Small AI Needs Strong Institutions

A small college licenses an artificial-intelligence assistant before it has reliable internet in every classroom. A ministry chatbot answers hundreds of questions but has no route to a trained person when someone discloses abuse or despair. A public institution automates summaries while its records remain incomplete. Leaders celebrate innovation, yet the people doing the work cannot explain what has improved.

The problem is not that these institutions began too small. The problem is that they began with the tool rather than the human purpose, the curriculum, the evidence and the responsibility surrounding it.

For the Global South, the decisive AI question is not whether we can reproduce the largest model or copy the most visible university. It is whether accessible technology can help a real teacher, learner, minister or public servant perform a valuable task more faithfully under local conditions.

Small AI can address large human problems. It will do so only when strong institutions choose the problem, prepare the people, protect those affected, test the result and retain the authority to stop.

On 4 August, the World Bank released World Development Report 2026: The Promise of Artificial Intelligence. It presents what the Bank describes as its first comprehensive assessment of AI's implications for developing economies and organises its recommendations around three stages: adopt available tools, adapt them to local conditions and advance towards frontier…