Can AI Help Poorer Nations Leapfrog a Century of Progress?

The World Bank Group says artificial intelligence could help developing economies increase productivity and improve essential public services, with a smaller share of their jobs exposed to automation than in high-income countries. But the report warns that shortages of electricity, connectivity, computing capacity, skills and effective institutions could prevent poorer countries from capturing those gains and deepen existing inequalities.

The World Bank says artificial intelligence (AI) could give developing economies a rare productivity advantage, but the countries with the most to gain are also those least equipped with the electricity, connectivity, skills and institutions needed to capture it.

According to the World Bank Group's World Development Report 2026: The Promise of Artificial Intelligence, 4.5% of jobs in low- and middle-income countries are exposed to automation by generative AI, compared with 14.2% in high-income economies. Yet the potential productivity gains are far closer: AI could meaningfully enhance 16.2% of jobs in developing countries, against 18.7% in wealthier ones.

The gap between automation risk and productivity potential suggests that poorer economies may not experience AI primarily as a wave of worker replacement. They could instead use it to compensate for shortages of doctors, teachers, agricultural advisers, engineers, administrators and other skilled professionals.

The Jobs Debate Looks Different Outside Rich Economies

In advanced economies, generative AI has intensified fears that software will absorb tasks once performed by office workers. Developing countries face a different labour-market structure. A larger share of employment in poorer economies remains concentrated in manual, agricultural, informal and service occupations that are difficult to automate using current generative-AI systems. It reduces immediate exposure to replacement, but it does not eliminate the relevance of AI.

Developing economies often cannot solve professional shortages quickly through conventional means. Training enough doctors, teachers, technicians and civil servants can take years and requires sustained public investment. AI cannot replace that investment, but it may increase the reach and effectiveness of the workers already in place. This is why the productivity figures deserve as much attention as the automation estimates.

The gains, however, will not be automatic. Giving workers access to an AI tool is not the same as improving productivity. Systems must be reliable, relevant to local conditions and integrated into actual workflows. Workers must know when to trust them, when to question them and how to identify errors.

The Real AI Divide Begins With Electricity

The global AI debate often focuses on advanced chips, foundation models and computing clusters. In many developing countries, the first barrier is far more basic: keeping the lights on and connecting to the internet.

In Sub-Saharan Africa, nearly one-third of rural schools still lack reliable electricity, while more than two-thirds lack dependable internet access, according to the World Bank report. In such settings, debates about sophisticated AI applications can become detached from the conditions in which teachers, students and public officials operate.

It creates a layered form of technological inequality. Countries without stable power cannot build reliable digital services. Those without affordable connectivity cannot distribute them widely. Those without computing access may depend on foreign providers. Those without locally relevant data may receive tools that work poorly in their languages or social contexts.

AI thus risks reinforcing the same divides it is promoted as helping to overcome. A clinic with intermittent electricity cannot depend on a cloud-based diagnostic assistant. A rural school without internet access cannot make routine use of an AI tutor. A government with fragmented records cannot easily deploy systems that depend on accurate, structured data.

Closing these gaps is no longer only an infrastructure objective. It is becoming a condition for economic participation in an AI-enabled world.

The World Bank Group's Mission 300 initiative, which aims to provide energy access to 300 million people in Sub-Saharan Africa by 2030, illustrates the scale of the foundational work required. Electricity expansion may not be labelled as AI policy, but it could determine which communities benefit from the technology and which remain excluded.

Adoption Matters More Than Technological Prestige

The report proposes a practical sequence for developing countries: adopt existing AI tools, adapt them to local conditions and only later advance towards developing frontier systems. The order challenges the political appeal of national AI ambitions built around expensive computing facilities or attempts to compete directly with the world's largest technology companies.

For most developing economies, the immediate question is not whether they can build the most advanced model. It is whether they can use available systems to solve specific economic and administrative problems.

Adoption could mean introducing tools that help teachers prepare lessons, improve agricultural forecasts or make public records easier to search. Adaptation would involve training systems on local languages, laws, crops, diseases and administrative processes. Frontier development would require far greater technical expertise, capital, energy and computing capacity.

Skipping the first two stages could produce costly projects with limited public value. Governments also face a less glamorous but more urgent challenge: distinguishing effective AI applications from impressive demonstrations. Pilot projects are relatively easy to launch. Proving that they improve learning, health outcomes, tax collection or business productivity is harder.

It will require clear procurement rules, independent evaluation and evidence that compares AI-enabled programmes with existing alternatives. Without such scrutiny, governments risk paying for tools that do not solve the intended problem, cannot operate at scale or create new dependencies on private suppliers. The strongest AI strategies may therefore be those that begin with narrowly defined public needs rather than broad national slogans.

Trust Will Determine Whether the Gains Last

AI can increase government capacity, but it can also magnify institutional weaknesses. A flawed system used by a private company may inconvenience customers. A flawed system used to determine welfare eligibility, school placement, medical treatment or tax liability can directly affect rights and livelihoods. This makes trust a central economic and political issue.

The World Bank report points towards voluntary industry standards, international cooperation and the use of existing laws when voluntary measures fail. It may provide a starting point, particularly for governments that lack specialised AI legislation, but the effectiveness of any framework will depend on enforcement capacity.

Rules have limited value if public agencies cannot audit systems, negotiate effectively with vendors or respond when harm occurs. There is also a risk that AI markets become highly concentrated. The most advanced systems are being developed by a small group of countries and companies. Developing economies may gain access to powerful tools without gaining meaningful control over the infrastructure, data practices or commercial terms behind them.

The key policy challenge is not simply how to accelerate AI adoption, but how to adopt the technology without creating new forms of dependence, exclusion or unaccountable decision-making.