The AI Development Divide: Will Artificial Intelligence Democratize Economic Opportunity or Concentrate It?

Two plausible futures

Every major general-purpose technology has raised the same question: will it help poorer economies catch up, or will it let richer ones pull further ahead? The answer has historically been both, depending on who could adopt the technology, how quickly and on what terms.

Artificial intelligence sharpens the question. On one hand, AI tools are distributed largely through software, often at low marginal cost, and can be accessed from a smartphone. A small business owner in Accra, a student in Nairobi and an analyst in Toronto can, in principle, use the same model. On the other hand, the most capable AI systems are built by a small number of firms in a small number of countries, using computing infrastructure, data and talent that are heavily concentrated.

This paper examines both possibilities. It argues that AI's distributional effect is not predetermined by the technology itself. It will be shaped by five conditions: financial inclusion, digital infrastructure, institutional capacity, labour market structure and access to the technology on fair terms.

The case for convergence

Lower costs of expertise

Much of the productivity gap between advanced and emerging economies reflects scarce expertise: doctors, engineers, lawyers, accountants and experienced managers. AI can make some forms of expertise cheaper and more widely available. Evidence from workplace studies suggests that AI tools deliver the largest gains to less experienced workers, effectively transferring some of the knowledge of top performers to everyone else. If that pattern holds across countries, AI could help close skill gaps faster than traditional education and training alone.

Leapfrogging is possible

Emerging economies have leapfrogged before. Mobile money is the clearest example. In Kenya, the spread of M-Pesa gave millions of people access to financial services without the branch networks that defined banking in advanced economies. Research has linked access to mobile money with meaningful reductions in poverty, particularly among female-headed households. Mobile networks allowed countries to skip fixed-line infrastructure; mobile money allowed them to skip conventional banking infrastructure.

AI could enable similar leaps: diagnostic support where specialists are scarce, agricultural advice tailored to local conditions, credit assessment for people with no formal credit history and public services delivered through conversational interfaces in local languages.

Smaller firms gain access to scale

AI tools can give small firms capabilities that once required large teams: marketing, translation, customer service, financial analysis and software development. In economies where small and informal enterprises account for most employment, that could raise productivity broadly rather than concentrating gains in a few large firms.

The case for divergence

Concentration of capability

Building frontier AI requires vast computing resources, specialized talent and large datasets. These are concentrated in a few countries and companies. Emerging economies may become consumers of AI built elsewhere, paying for access and exporting data, while the economic rents from the technology accrue to its owners. Dependence on foreign platforms also creates strategic risks, from pricing power to the possibility that services are restricted or withdrawn.

Adoption lags compound

Historical research on technology diffusion has found that while new technologies now reach poorer countries faster than in the past, the intensity of their use, meaning how deeply they are integrated into production, still lags considerably. Access is not the same as adoption. If advanced economies integrate AI deeply into firms and public institutions while emerging economies use it superficially, the productivity gap could widen even as access appears to equalize.

Pressure on development pathways

Many economies have grown by moving workers from agriculture into manufacturing and, later, into tradable services such as call centres, business process outsourcing and software services. Economists have already documented "premature deindustrialization," in which manufacturing's share of employment peaks earlier and at lower income levels than it did for today's rich countries. AI could add pressure to services-led growth, automating some of the tasks that made outsourcing an entry point into the global economy. Research on automation shows that its effect on workers depends heavily on whether new tasks are created alongside the ones that are automated. If new tasks emerge mainly in advanced economies, emerging economies lose a development ladder without gaining a new one.

Data and language gaps

AI systems perform best on the languages, contexts and populations best represented in their training data. Many African and South Asian languages, local business practices and informal economic activity are underrepresented. Systems that work less well for these populations can reinforce existing disadvantages, for example in credit scoring, hiring or access to public services.

Five conditions that decide the outcome

Whether AI democratizes or concentrates opportunity depends less on the technology than on the conditions into which it is introduced.

Condition	Why it matters	What raises the odds of convergence
Financial inclusion	AI-enabled services need customers who can pay, save and borrow digitally	Interoperable mobile money, digital identity, inclusive credit data
Digital infrastructure	AI depends on connectivity, affordable devices and computing capacity	Broadband expansion, affordable data, regional data centres
Institutional capacity	Governments must regulate, procure and deploy AI effectively	Technical talent in the public sector, clear data governance, sound procurement
Labour markets	Workers need pathways into the new tasks AI creates	Reskilling linked to employer demand, support for small enterprises
Access on fair terms	Dependence on a few foreign providers creates pricing and strategic risk	Open models, local language data, regional partnerships, competitive markets

Financial inclusion as the foundation

Financial inclusion deserves particular attention because it is both a precondition for and a beneficiary of AI. Digital payment histories create the data that allow AI-based credit models to serve people without formal credit records. In turn, better credit assessment can expand lending to small businesses and households excluded from traditional finance. The risk is that the same tools, poorly governed, can entrench bias or enable predatory lending. Inclusion requires both innovation and consumer protection.

Institutions determine whether gains are shared

AI's benefits will depend heavily on the capacity of public institutions to set rules, invest in infrastructure and deploy AI in public services. Countries with stronger institutions will be better able to negotiate fair terms with technology providers, protect citizens' data and direct AI toward priorities such as health, education and agriculture. Building that capacity, including technical talent inside government, may be among the highest-return investments a developing economy can make.

Implications for policymakers and businesses

For governments in emerging economies, the priority is to treat AI readiness as development policy: invest in connectivity, digital identity and payments infrastructure; build public-sector technical capability; support local language datasets; and avoid lock-in to a single provider.

For development finance institutions, AI creates an opportunity to finance the complementary infrastructure, such as data centres, connectivity and digital public goods, that determines whether countries can adopt AI deeply rather than superficially.

For financial institutions and payment networks, emerging markets offer some of the clearest opportunities for AI to expand access to credit, savings and insurance. Responsible deployment, with transparent models and strong consumer protection, is both an ethical requirement and the basis for long-term growth in these markets.

For technology companies, fair pricing, local partnerships and investment in underrepresented languages and contexts are not only matters of corporate responsibility. They determine the size of the market AI can eventually serve.

Conclusion

AI does not have a fixed distributional destiny. The same technology that could give a smallholder farmer expert agricultural advice could also concentrate economic power in a handful of firms and countries. History suggests that technologies spread fastest where complementary conditions are in place: infrastructure, institutions, skills and access to finance.

The question is therefore not whether AI will help emerging economies catch up, but whether the conditions for broad adoption will be built in time. Mobile money showed that emerging economies can design around the constraints of older systems and, in some respects, move ahead of advanced ones. AI offers a similar opportunity, but only for the economies and institutions that prepare for it deliberately.

References

Acemoglu, D., & Restrepo, P. (2019). Automation and new tasks: How technology displaces and reinstates labor. Journal of Economic Perspectives, 33(2), 3–30.
Brynjolfsson, E., Li, D., & Raymond, L. R. (2023). Generative AI at work (NBER Working Paper No. 31161). National Bureau of Economic Research.
Comin, D., & Hobijn, B. (2010). An exploration of technology diffusion. American Economic Review, 100(5), 2031–2059.
Rodrik, D. (2016). Premature deindustrialization. Journal of Economic Growth, 21(1), 1–33.
Suri, T., & Jack, W. (2016). The long-run poverty and gender impacts of mobile money. Science, 354(6317), 1288–1292.