World Development Report 2026

AI’s expanding set of capabilities to perform or help accomplish more expert tasks effectively expands the supply of skills available in developing countries in areas such as health care, agricul­ture extension, and government administration. However, AI’s benefits are neither equally distributed among geographic areas and population segments nor guaranteed to materialize. AI can also enhance the harmful capabilities of malicious actors. 

The layers of the AI value chain—from hardware and infrastructure to models and applications—are concentrated in a few countries, often in only a handful of companies. Building the most advanced layers of the AI value chain is difficult, but accessing the capabilities of AI by adopting and adapting what has been built by others is more feasible and less expensive. Nevertheless, access to AI’s capabilities alone does not guarantee that coun­tries will benefit from them, because the effective­ness of AI solutions depends on how well they fit specific contexts.

AI’s potential depends not only on the context in which it will be used, but also on the context in which it has been developed. If those contexts are mismatched, AI will not work well. Realizing the full potential of AI therefore requires the right complements to be in place to adopt AI applications and to tailor them to the context in which they are deployed. Whether AI ultimately does more good than harm will depend on how quickly the complements are put in place.

AI is transforming jobs, productivity and growth. With most people employed in manual rather than cognitive work in developing countries, the immediate impact of AI adoption is to complement rather than displace workers—with early signs of job disruption concentrated in knowledge-intensive services. Over time, as complementary factors—infrastructure, skills, and institutions—develop, AI can enable productive busi­nesses to grow. In the absence of the right com­plements, AI may widen gaps in productivity between developing countries and high-income countries, where AI adop­tion is higher. AI can also have cascading effects through disrupting trade patterns, increasing human capital, raising energy costs, and creating economic volatility.

AI is already transforming the delivery of public services at the back end of public administration, such as weather forecasting and tax enforcement, largely through the adaptation of machine learning algorithms to local contexts. Adapting and adopting AI can also help overcome critical shortages of skilled personnel, such as teachers and health care workers, at the citizen-facing front end of public service delivery. The lack of evidence for selecting, procuring, and evaluating AI applications, especially at the front end, limits AI’s full potential to improve the delivery of public services. The lack of complements, such as digital connectivity and skills, also constrains AI adoption at the front end of public service delivery.

AI is driving power asymmetries between countries, governments, corporations, and indi­viduals. Between countries, control over key parts of the AI value chain—frontier models, advanced chips and data centers—by a few companies creates dependency risks, but pursuing “AI sovereignty” comes at a high cost and with uncertain benefits. Between governments and corporations, finan­cial incentives to attract AI investment squeeze public budgets yet provide uncertain returns. Between corporations and individuals, AI can benefit owners of capital more than people because of the concentration of market power. Between governments and individuals, AI can spread propaganda and enable the surveillance of citizens.

 As enablers of AI, governments can shape its economic impact by building the analog and digital complements, such as infrastructure, data, skills, and institutions. This starts with establishing the analog complements for the productive adoption of AI by businesses and people—foundational skills, infrastructure, and informa­tion. Governments should also support local firms’ ability to access key digital and analog complements—computing infrastructure, training data, AI talent, and access to capital—for adapting AI to local contexts and for advancing AI at a later stage. These policies must be pursued systematically because returns to investments in one domain are reduced when other domains are neglected.

As users of AI, governments can shape its social impact through improving public service delivery. Governments should use their buying power to spur the adaptation, and potential advancement, of AI solutions across use cases—such as those in educa­tion, health, and the legal system—in which local contexts are paramount. This requires making public administrative data AI-ready. This also requires establishing frameworks for testing, procuring, and evaluating AI solutions to move from pilots to large-scale deployment. Government procurement can crowd in local AI solutions if those developed elsewhere are either less effec­tive or even harmful in high-stakes applications, such as health care. All of this will require investment in staff capacity.

As regulators of AI, governments can shape its political impacts. Because regulatory capacity is often limited and the technology is evolving rapidly, governments should rely on voluntary private standards as a starting point to build trustworthy AI. Where market forces are not aligned with the responsi­ble adoption, adaptation, and advancement of AI, governments should use existing regulations to address social harms (such as violations of privacy and cybersecurity failures). Over time, governments should assess the current landscape of regulations, identify and address gaps, and build enforcement and compliance capacity, while empha­sizing international cooperation to guard against fragmentation of regulations across countries.