Generative AI, productivity, and inequality in the Global South

Abstract

Generative artificial intelligence (GenAI) is rapidly diffusing across economies that remain structurally unequal in data, compute, skills, and institutional capacity. Using an integrative, multi-disciplinary secondary-data design, this article develops a decolonial, capability-centred political economy of GenAI and the future of work in the Global South. Drawing on ILO Working Papers 96 (2023) and 140 (2025), we show that roughly one in four workers globally are in occupations with some GenAI exposure, but only about 3.3% of jobs fall into the highest-risk category; exposure is concentrated in high-income economies, where 34% of employment is in exposed occupations, compared with 11% in low-income countries, and is disproportionately borne by women and clerical workers. Productivity experiments in the United States and other high-income contexts report average gains of 14–15% in customer support and substantial gains in writing quality, with especially large benefits for lower-skilled workers. Yet cross-country readiness for AI remains starkly unequal: the Oxford Insights Government AI Readiness Index 2022 reports a global average score of 44.61/100, compared with 29.38 in sub-Saharan Africa and 38.59 in North Africa, with 21 of the 25 lowest-scoring countries located in sub-Saharan Africa.

Anchored in the task-based approach to labour markets, the human-development capabilities framework, and decolonial theories of AI and data colonialism, we articulate a conceptual model linking GenAI exposure, structural AI readiness, and governance regimes to capability expansion or erosion for workers and communities in the Global South. Methodologically, we combine critical synthesis of peer-reviewed economics, philosophy, and science-and-technology studies with analysis of cross-national indicators from the ILO and Oxford Insights, and interpret these through emerging decolonial AI governance proposals from multilateral bodies and Global South scholars.

We find that: (1) the immediate risks of GenAI-driven net job destruction in low-income countries are lower than often claimed, but (2) the risk of deepening “capability inequality” via data colonialism, concentration of AI infrastructure, and exclusion from rule-setting is substantial; and (3) decolonial, capability-oriented governance—emphasising data sovereignty, investment in labour-augmenting applications, and Global South leadership in agenda-setting—offers a viable alternative trajectory. The article proposes a multi-level impact-assessment and monitoring framework that aligns GenAI deployment with decent work, human development, and decolonial justice.