Key Details
Most countries should first adopt existing tools and adapt them to local languages and needs. Only countries with sufficient computing capacity, data and specialised talent should prioritise advancing frontier AI models.
Area | Principal Finding | Development Significance |
|---|---|---|
Policy Pathway | Countries can adopt, adapt and advance AI, with each pathway requiring progressively greater capabilities. | Most developing economies can secure gains without building frontier models. |
Employment Exposure | Around 4.5% of jobs in low- and middle-income countries are amenable to automation, while 16.2% could be complemented by AI. | Aggregate displacement may initially be limited, but particular white-collar occupations face greater exposure. |
Small Businesses | Approximately one-fifth of surveyed small firms across six developing economies used AI chatbots. | Adoption is spreading, but remains concentrated in information, writing and translation. |
Technology Concentration | Five US hyperscalers were projected to spend more than US$750 billion on AI infrastructure in 2026. | Few countries can compete directly at the frontier, increasing the importance of adaptation and supplier diversity. |
Public-Service Evidence | Only 17 of 9,762 reviewed healthcare-AI studies provided causal evidence on outcomes in low-resource settings. | Pilots and technical demonstrations substantially exceed evidence supporting large-scale deployment. |
Government Evaluation | Around 80% of surveyed low- and lower-middle-income countries had no or only basic mechanisms for evaluating public-sector AI. | Government adoption is advancing faster than institutional oversight. |
Environmental Costs | Data-centre electricity use could more than double by 2030, while generative-AI-related e-waste could reach 5 million tonnes. | AI strategy increasingly intersects with energy, water, mineral and waste policy. |
AI Can Extend Scarce Expertise
The World Development Report 2026: The Promise of Artificial Intelligence examines AI through three defining features: its expanding capabilities, the growing concentration of the technology and the complements needed to use it productively.
The report argues that AI's greatest development opportunity lies in extending scarce expertise. It highlights applications that support farmers, health workers, teachers and small businesses, including through basic phones, text messages and offline systems.
Examples include:
Bangladesh: AI-assisted diabetic eye screening increased daily assessments by ~40%.
Ghana: An AI mathematics tutor delivered by SMS produced learning gains approaching one additional year of schooling at around US$5 per student.
Telangana: AI-enabled monsoon forecasts helped farmers improve production decisions and, in some cases, save up to US$560.
The report stresses that AI delivers results only when users can act on its advice through credit, inputs, insurance, market access and capable public institutions.
Adaptation Is the Main Opportunity for Developing Economies
The report identifies three pathways for AI development:
Adopt — deploy AI developed elsewhere.
Adapt — customise AI for local languages, data, laws and operating conditions.
Advance — develop frontier AI models and infrastructure.
It argues that adaptation offers the greatest opportunity for most developing economies, enabling existing models to work effectively under local conditions. By contrast, advancing frontier AI requires expensive chips, data centres, specialised talent and extensive datasets, limiting the number of countries able to compete at the technological frontier.
Labour Markets Will Change Unevenly
The report estimates that only 4.5% of jobs in developing economies are directly exposed to automation by generative AI, while 16.2% could be enhanced.
Although overall disruption may be limited, it is likely to be concentrated in ICT, finance, business services and other knowledge-intensive occupations. The report also cites evidence from South Asia, where job postings in highly exposed white-collar occupations declined by around 20%, raising concerns about entry-level employment opportunities.
Governments Must Enable, Use and Regulate AI
The report assigns governments three complementary roles:
Enable AI through infrastructure, skills, data systems, computing capacity and competitive markets.
Use AI to improve public administration and service delivery where institutions are ready.
Regulate AI to address bias, unsafe outputs, data misuse, worker exploitation and excessive market concentration.
Rather than deploying AI for its own sake, the report recommends matching technology to clearly defined public problems and institutional capability, noting that predictive and decision-support systems may often be more appropriate than direct generative AI interfaces.
What Are AI Complements?
AI complements are the conditions that allow an AI system to work effectively. They include electricity, connectivity, affordable devices, literacy, technical skills, usable data, competent institutions and competitive markets. Access to a model without these foundations may produce little value or reinforce existing inequalities.
Policy Relevance
AI strategy is principally an institutional strategy: Model access produces development gains only when users and public agencies possess the infrastructure, skills and authority to act on its outputs.
Adaptation offers a wider route than frontier competition: Local-language tools and specialised applications can address specific development problems without replicating the entire global AI stack.
Employment policy must account for concentrated disruption: Entry-level service-sector jobs can be significantly affected even when the national share of automatable employment remains small.
Public deployment requires evidence beyond technical accuracy: Health, education or welfare systems should be evaluated by their effects on access, quality and outcomes—not the number of AI pilots launched.
Interoperability can reduce technological dependence: Governments can diversify suppliers and retain control when data, models and applications can be moved between providers.
AI investment carries environmental trade-offs: Data-centre expansion must be assessed alongside electricity availability, water stress, electronic waste and critical-mineral requirements.
Follow the Full Report Here: World Development Report 2026: The Promise of Artificial Intelligence