Paschal Donohoe, the World Bank’s chief knowledge officer, made the case for specialised AI models that developing countries could use on mobile phones with limited connectivity in an article published on 18 October, AOL reported. Donohoe argued that adapting AI to local languages and needs could accelerate economic development in the countries the bank serves.
Key points
- Donohoe is calling for narrow-purpose models that people can use on phones in their own languages, even where connectivity is limited.
- He cited Indian farmers seeking soil analysis through the World Bank’s AgriConnect scheme as a possible use.
- The bank’s World Development Report 2026 recommends adapting existing tools, evaluating applications and using public procurement to expand those that work.
- World Bank President Ajay Banga has argued that small AI could create jobs and raise productivity in middle- and lower-middle-income countries.
Donohoe’s case for AI on phones
“Small AI” consists of specialised models trained on narrower datasets for a particular purpose. He said they could work locally, in the language of the community using them, with little connectivity. For countries considering where to spend on AI, his proposal puts the intended user and the conditions in which a tool must operate ahead of the size of the model.
He cited farmers in India involved in the World Bank’s AgriConnect scheme who needed soil analysis in their own language. Donohoe argued that there would not be enough agricultural experts and scientists to serve every small farm individually, and that a phone-based tool could help meet those information needs. He suggested that the bank, governments and private companies could collaborate on such products.
Donohoe said economic development that might otherwise take centuries could potentially happen within decades. That is his forecast for what locally useful AI might make possible, rather than a measured outcome from AgriConnect. The example identifies a specific service — soil information a farmer can use — and a means of reaching people for whom a consistently connected computer is an unsuitable starting point.
World Development Report 2026 favours adaptation
The bank’s World Development Report 2026 recommended adapting existing AI to local conditions rather than spending billions of dollars to build advanced systems from scratch, Open Magazine reported. The report called for investment in infrastructure and skills, alongside affordable applications that produce measurable results in public services. Its advice gives governments a role beyond encouraging developers: they would have to decide what to buy, how to assess it and whether a trial deserves wider use.
Among the applications described in the report were AI-assisted weather forecasts reaching more than 30 million farmers in India. In Kenya, a predictive algorithm was used to assign more than 10,000 court cases a year among over 1,500 mediators. In Bangladesh, an AI medical-imaging application increased daily screening for a diabetes-related complication by 39.5 per cent. These are different tasks, but each is tied to a service with an identifiable user and a result that can be counted.
The report recommended that governments establish ways to procure, evaluate and expand pilot projects, citing limited reliable evidence about which approaches work. It proposed beginning regulation with voluntary standards and flexible rules built on existing law. It also advocated interoperable systems to limit reliance on major AI firms in the United States and China. Those recommendations leave governments choosing both the applications they support and the evidence they require before spending more on them.
Banga sets small AI beside jobs risk
World Bank President Ajay Banga said on 12 October that small AI offered a substantial opportunity to build jobs and productivity in middle- and lower-middle-income countries, CNBC reported. He contrasted that prospect with estimates of job disruption from large language models in developed economies. Banga also estimated that 1.2 billion young people would turn 18 over the next 15 years and said the private sector would have to absorb most of them.
At the same Bangkok discussion, IMF managing director Kristalina Georgieva warned that AI was affecting labour markets like a “slow-moving tsunami”, with young entrants particularly exposed. The IMF estimates that AI will eventually augment, downgrade or eliminate about 40% of jobs worldwide. Georgieva put exposure at about 60% in advanced economies and 23% in low-income countries, and attributed the lower figure in poorer countries to AI being further away. AI Affairs previously reported her warning about the worldwide jobs figure.
Singapore President Tharman Shanmugaratnam, who co-chairs the World Bank’s Jobs Council, called for governments to make employment support of the kind used during crises a permanent part of policy. He urged incentives for companies to hire new graduates and argued that workers could adapt as AI displaced tasks, provided they had opportunities to learn at work and universities shifted towards applied training. His concern sits alongside Banga’s jobs argument: new applications may create demand for labour, while other uses of AI change the work available to people entering employment.
Donohoe said the World Bank had mobilised $120bn of private capital over the past year through guarantees and other financial instruments. He linked that work to employment and economic development, saying the bank wanted to draw attention to the connection between private investment and jobs.
