A boost from Artifical Intelligence in ageing societies?

Reading Time: 5 minutes

Demographic headwinds are set to weaken economic growth in OECD countries over the coming decades. At the same time, artificial intelligence (AI) provides opportunities for productivity gains, potentially alleviating labour shortages and boosting economic growth.

By Christophe André and Matthias Schief, OECD Economics Department

Population ageing and AI adoption across a growing number of economic activities are affecting growth prospects in opposite directions. While population ageing could reduce real per capita income growth by up to 0.8 percentage points per year in some OECD countries over the coming decades (André et al., 2024), AI could boost it by up to one percentage point per year over the next ten years, assuming an adoption speed similar to that of previous ICT technologies, such as computers and the internet (Filippucci et al., 2026). However, little is known about the interactions between these two megatrends. 

In a new paper (André and Schief, 2026), we provide early insights on how exposure to AI varies over the life cycle and what this may imply for AI deployment in ageing societies. In recent years, several studies have assessed occupation-level and task-level exposure to AI in the United States (Felten et al., 2021; Eloundou et al., 2024). Other studies (Georgieff and Hyee, 2021; Lewandowski, Madoń and Park, 2025) have used the OECD Survey of Adult Skills (PIAAC) to derive AI-exposure indicators for broad sets of countries. The PIAAC dataset contains rich information on respondents, which allows relating AI exposure to demographic and socio-economic characteristics, including age. The detailed description of tasks performed by respondents also allows distinguishing jobs exposed to automation, where AI may directly replace workers, from jobs exposed to augmentation, where AI complements workers’ abilities to make them more productive.   

We construct a new index of AI exposure at the individual level, using the latest vintage of PIAAC data. We distinguish between individuals that are exposed to AI (types II and III) and those that are not (type I) using information about two broad aspects of their job: the presence of a strong physical component and the daily use of computers or other electronic devices (Table 1). Responses indicating the degree to which work is proceduralised versus context dependent are used to differentiate between exposure to automation versus augmentation. Proceduralised work refers to activities whose essential functions can be codified into relatively stable and well-defined sequences of actions performed within a relatively rigid production process (for example, standard insurance claim processing or legal document drafting). By contrast, context dependent work is not fully specified ex ante, and the sequence of tasks must often be adapted in real time to changing circumstances, involving human judgement (for example, project management under changing constraints).

Table 1. A classification of AI exposure

	Proceduralised work	Non-proceduralised work
Strong physical component or no computer use	Not exposed to AI (Type I)	Not exposed to AI (Type I)
No strong physical component and daily computer use	Exposed to automation (type II)	Exposed to augmentation (type III)

While our classification has the merit of parsimony and transparency, it is also somewhat coarse, as it relies on a limited number of job characteristics and classifies all workers in only three categories, which cannot fully capture the wide variety of occupations in the economy. Nevertheless, our classification correlates strongly at the occupation level in the United States with the respective exposure measures in Felten et al. (2021) and in Eloundou et al. (2024), which are derived from more detailed information on work tasks and AI capabilities. The distribution of automation and augmentation exposure within industries is also consistent with early evidence on adoption and employment patterns across economic sectors.

Our overall (automation and augmentation) exposure to AI indicator exhibits an inverted U-shaped pattern across age groups (Figure 1, left panel). However, this pattern is less pronounced when controlling for education, occupation and country, pointing to the role of differences across cohorts in education and occupational choices in explaining differences in AI exposure between age groups. Exposure to automation is highest in younger age groups and declines rapidly with age, with middle-aged and older workers less exposed to substitution as their experience tends to complement AI (Figure 1, right panel). Exposure to automation remains similar when controlling for education, occupation and country, suggesting only minor influence of cohort effects. The strong exposure of youth to automation is in line with the weak entry-job creation in AI-exposed sectors observed recently in several OECD countries, even though evidence on the role of AI remains mixed.

With AI more likely to complement middle-aged and older workers than replace them, ageing societies may receive a boost from AI. The technology could help lengthen working lives, if it improves job quality by reducing repetitive tasks and facilitating skills updating. This would alleviate potential labour shortages and is key to preserving growth potential and fiscal sustainability (Koutsogeorgopoulou and Morgavi, 2025). Productivity gains would also contribute to mitigating the demographic drag on GDP per capita.

Nevertheless, these potential benefits do not come without challenges. As a general-purpose technology, AI is bound to be disruptive, imposing significant changes in work processes. Even workers with skills that complement AI may need to change jobs in response to changing demands for their competences and need reskilling and upskilling to efficiently use AI. This may prove more difficult in ageing societies, as job mobility and occupational transitions fall sharply with age (Figure 2) and older workers tend to receive less training than younger ones, which could lead to skill obsolescence as technological change accelerates. In addition, the intergenerational transmission of knowledge may be hampered if a shortage of entry jobs reduces opportunities for young people to learn on the job.  

Furthermore, business dynamism and innovation may weaken in ageing societies, which could slow the diffusion of AI and hold back productivity gains. Weakening workforce growth has been linked to declining firm entry rates in the United States (Karahan et al., 2024), but studies on other OECD countries remain scarce.

To sum up, AI may offer opportunities for ageing societies to at least partly offset the demographic drag on economic growth. However, this will require overcoming major challenges in lifelong learning, labour reallocation and business dynamism and innovation. Furthermore, while demographic headwinds will persist for decades, it is unclear whether AI will lead to a level shift in productivity or to persistently higher growth, for example through continuous development in robotics or fast progress towards AGI (Artificial General Intelligence).

References

André, C. and M. Schief (2026), “A boost from AI in ageing societies? Early insights”, OECD Economics Department Working Paper, No. 1870, OECD Publishing, Paris.

André, C., P. Gal and M. Schief (2024), “Enhancing productivity and growth in an ageing society: Key mechanisms and policy options”, OECD Economics Department Working Papers, No. 1807, OECD Publishing, Paris, https://doi.org/10.1787/605b0787-en.

Eloundou, T., S. Manning, P. Mishkin and D. Rock (2024), “GPTs are GPTs: Labour market impact potential of LLMs”, Science, 384 (6702), 1306-1308.

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, 42(12), 2195-2217.

Filippucci, F., P. Gal, K. Laengle, M. Schief and M.A. Yildirim (2026), “AI meets trade: Global linkages and the cross-country distribution of the gains from AI”, OECD Artificial Intelligence Papers, No. 57, OECD Publishing, Paris, https://doi.org/10.1787/13081644-en.

Georgieff, A. and R. Hyee (2021), “Artificial intelligence and employment : New cross-country evidence”, OECD Social, Employment and Migration Working Papers, No. 265, OECD Publishing, Paris, https://doi.org/10.1787/c2c1d276-en.

Karahan, F., B. Pugsley and A. Şahin (2024), “Demographic Origins of the start-up Deficit”, American Economic Review, 114(7), 1986-2023, https://doi.org/10.1257/aer.20210362.

Lewandowski, P., K. Madoń and A. Park (2025), “Workers’ exposure to AI across development”, IBS Working Paper, 02/2025, Institute for Structural Research (IBS), Warsaw, https://ibs.org.pl/wp-content/uploads/2025/03/Workers_AI_exposure_across_development_IBS_WP_202502.pdf.