before making internal staffing adjustments. The OECD case studies on AI (Milanez, 2023[3]) captured instances where companies in the manufacturing and financial sectors relied on slowed hiring and attrition rather than immediate redundancies to manage diminished need for labour due to AI.

Figure 3.6. Most SMEs report generative AI has had no effect on overall staff need

Percentage of SMEs that report generative AI decreased, had no effect or increased the overall number of staff needed, by country

Note: SMEs using generative AI were asked: “Can you tell me whether the use of generative AI within your company has increased or decreased or had no effect on the overall number of staff your company needs?”

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

There is some link between how generative AI is used within the company and the impact on overall staff needs. SMEs using generative AI mostly for recurring tasks are over 50% more likely to say that staff needs decreased compared to those using it for one‑off tasks. SMEs using generative AI mostly for complex tasks are almost twice as likely to say that staff needs have increased compared to those using it for simple tasks. Whether the tasks are considered core to the business or not makes little difference for overall staff needs.

Higher revenue is found to be associated with lower staff needs.6 In other words, SMEs reporting that generative AI increased revenue are more likely to say that overall staff needs decreased (13.3% vs. 7.5%) and less likely to say that overall staff needs increased (4.1% vs. 6.0%). One interpretation is that generative AI enables SMEs to scale up operations without increasing staff needs (even decreasing them).

Furthermore, improved employee performance does not appear to be linked to the impact on staff needs, undermining the idea that labour-driven productivity gains are driving up staff needs (through the productivity effect or otherwise) or driving down staff needs (through efficiencies). While theoretical models of the labour market impact of AI say that it is by boosting productivity that AI can drive up demand for labour and potentially counter the displacement effect (OECD, 2023[4]), there is no sign of this mechanism in the survey results. Fewer SMEs report an increase in staff needs than report a decrease, and where SMEs do report an increase, this does not appear to be because they are experiencing improved employee performance. Other studies (discussed in Box 2.3) have estimated time savings associated with the use of generative AI at work at between 2.8% and 5.4% of users’ work hours. If SMEs interviewed as part of this survey experience the same, this may be too small to influence staff needs.

SMEs in the UK report the biggest change in staff needs due to generative AI: they are more likely than SMEs in any other country to report either a decrease (18.3%) or an increase (8.6%). SMEs in Japan, Germany and Korea are the least likely to report any change, either an increase or a decrease. The divide in the UK results could signal heterogeneous impacts across UK SMEs or a greater readiness among UK SMEs to adjust their staffing needs (which could be a function of the flexibility of the UK labour market). The sectors where SMEs are most likely to report a decrease in staff needs due to generative AI are administrative and support service activities, real estate activities, and information and communication (all 14‑15%), while increases are most common in the education sector (17.1%).

Midsize SMEs with between 10 and 49 employees are least likely to say that their overall staff needs have decreased due to generative AI (6.5%) while one‑person businesses are the most likely to say this (10.4%). Possible interpretations for the latter result are that these businesses previously had two workers which had reduced to one by the time of the survey or that the business previously had an unmet staffing need which generative AI had eliminated.

SMEs associate generative AI with increased skill needs

19.7% of SMEs report that generative AI increased skill needs (i.e. an increased need for highly skilled workers relative to lower-skilled workers), twice as many as reported that it decreased them (9.4%) (Figure 3.7). Increasing skill needs dominate in all countries except for Ireland, where the two opposing effects are roughly equal (18‑19%).

Figure 3.7. Twice as many SMEs associate generative AI with increased skill needs as with decreased skill needs

Note: SMEs using generative AI were asked: “I’m going to list some aspects of a company’s staffing needs. For each of these, can you tell me whether the use of generative AI within your company has increased, decreased, or had no effect on this aspect? The number of highly skilled staff your company needs; The number of lower skilled staff your company needs". An SME was considered to be experiencing increasing skill needs (or decreasing skill needs, for the opposing case) due to generative AI if it reported an increase in the number of highly skilled staff (or lower-skilled staff) needed without reporting a simultaneous increase in the number of lower-skilled staff (or highly skilled) needed; or if it reported a decrease in the number of lower-skilled staff (or highly skilled) needed without reporting any effect on the number of highly skilled staff (or lower-skilled staff) needed. 38 out of 2,085 responses marked as “Don’t know” or “Not applicable” were excluded from the analysis.

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

These results explore the idea that generative AI changes the relative need for high- and low-skilled workers,7 by modifying the complexity of tasks within an occupation or by compositional shifts across occupations (i.e. changes in the share of jobs requiring a certain level of skills). For instance, a relative increase in the need for higher-skilled workers could be due to generative AI automating simple tasks (consistent with the findings in Chapter 2), to the complementarity of more advanced skills to the use of generative AI, or to generative AI causing jobs with high skill requirements to grow faster than jobs with low skill requirements (e.g. if increased consumer demand induces labour demand in such a manner8) – or indeed a combination of these factors. Inversely, a relative increase in the need for lower-skilled workers could reflect the automation of complex tasks (e.g. a deskilling within occupations9), the complementarity of skills associated with occupations classified as low skills (e.g. manual skills), or generative AI causing jobs with low skill requirements to grow faster than jobs with high skill requirements.

The association between generative AI and increasing skill needs is consistent with the findings of the OECD AI survey of employers (Lane, Williams and Broecke, 2023[5]), in which more than 50% of firms said that AI had made it more important to have highly educated workers.10

Generative AI increases the importance of data analysis and interpretation skills

Generative AI has increased, on net, the importance of all skills, but particularly those of data analysis and interpretation skills (46.4%) and creativity and innovation skills (41.9%) (Figure 3.8). This is in line with findings from Green (2024[6]) showing that some of the skills that experience the largest increase in demand in occupations exposed to AI are related to originality and digital office tools such as Microsoft Excel. Decreases in importance of skills, while non-negligible, are reported by a minority of respondents, with the largest share of decrease responses associated with critical thinking and problem-solving skills, followed by clerical and administration skills. Other results from Green (Green, 2024[6]) show that AI exposure is associated with declining demand for general skills commonly used in office settings, as word processing, basic computer programming skills, as well as administrative and clerical tasks. While the survey did not ask about digital security skills specifically, the OECD 2025 D4SME survey highlights the importance of these skills for ensuring safe and responsible digital uptake by SMEs (Bianchini and Lasheras Sancho, 2025[7]).

Figure 3.8. Generative AI has made data analysis and interpretation skills more important, along with other skills

Percentage of SMEs that report generative AI made each skill more or less important

Note: Respondents were asked: “I’m going to read to you a number of skills. For each of them, can you tell me whether you think generative AI has made the skill more important, less important or whether it has made no difference to the importance of the skill for workers in your industry?” Results include users’ and non-users’ responses.

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

References

[9] Bessen, J. (2015), Learning by Doing: The Real Connection between Innovation, Wages, and Wealth, Yale University Press.

[7] 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.

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

[6] Green, A. (2024), “Artificial intelligence and the changing demand for skills in the labour market”, OECD Artificial Intelligence Papers, No. 14, OECD Publishing, Paris, https://doi.org/10.1787/88684e36-en.

[8] Jiang, W. et al. (2025), AI and the extended workday: productivity, contracting efficiency, and distribution of rents, National Bureau of Economic Research, https://www.nber.org/papers/w33536.

[5] 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] Milanez, A. (2023), “The impact of AI on the workplace: Evidence from OECD case studies of AI implementation”, OECD Social, Employment and Migration Working Papers, No. 289, OECD Publishing, Paris, https://doi.org/10.1787/2247ce58-en.

[2] 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.

[4] OECD (2023), OECD Employment Outlook 2023: Artificial Intelligence and the Labour Market, OECD Publishing, Paris, https://doi.org/10.1787/08785bba-en.

Notes

← 1.  The causality is unlikely to run in the opposite direction (i.e. a scenario in which SMEs experiencing shortages would have used generative AI to eliminate the shortage and therefore would have no shortage to report at the time of interview) due to the wording of the question, which asked respondents if their SME had experienced a shortage at any point in the previous two years. The start of the reference period thus coincides with the release of ChatGPT in November 2022.

← 2.  Logistic regression is used to examine the relationship between reported benefits (as included in Figure 2.6) and generative AI helping to compensate for labour shortages and skill gaps, controlling for country, company size, sector and whether generative AI is used in a core activity. SMEs that say generative AI helps them increase revenue and perform new tasks are also more likely to say that it helps address skill gaps, but the links are weaker. SMEs that say generative AI saves money are more likely (31.5%) to say it helped address worker shortages compared to those that do not (19.5%).

← 3.  While less commonly reported, an increase in workload could signal that generative AI is creating additional tasks for workers or that productivity gains resulting from automating tasks or improving operations accrue to the employer at the expense of the worker. A recent working paper (Jiang et al., 2025[8]) demonstrates the latter dynamic by showing that occupational exposure to generative AI is associated with an overall increase in weekly working hours (which differs from the overall pattern observed in this survey), and that this effect is even more pronounced where AI is complementary to workers and where the labour market is competitive. In other words, when generative AI makes each working hour more productive and workers’ bargaining power is low, employers extract most of the technology-enabled productivity gains by demanding more working hours (i.e. contrasting from the case in which workers’ bargaining power enables them to reduce their hours thanks to the same productivity gains).

← 4.  Ordered logistic regression is used to examine the relationship, controlling for country, company size and sector. No other benefits are statistically significant predictors.

← 5.  Ordered logistic regression is used to examine the relationship, controlling for country, company size and sector. No other benefits are statistically significant predictors.

← 6.  Ordered logistic regression is used to examine the relationship between reported benefits (as included in Figure 2.6) and impact on staff need, controlling for country, company size and sector. No other benefits are statistically significant predictors of impact on staff need.

← 7.  While labour economists often use educational attainment, wage levels or occupation classification as a proxy for skill level (mostly due to availability of data), it is not clear that respondents would have had the same concepts in mind when answering. In the survey, no definition of “highly skilled” or “lower skilled” staff was provided to respondents, so interpretation may have varied across respondents, including e.g. by sector, by country, by the respondent’s personal characteristics.

← 8.  For instance, the introduction of ATMs in the banking industry in the 1970s produced cost savings and increased consumer demand, leading banks to open additional branches, increasing the number of bank tellers needed, shifting their tasks away from cash-handling and towards marketing and customer relations, and increasing the proportion of bank tellers with a college degree (Bessen, 2015[9]).

← 9.  The OECD case studies on AI (Milanez, 2023[3]) captured some instances of deskilling, including one in which steel rods were straightened by a machine controlled by AI software, while workers only had to start the machine and load and unload it, removing the weeks or months of experience that were previously required to operate the machine.

← 10.  Although direct comparison is not advisable because companies were not asked about whether AI had made the same skills less important.