Skills in the AI age

AI adoption by firms can help reduce costs and boost productivity, supporting increased competitiveness. However, it also has far-reaching societal implications for individuals as consumers and workers.

AI adoption by firms can transform the approach to operational efficiency, decision-making processes and consumer engagement. AI technologies, including machine learning, natural language processing, computer vision and predictive analytics, enable businesses to automate routine tasks, optimise resource allocation and enhance productivity (Brynjolfsson, 2017[3]; Calvino and Fontanelli, 2023[4]; Filippucci et al., 2024[5]). For example, AI-powered predictive maintenance is transforming the manufacturing industry by reducing downtime, lowering maintenance costs and enhancing operational efficiency (Lalwani, 2025[6]).

AI can also help companies improve product or service quality (OECD, 2023[2]), enabling workers to benefit simultaneously through improved job quality and satisfaction. Indeed, AI could reduce or eliminate dangerous or tedious tasks, and create more complex and interesting ones instead. It can boost worker engagement, giving them greater autonomy and even improving their mental health. Some workers may also benefit from higher wages if there is a higher demand for specific skills on the job market. Additionally, AI has proven to be instrumental in enhancing workplace wellness and ergonomics (Fiegler-Rudol et al., 2025[7]). AI technologies can automate physically demanding and repetitive tasks, reducing the risk of burnout and improving employee well-being (García-Madurga et al., 2024[8]).

Recent evidence also shows that GenAI tools could significantly enhance productivity, both at individual and organisational levels, with relevant aggregate implications (Calvino, Reijerink and Samek, 2025[9]).

In OECD Member countries with available data, businesses have significantly increased their uptake of AI in recent years. As measured in ICT usage surveys, use of AI in these businesses grew from around 7% to 20% between 2021 and 2025 (Figure 1). This increase is likely driven by the adoption of “general-purpose” GenAI tools, such as ChatGPT and Copilot, which were made available as of 2022.

AI uptake in firms differs considerably by age, sector and size. Younger firms, including start-ups, have a strong tendency to use AI. Indeed, start-ups are often at the forefront of radical innovations, especially during the emergence of new technological paradigms such as AI. As a result, the propensity of young firms to adopt AI often surpasses that of older firms (Calvino and Fontanelli, 2023[4]).

A recent OECD taxonomy also reveals significant heterogeneity across sectors with respect to AI human capital, AI innovation, and exposure to and use of AI (Calvino and Fontanelli, 2023[4]). Sectors such as information technology (IT) services score high along all the dimensions considered. Others, such as pharmaceuticals, exhibit considerably more heterogeneity (high AI human capital but low AI innovation).

Without targeted and effective support to develop the skills needed to work with AI, workers and businesses will not be able to seize its benefits, while disparities in job opportunities, job quality, wages and productivity may widen. With respect to size breakdowns, larger firms with greater financial and technical resources appear more likely to adopt AI (Calvino and Fontanelli, 2023[4]). SMEs often report barriers related to cost, expertise and infrastructure among the reasons for not adopting AI technologies. Lack of skills is the second most common barrier to adoption of AI reported by firms in finance and manufacturing (after costs) in the OECD AI surveys of employers and workers (OECD, 2023[2]). Similarly, half of SMEs not using GenAI tools report lack of skills as a barrier (OECD, 2025[11]). Other major deterrents include concerns about legal and regulatory issues (e.g. concerns about ineffective protection and enforcement of copyright and other intellectual property), data privacy, and manipulated and deliberately misleading content online (Bianchini and Lasheras Sancho, 2025[12]).

AI is increasingly recognised as a transformative force in labour markets, driving significant shifts in job creation, skills demand and overall workforce dynamics. While concerns about job displacement persist, research suggests that AI also creates numerous opportunities for employment growth and economic resilience. One study, for example, argues that AI technologies can lead to net job gains by complementing human labour rather than substituting it, thereby creating new tasks and industries (OECD, 2023[2]).

Building on research by Lane and Saint-Martin (2021[13]) within the OECD’s Artificial Intelligence in Work, Innovation, Productivity and Skills (AI-WIPS) programme, OECD (2023[2]) identifies different types of impacts of AI on employment. In one scenario, AI affects labour demand by automating certain tasks, complementing humans in completing existing tasks but also creating new ones. When AI automates tasks previously done by humans, it may replace human labour. In parallel, AI may create new tasks that require human labour, leading workers to undertake alternative tasks compared to previous ones. Finally, AI may complement workers, uniformly improving human productivity in all assigned tasks. In this scenario, AI neither replaces workers in the completion of tasks nor creates new tasks; it allows workers to perform the same tasks more efficiently. The three channels are not mutually exclusive, and all may be present at the same time.

The implications for labour demand depend on which effects dominate. The creation of new tasks should lead to increases in labour demand. In particular, the creation of new tasks results in a reinstatement effect that creates new jobs and leads to more employment (OECD, 2023[2]). These jobs can come from anywhere, but literature usually links them to the operation of the new technologies themselves (Acemoglu and Restrepo, 2018[14]; Autor et al., 2022[15]).

These new jobs likely require skills to develop, train or maintain AI systems. However, they also include new jobs for the larger set of workers who will only interact with AI applications. The automation channel, or the replacement of tasks previously performed by human labour, is ambiguous with respect to whether it increases or decreases demand for labour. It results in a displacement effect, when labour demand decreases as AI replaces human labour (OECD, 2023[2]).

Finally, in addition to the displacement effect, the automation channel can give rise to a productivity effect. This can result in increases in labour demand for tasks or jobs not automated by AI, such as packers and forklift operators. The productivity effect neither arises (directly) from the automation of tasks, nor the ability to perform tasks more efficiently. Rather, it stems from the induced demand for tasks or jobs generated from the cost savings of automation. Whether the productivity effect dominates the displacement effect, and therefore whether automation increases or decreases labour demand, is the core question facing the future of labour and AI (OECD, 2023[2]).

Using data from labour force surveys, Lane (2024[16]) shows that on average across OECD Member countries, “IT professionals”, “Business professionals”, “Managers”, “Chief executives” and “Science and engineering professionals” are the occupations most exposed to AI (Figure 2).1 This follows intuitively from the technological advances in which AI has made the most progress (OECD, 2023[2]). In contrast, “Cleaners, helpers”, “Agriculture, forestry and fishery workers”, “Food preparation assistants”, “Labourers”, “Refuse workers and other elementary workers” are the least exposed to AI. Therefore, high-skill, white collar occupations are the ones most exposed to recent progress in AI.

Recent advances in AI may complement rather than replace human labour. The verification of displacement and reinstatement scenarios requires putting findings in parallel with occupations that are subject to automation. Most AI exposure indicators that are used to estimate the effect of AI on employment measure progress in the capabilities of AI as they relate to abilities needed in occupations. However, progress in AI capabilities is not equivalent to the probability of automation (OECD, 2023[2]). Recent progress in AI may complement human labour rather than automate it. More importantly, AI is just one of the many advanced ICT technologies that can lead to automation of tasks previously done by human labour.

Aggregate changes in employment will ultimately depend on all sources of automation. The predicted effect of AI on certain groups may differ unless all sources of automation are carefully considered. Recent evidence shows that, in some cases, the creation of new employment opportunities has more than offset job displacements. However, this dynamic has unfolded unevenly across regions within OECD Member countries. Moreover, the new jobs have not necessarily benefited displaced workers (OECD, 2024[17]).

In an OECD study to reflect advancements in automation technologies, Lassébie and Quintini (2022[18]) re-evaluates how various occupations are exposed to automation. This assessment, which draws on a unique survey measuring the automatability of around 100 distinct skills and abilities, was developed and completed with input from experts across multiple AI research domains. The survey considers both AI and also other automation technologies (such as robotics) that are increasingly integrated with AI, allowing these technologies to complement one another. The study concludes that high-skilled occupations face the lowest overall risk of automation,2 although the full impact of the widespread availability of GenAI tools is not entirely captured given the period covered.

While AI has made some high-skill job requirements more susceptible to automation, many critical skills in these roles remain difficult to automate. As a result, despite greater exposure to recent AI advancements, high-skill jobs generally continue to be among the least vulnerable. In contrast, low- and middle-skill roles, particularly in “Construction and Extraction”, “Farming, Fishing and forestry”, and to a lesser extent, “Production” and “Transportation and Material Moving”, are more likely to be at high risk of automation. In general, the least at-risk jobs are those in “Community and Social Service”, as well as in “Management” (Figure 3). Overall, jobs requiring non-routine cognitive, social and creative skills are less susceptible to automation, while routine and low-skilled jobs are at higher risk (OECD, 2023[2]).

More specifically, recent OECD evidence also shows that exposure to GenAI tools varies widely across occupations, reflecting differences in how much job tasks can be accelerated by current and emerging AI technologies. Occupations involving cognitive, non-routine tasks – such as programming, translating, interpreting and other highly skilled professional work – tend to exhibit higher exposure. In these occupations, GenAI could perform a significant share of tasks at least twice as fast today or in the near future. By contrast, roles reliant on physical or manual tasks, such as cleaning or equipment operation, show minimal exposure. Overall, about a quarter of workers across the OECD were exposed to GenAI (defined as at least 20% of their tasks being amenable to AI assistance) in 2022-2024. This share is projected to grow substantially as GenAI tools become more integrated into workplace software and systems (OECD, 2024[17]).