In developing economies, the AI debate has focused less on border risks and more on how to ensure that the technology actually serves people. This requires not only designing AI tools for use in government, but also investing in the necessary national capacity to use the technology.
WASHINGTON, DC—Many states are trying to do more for their citizens, despite outdated systems. That means accessing government services can feel like a second job. A mother walked half a day to the clinic, only to be turned away because she needed a form from an office in another city. A farmer who qualified for financial support gave up after passing from queue to queue. AI, if used properly, can help close that gap, expanding the reach of the nation it can.
For low- and middle-income countries, the stakes couldn’t be higher: AI can be a great equalizer, or it can exacerbate inequality. While AI’s contribution to agriculture, health care, and education holds great promise, the World Bank’s 2026 World Development Report estimates that AI’s productivity gains could more than double in advanced economies if current trends continue. The future trajectory of technology depends on the decisions developers, investors, and governments make today about how it’s built, funded, and deployed.
As Kristalina Georgieva, managing director of the International Monetary Fund, explained at the IMF-World Bank Annual Meeting in October, AI adoption depends on access to electricity, digital connectivity, and workers equipped with the right skills. But after decades in development finance, I am convinced that access, while a legitimate concern, is not the biggest obstacle. Above all, the adoption of AI requires building the capacity of countries to use these tools.
In advanced economies, the AI debate has largely focused on safety, security, and the threat of job displacement. But in emerging economies, the more pressing question is how to ensure that AI tools actually serve people. They must work with local languages, train on local data, and build for local realities. And it must be combined with investments in institutions and people closest to the problem, who know where AI tools can provide real benefits and can guarantee them. Much of this important work lies with the companies that make these tools and the funders that support them.
But even the best tools will struggle to produce results if governments lack the capacity to implement them. Creating that capacity requires integrating AI systems into existing programs. Consider MahaVISTAAR, a free AI advisory service built by the Indian state of Maharashtra for farmers to ask about crops, weather, and pests in their own language via an app, voice call, or chat. To date, more than 740,000 farmers have joined, with 70,000 new users signing up every month, at a cost to the government of less than 18 cents per farmer—a fraction of the cost of reaching them. The tool works because the government sees it as a new layer of public infrastructure built on a network to provide advice to farmers, rather than a single application.
On a recent visit to Pakistan, I witnessed provincial governments working to build this kind of capacity, which they consider a public good. Punjab has created a government office dedicated to AI and is collaborating with local researchers to develop public health applications, including local language tools that help frontline workers reach mothers with the right guidance at the right time.
AI tools meant to serve citizens should work in government, not just be adapted. They must be built to open public systems and data, adhere to the data sharing and security protocols that policymakers choose for their countries, and be affordable for governments to acquire and maintain.
In my experience, three elements are essential to building AI capacity. First, governments need leaders who can define the best use cases for AI, along with institutional ownership and cross-ministerial coordination. AI strategies require clear budgets and detailed timelines, and must be supported by empowered ministries and binding decision-making mechanisms across agencies. This can reduce the possibility of parallel projects, which can lead to duplication and suboptimal results, undermining existing limited capabilities.
Second, countries need civil servants with sufficient AI expertise to commission, acquire, integrate, evaluate, and manage these systems. This means developing a talent pipeline that funnels people with the necessary technical skills – as opposed to general digital literacy – into government roles. Although rich countries do not know how to meet these demands, developing countries can hardly afford to ignore them.
Third, the government must build an integrated digital public infrastructure, including e-government service delivery, digital ID, payment, and a secure data exchange layer. Weak digitization and fragmented vertical systems are immediate obstacles to AI performance and safety. Once that infrastructure is in place, AI can help make government systems more secure — detecting fraud, cyberattacks, and anomalous patterns.
Creating tools can be the easier part. Whether AI reaches that mother or that farmer will be determined less by the pace of technology than by the choices governments make today. If policymakers get it right, AI can help states accomplish their most basic and most important tasks: serving their citizens effectively.