Decoding AI for Development

World Development Report 2026, with the working theme Artificial Intelligence for Development, is forthcoming and not yet published; the World Bank's concept note states it will be published in late 2026. The Report sets out to investigate the development implications of artificial intelligence as a general-purpose technology and to assess what policy choices can capture its benefits while offsetting its risks. Its stated aim is to bring a developing-country perspective to a debate that, so far, has focused primarily on high-income countries. Ufuk Akcigit of the University of Chicago serves as one of the two Lead Academics alongside Susan Athey of the Stanford Graduate School of Business, with Gaurav Nayyar as Director and overall guidance from Indermit Gill and Somik Lall. That academic leadership connects the Report directly to the Growth Academy's work on innovation, firms, and growth.

The concept note frames AI not as science fiction but as a new wave of technological change whose central function is to enhance the capabilities of people, firms, and governments. It organizes real-world AI into three cognitive functions that mirror the technology's evolution: predictive AI for pattern detection and forecasting, generative AI for creating text and other content, and emerging agentic AI that plans and executes sequences of tasks. The Report situates AI against earlier general-purpose technologies such as the steam engine, electricity, and the internet, asking how it is similar and how it differs. Two distinguishing features are emphasized: access barriers appear high for frontier model development but low for adoption of off-the-shelf tools, and the local context, including the quality and relevance of training data, is central to whether AI solutions actually work in low- and middle-income settings.

A first part decodes AI as a technology and lays out a conceptual framework for how its productive use can raise, skew, or lower societal welfare. The framework distinguishes three points along the AI value chain that matter for development: adopting ready-made tools, adapting open-source models to local contexts and data, and advancing frontier models. The concept note's working judgment is that most developing countries are positioned to benefit from adoption and adaptation rather than from advancement, given the computing infrastructure, data, and talent that frontier development requires. It also poses cautionary questions, including whether countries can leapfrog earlier general-purpose technologies, whether efficiency gains can be balanced against equity, and whether weaker institutions can keep regulatory pace with a fast-moving technology.

A second part assesses AI's implications across three pillars of the development process: economic growth and jobs (markets), the delivery of government services (state), and broader sociopolitical change (society). On growth and jobs, the Report intends to examine how AI may displace workers in some tasks, expand others, and create new ones, while weighing concerns that gains could concentrate among owners of capital or skilled workers and that a lack of complementary factors could blunt productivity effects. On government services, it considers how AI could guide the allocation of scarce resources, improve transparency, and personalize service delivery, alongside risks of entrenched bias, dependency, and weaker accountability. On sociopolitical change, it weighs the acceleration of new ideas and wider consumer benefits against risks to human capital, the environment, social cohesion, and geopolitical contestation.

A third part turns to policy priorities for low- and middle-income countries, organized around three directions: empowering entrepreneurs to adapt and apply AI through competition and local innovation; educating people through broad-based skills and AI-specific competencies, including engagement with diaspora and foreign talent; and ensuring trust through governance and institutional frameworks that can be tested, evaluated, and adjusted as the technology evolves. The concept note also signals that the Report will rest on new evidence rather than existing data alone, including surveys of AI use within governments and firms across World Bank regions, analysis of online job postings, research syntheses on human capital, and a planned cross-country measure of generative AI adoption. Because the Report is still in preparation, these are stated questions and intended contributions rather than final findings.