Key Details
The IMF working paper Aging Economies and AI Adoption: Firm-Level Evidence from the World Bank Enterprise Surveys examines whether firms in older economies are more likely to adopt labour-saving technologies.
Study sample: 89,380 firms across 144 countries (including India), surveyed between 2022 and 2025.
Adoption identified: 1,656 firms, or 1.9%, reported process innovations involving AI, automation or robotics. Only 221 firms, or 0.2%, explicitly mentioned AI or machine learning.
Demographic effect: A 10-percentage-point increase in the old-age dependency ratio was associated with a 0.6-percentage-point rise in process adoption after controlling for income, internet access, firm size, sector, region and survey year.
Where the effect is strongest: Broad AI and automation adoption is concentrated among large firms, manufacturers and firms in developing economies. Explicit software-based AI is more common in services.
India in the study: India contributes 9,111 firms surveyed in 2022, the largest national sample. The paper does not report a separate India estimate.
South Asian context: The region recorded 0.16% process adoption and 0.04% product adoption, although these figures partly rely on surveys conducted before the widespread uptake of generative AI.
Aging Changes Both How Firms Produce and What They Sell
The paper identifies two distinct channels.
Through the process channel, firms respond to labour scarcity by introducing AI, automation or robotics into production. This relationship is strongest in manufacturing, where machines can more readily substitute for repetitive physical work.
Through the product channel, firms develop AI-enabled products or services for customers facing labour constraints of their own. This effect appears in both manufacturing and services.
Only 82 firms were classified as adopters through both channels, suggesting that aging affects two largely separate groups: businesses automating their own operations and businesses creating labour-saving solutions for the market.
Hardware Automation and Software AI Follow Different Patterns
When AI, automation and robotics are combined, the demographic effect is concentrated in manufacturing. When the measure is narrowed to firms explicitly mentioning AI or machine learning, service firms are more likely to adopt.
The distinction reflects different technology profiles:
Manufacturing makes greater use of machinery, robotics and automated production lines.
Services are more exposed to software-based applications such as language models, predictive analytics, sentiment analysis and automated customer interactions.
Policy directed towards “AI adoption” must therefore distinguish between the capital requirements of industrial automation and the data, software and skills required for service-sector AI.
Large Firms Are Better Placed to Respond
The effect is strongest among firms with 100 or more employees. Small firms show a considerably weaker response, while the result for medium-sized firms is not statistically significant in the main process-adoption analysis.
Automation requires investment in equipment, software integration, organisational redesign and worker training. Demographic pressure may create the incentive to adopt technology without providing firms the capacity to do so.
This produces a risk that aging could widen productivity differences: large businesses automate, while smaller firms continue facing labour and skill shortages.
The Relationship Is Evident in Developing Economies
The demographic effect is statistically significant in both low-income and middle-income developing economies. The estimate for advanced economies is less precise, partly because their old-age dependency ratios are already high and vary less within that group.
Descriptive results reinforce the broad pattern. Process adoption was highest in several older economies, including Denmark at 15.5%, Switzerland at 10.6% and Belgium at 9.4%. China recorded 7.9%.
South Asia reported only 0.16% process adoption and 0.04% product adoption, the lowest regional rates in the study. These are not current India estimates: India’s underlying survey was conducted in 2022, before widespread business use of generative AI.
What Is the Old-Age Dependency Ratio?
The old-age dependency ratio is the number of people aged 65 and above relative to every 100 people of working age. A higher ratio indicates a smaller potential workforce relative to the elderly population. The paper’s proposition is that this shift makes workers and skills scarcer, strengthening the business case for labour-saving technology.
Firms Report Greater Labour Constraints in Older Economies
The paper tests the proposed mechanism rather than relying only on the correlation between aging and adoption.
Firms in older economies are more likely to identify an inadequately educated workforce as their biggest business obstacle and to report more severe skill gaps. The demographic effect is also stronger among businesses whose labour costs account for a larger share of sales.
These findings are consistent with a sequence in which aging tightens labour availability and firms respond by adopting technology, although reports of inadequate skills may also reflect education quality and skill mismatches.
The adoption result remains positive across alternative classifiers, lagged demographic measures, different regression approaches and an instrumental-variable exercise based on countries’ earlier demographic structures. This strengthens the association, but the working paper remains research by its author rather than an official IMF policy position.
Policy Relevance
For India, the paper highlights a distributional challenge before aging becomes a stronger economic driver.
Prepare smaller firms: Shared technology facilities, affordable finance, vendor-neutral advisory support and workforce training could reduce the fixed costs that currently favour larger businesses.
Distinguish AI from automation: Manufacturing policy must account for equipment and robotics, while service-sector policy should address software, data access and AI-enabled business processes.
Track adoption directly: Future enterprise surveys should ask firms which technologies they use, for what functions and with what effects on employment, productivity and wages.
Anticipate labour-market change: Automation may help maintain output when workers are scarce, but it can also concentrate displacement in routine occupations. Reskilling must develop alongside adoption incentives.
Examine regional demographics: National averages can conceal States and industries experiencing different combinations of aging, migration and skill shortages. Technology and workforce planning should reflect those differences.
Follow the Full Paper Here: Aging Economies and AI Adoption: Firm-Level Evidence from the World Bank Enterprise Surveys