Why Bigger Is Not Always Better for Emerging Markets
Most of the artificial intelligence conversation revolves around ever larger models, data centers the size of small towns, and chatbots that promise to write, code, and reason at superhuman speed. The World Bank is pushing in a different direction. Its chief knowledge officer, Paschal Donohoe, is urging developing countries to build what he calls “small AI”: specialized models trained on narrow datasets for a specific job, designed to run on a basic mobile phone even when connectivity is weak. The argument is that this kind of tool could do more for a farmer in rural India or a clinic worker in East Africa than any frontier model ever will.
The ambition is striking. Donohoe suggests that economic development that once could have taken centuries might now unfold within decades. That is a bold claim, and it deserves scrutiny. Yet the logic behind it is grounded in a practical observation. A small farm does not need a general-purpose genius. It needs soil guidance in its own language, delivered on a device it already owns, at a cost it can afford.
The Infrastructure Gap Nobody Can Ignore
The World Bank’s own annual development report notes that almost a third of rural schools in sub-Saharan Africa lack electricity, and more than two-thirds lack a consistent internet connection. That reality makes cloud-dependent AI a poor fit for much of the world’s population. IMF managing director Kristalina Georgieva made the same point from a different angle, insisting that electricity, digital infrastructure, and public safeguards against bad actors are preconditions for AI to spread. Her comparison to the cellphone is telling. Mobile phones spread faster across the developing world than landlines ever did, precisely because they leapfrogged the wired infrastructure that rich countries had spent a century building. Small AI could follow the same path, but only if someone builds the local models, the language support, and the offline capabilities first.
What the Jobs Question Really Looks Like
Georgieva also warned that around 40 percent of jobs worldwide could be augmented or eliminated by AI, with young workers facing the sharpest risk. Entry-level roles that once served as ladders into professional careers are the ones most exposed to automation. Her blunt remark that no young person wants to be a barista as a professional future captures a real anxiety. Small AI offers a partial answer here, since tools that extend the reach of agronomists, teachers, and health workers can create new kinds of skilled local employment rather than simply replacing existing roles. Whether that happens will depend on investment choices made in the next few years, not on the technology alone.
Where Private Money and Public Purpose Meet
The World Bank says it has mobilized $120 billion in private capital over the past year using guarantees and other financial tools, even as several wealthy nations trimmed their aid budgets. That shift toward private investment makes the small AI vision more plausible, because commercial developers have the incentive to build products people will actually buy. For consumers in emerging markets, the most important question is not whether AI exists in their country but whether it speaks their language, works on the phone already in their pocket, and fits their budget. The companies that solve those problems will likely find some of the largest unmet demand in the technology sector today.
For IntentBuy readers tracking consumer adoption, this story is a reminder that the next wave of AI buyers may look nothing like the early adopters in Silicon Valley. They may be a farmer comparing affordable soil-testing apps, a small shopkeeper evaluating a multilingual assistant, or a parent choosing a low-cost learning tool that works without reliable internet. Watch for the devices, price points, and local-language features that win these buyers over, because that is where tech adoption is likely to surprise us most in the years ahead.
