The World Bank is urging developing countries to begin building «small AI» models that can run on a mobile phone with limited connectivity, arguing the technology could speed up economic development from centuries to decades.
Paschal Donohoe, the World Bank's chief knowledge officer, said artificial intelligence could accelerate development for emerging economies, but only if it is adapted to local needs. His intervention comes as Silicon Valley's largest technology firms concentrate on ever more sophisticated consumer products, a focus that leaves smaller markets on the margins of the AI boom.
The bank's argument is that the most advanced models, which require vast computing power and reliable high-speed internet, are poorly suited to countries where many people access the internet through mobile phones and where connections can be intermittent. Small AI models, by contrast, can operate on a handset with low connectivity, bringing useful tools directly to users rather than depending on distant data centres.
For small island states such as the Commonwealth of Dominica, the appeal is practical. Public services, agriculture, fisheries, tourism and disaster preparedness all depend on timely information, and a model that works on a phone in a village with weak signal could support farmers, fishers and local officials alike. The World Bank's emphasis on local adaptation suggests that tools built elsewhere and simply imported may deliver less than systems designed around the language, data and conditions of the communities they serve.
Donohoe's framing places AI alongside more familiar development priorities rather than above them. The claim is not that technology alone will transform an economy, but that appropriately scaled tools could compress the time it takes to improve services and spread knowledge. That distinction matters for governments weighing scarce resources against competing demands in health, education and infrastructure.
The World Bank's call also raises questions about skills, data and regulation. Building even a small model requires people who understand the technology, reliable local data and a policy environment that protects users. For developing countries, the risk is that the benefits arrive slowly while the costs of dependence on foreign platforms accumulate.
The bank's intervention is likely to be read closely by finance ministries and regional bodies across the Caribbean, where digital transformation is already on the policy agenda. Whether small AI becomes a genuine development tool or another unmet promise will depend on investment in local capacity and on whether the technology is shaped around the needs of the people it is meant to serve.
3
