Lire en français African countries face hard economic choices. Over the last 50 years, manufacturing has become more automated, capital-intensive, and dependent on high quality infrastructure. Its share of global employment has been declining for decades, even in China.
Our recent whitepaper1, makes the case that an AI powered services sector strategy is a plausible component of a growth strategy as advances in artificial intelligence are beginning to alter what services are, how they are produced, and where they can be delivered.
It is useful to distinguish between two phases of service globalisation. The first, enabled by digitization in the 2000s, expanded the set of services that could be delivered across borders, particularly in finance, ICT, and business process outsourcing. AI may represent a second shift, not only making services tradable, but reshaping how they are produced. Through task-level automation and augmentation, AI can reduce frictions related to language, coordination, and compliance, while enabling new forms of human-AI collaboration.
In this setting, tradability is less a fixed characteristic than an outcome shaped by how tasks are structured, outputs standardised, and technologies deployed. Three forces are particularly relevant: task augmentation, tradability, and technology. Together, they influence whether economies move into more cognitive, AI-complementary work and whether services can be delivered across borders, effectively enabling a form of “virtual migration.”
Emerging evidence shows AI may expand the range of services that are exportable by reducing barriers like language, search friction, and compliance costs. At the same time, the opportunity is not just to sell the same services more cheaply, but to build new production models, where human workers supervise, validate, and extend AI outputs for global markets — providing the judgment and quality assurance that fully automated systems lack.
What appears to matter as much as access to AI systems is the capacity to use them effectively. This depends on what might be described as ‘cognitive capital’, the institutional and organisational capabilities that allow AI to be embedded into real workflows. These include data standards, interoperable systems, regulatory adaptability, and domain expertise within firms and governments.2
This framing also shifts attention within current policy debates. Government investments in specialised computing infrastructure may be less decisive in many contexts than commonly assumed, particularly where private sector provision is available. Conversely, an exclusive focus on risks such as bias or privacy, can obscure the central challenge of generating employment at scale.
For policymakers, AI can be seen as a tool inside a larger jobs-and-exports strategy, not as a standalone “AI strategy.” This could tap into the same dynamics that made manufacturing powerful, targeting large export markets and applying technology at scale with the result of absorbing large numbers of lower-skill workers. This suggests several practical directions:
Progress is likely to depend on augmenting existing capabilities in a small number of sub-sectors then expanding into adjacent sectors where there is global demand.3 Examples might be cross border banking and mobile money in Kenya, or Nollywood films and Afrobeats in Nigeria, as well as newer, AI-native services such as data curation or AI-assisted compliance, where human judgment remains essential.
This will often require working with larger incumbent firms, rather than just startups, with targeted enabling investments such as reliable power, connectivity, digital public infrastructure, and training where it can accelerate growth.
Support can be made conditional using “export discipline”, as used by the Asian tiger economies: to target public goods and other forms of support at companies that show they can win in global markets via export receipts.
Services automation by AI does present risks to jobs, but the best approach may be to face them head-on rather than shying away from services altogether. The available evidence is mixed, but not discouraging. Some studies suggest that AI tools can disproportionately raise the productivity of less-experienced workers, and that augmentation may be more common than full automation.
Nevertheless, uncertainty remains high. AI technologies are evolving rapidly, and no single sector or model is likely to dominate. A portfolio approach—combining services with continued investments in agriculture, manufacturing niches, and regional markets may prove more robust than a singular focus on any one pathway.
Given the complexity and risk involved in selecting sectors, policymakers may wish to use data driven methodologies. Our paper cites a ‘progression network’, a new, data-driven methodology that links technology subfields with the goods and service categories to grow into export. strengths. The right posture is ambition matched with data-driven discipline: make targeted bets, measure performance, and adapt as conditions evolve.
There are real risks, yet there is also a credible opportunity. No single strategy will absorb the full scale of Africa’s workforce expansion, but a competitive, AI-enabled service sector, one built around human-AI complementarities, could still anchor a broader growth model that generates foreign exchange, fiscal revenue, and rising demand for domestic goods and services. This entails a more complex political economy than simply writing a national AI strategy but also closer to what job creation requires.