Artificial intelligence for low income countries

Currently, artificial intelligence (AI) holds a prominent place in discussions. The adoption of AI and its impact on organizations, businesses, and society have been extensively explored. Notably, the IBM Report shows the global AI adoption rate reached 35% in 2022, a four-point increase from 2021, and McKinsey reports adoption has more than doubled since 2017. Generative AI is expected to further accelerate uptake. Despite widespread benefits across private and public sectors, most AI research and development remain centered on high-income countries (HICs), overlooking low-income countries (LICs) defined by the World Bank as those with GNI per capita of $1145 or less (2024–2025). LICs have significant needs in health, education, energy, and governance, yet lag in adoption and in the academic discourse. This position article argues for AI adoption in LICs grounded in distributive justice, technological equity, and digital decolonization; contends that AI catch-up is feasible via leapfrogging and absorptive capacities; identifies high-impact domains (health, education, energy, governance); and proposes strategies to bridge the AI gap. The contribution is to re-center LICs in the global AI conversation, outline where AI can have near-term impact, and specify conditions required for success. The paper is structured with a background on AI inequality, methodology, arguments for AI’s importance in LICs, theoretical justification for catch-up, domains of effective use, policy strategies to narrow the gap, and a conclusion.

Background. Technology (used synonymously here with AI) drives socio-economic development. The AI divide arises from: (1) lack of software/hardware to develop and use AI and (2) insufficient knowledge and training—echoing the digital divide literature. LICs have fewer AI tools and skilled professionals than HICs, and current AI applications and researchers are concentrated in HICs. Over 60 countries (over 70% developed) have issued national AI policies in the past five years, while most LIC governments rank at the bottom of AI readiness (Oxford Insights, 2022). LICs are underrepresented in AI research: a negligible 0.03% of AI conference publications in 2020 came from sub-Saharan Africa, while the US, EU, UK, and China dominate publications, citations, and leading institutions. Business adoption varies widely by country; however, surveys often omit LICs altogether. This AI divide threatens to widen productivity and income gaps, and impair growth in education, healthcare, energy, and governance if LICs do not plan proactively. Neglecting LICs in AI contradicts distributive justice, technological equity, and digital decolonization. The paper argues for theoretical and practical pathways for LIC AI catch-up via leapfrogging and absorptive capacities, envisioning more uniform, global AI use. The article adopts a broad definition of AI across rule-based systems to machine learning and generative AI.

This is a position paper grounded in established economic theory (e.g., Abramovitz’s catch-up), technological equity, and absorptive capacity frameworks from firm- and national-innovation literatures. Given limited AI presence and scarce comparable data in LICs, the authors refrain from quantitative cross-country analysis. The evolving nature of AI and lack of consensus on adoption metrics further complicate empirical approaches. Instead, the paper synthesizes academic and policy literature, examples from LICs/LMICs, and the authors’ contextual knowledge, including feedback from anonymous reviewers, special issue editors, and seminars (e.g., GEI/World Bank/UNDP 2024). The approach articulates theoretical feasibility and policy prescriptions, focusing on domains with existing AI tools and urgent need.

• AI adoption is accelerating globally (IBM 2022: 35% adoption; +4pp YoY; McKinsey: more than doubled since 2017), with generative AI further boosting uptake, but LICs trail in readiness and policy development.
• LICs are largely absent from AI policy and research leadership: over 60 countries have national AI strategies (70%+ developed); LIC governments mostly rank at the bottom of AI readiness (Oxford Insights, 2022). Sub-Saharan Africa produced ~0.03% of AI conference publications in 2020; US/EU/UK/China dominate outputs and citations.
• Normative case for AI in LICs: distributive justice, technological equity, and digital decolonization argue that AI progress should be accessible and beneficial across countries and social strata; AI can drive SDGs, serve underserved populations, catalyze growth (technology-led development), and mitigate institutional weaknesses (e.g., corruption) through transparency and efficiency.
• Theoretical feasibility of AI catch-up: defined as narrowing the AI capability gap with leaders. Two complementary pathways: • Leapfrogging: some LICs can bypass stages by adopting advanced AI solutions directly, aided by globalization, diffusion, and lessons from early movers; precedent in mobile e-commerce and e-banking adoption in LMICs. Preconditions include digital infrastructure, skills, and research capacity. • Absorptive capacity: for most LICs, follow leader trajectories by acquiring, assimilating, and applying external AI knowledge; requires strong institutions, human capital, tailored policies, integration with advanced economies, and sustained investment.
• High-impact application domains for LICs with near-term feasibility: • Health: AI-enabled telemedicine can bridge geographic and gender gaps; autonomous drones (e.g., Zipline) supply remote areas—Zipline delivers ~75% of Rwanda’s blood supply outside the capital, reducing delivery times/wastage; AI for diagnostics (e.g., X-ray analysis), decision support, triage, and chatbots. Challenges: data availability, trust, cost-effectiveness. • Education: AI can support distance/personalized learning and teacher planning; examples include RoboTutor and AI chatbots (African Virtual University). Barriers: device/connectivity/electricity access, content localization, cultural acceptance, teacher training, financing. • Energy: AI-driven smart energy (smart grids, smart meters, IoT) can reduce losses, improve planning, and enable predictive maintenance; Kenya’s SteamaCo uses AI-powered smart meters/distributed grids to deliver affordable, reliable power. Requires robust digital/energy infrastructure and public training. • Governance: AI automation and e-government improve service delivery, transparency, and citizens’ trust; Karnataka’s e-land records eliminated ~1.32 million days of wait time for ~7 million farmers and saved ~INR 806 million in potential bribes; mobile bill payment in Tanzania reduced corruption; data-driven analytics improve targeting and forecasting under resource constraints.
• Policy strategies to narrow the AI gap: • Leapfrogging-oriented: build agile infrastructure (high-speed internet/5G, cloud, data systems; e.g., Rwanda’s Kigali Innovation City), foster context-driven innovation (e.g., Brazil’s AI in agriculture), leverage digital platforms and open-source libraries (e.g., Kenya’s M-Pesa ecosystem), accelerate workforce skills via online learning (e.g., Nigeria’s tech communities), and pilot AI initiatives (e.g., Latin American health pilots) to demonstrate value and scale. • Absorptive capacity-oriented: strengthen institutions (e.g., AI Singapore model), invest in human capital (e.g., Estonia’s education partnerships), craft tailored policies to local contexts (e.g., India’s CPS/smart cities), build international collaborative networks (e.g., South Korea’s partnerships), and facilitate technology transfer (e.g., Mexico’s partnerships with global tech firms).
• No one-size-fits-all: LIC heterogeneity implies choosing between or combining leapfrogging and absorptive-capacity approaches based on foundational capabilities.
• International support is pivotal: advanced AI economies and organizations (UNESCO, OECD, USAID, World Bank) should aid LICs via technology transfer, grants, and technical assistance to enable catch-up.

The paper addresses whether and how LICs can benefit from and catch up in AI. By grounding the analysis in distributive justice and technological equity, it reframes AI as a global public good whose benefits should extend to LICs. It then demonstrates theoretical feasibility through leapfrogging and absorptive capacity frameworks, acknowledging heterogeneity in foundational capabilities across LICs. Mapping four domains—health, education, energy, and governance—shows where proven AI tools from HICs can be transferred with contextual adaptation to deliver near-term developmental gains. Empirical examples (e.g., drone-enabled supply chains, e-government reforms) illustrate real-world impact on service delivery, corruption mitigation, and resource optimization. The proposed strategies translate theory into actionable steps: leapfrogging for capable LICs with agile infrastructure and pilots; absorptive capacity for most LICs via institution building, human capital development, tailored policy design, international collaboration, and technology transfer. Collectively, these findings show that targeted AI adoption can mitigate structural constraints in LICs, support SDGs, and reduce innovation disparities, provided that domestic reforms and international support align.

Global enthusiasm for AI contrasts with uneven integration that leaves LICs behind due to gaps in infrastructure, data, talent, and governance. The paper argues that AI catch-up is feasible but incremental, contingent on LICs’ diverse starting points. Using leapfrogging and absorptive capacity as guiding frameworks, it identifies high-impact domains—health, education, energy, and governance—where AI can deliver rapid development benefits. It recommends building digital infrastructure and institutions, investing in skills, tailoring policies to local contexts, piloting and scaling proven AI applications, and partnering internationally for technology transfer and assistance. There is no single pathway; LICs with stronger foundations may leapfrog, while others benefit more from absorptive capacity approaches. Future work should empirically evaluate AI adoption processes and outcomes in LICs to validate or refine these prescriptions. International organizations and AI-leading economies have a crucial role in enabling equitable AI diffusion through funding, cooperation, and knowledge sharing.

The study is an ex-ante, theory-driven position paper without quantitative cross-country analysis due to limited, evolving AI data in LICs and a lack of consensus on operationalizing AI adoption. Findings may not generalize across heterogeneous LIC contexts and require ex-post empirical validation as AI diffuses. Resource constraints, institutional weaknesses, data scarcity, and dependency risks in international collaborations can affect feasibility and scalability of proposed strategies.