Arthur Lewis Was Right About Everything, Except The Robots

In 1954, economist W. Arthur Lewis, the first person of African descent to win a Nobel Prize in Economics, published a seminal paper in development economics. His insight was simple: poor countries had many underemployed workers in low-productivity agriculture, and industrialization could move this surplus labour into more productive sectors, generating profits that could be reinvested to expand production and create more jobs. Development was not just about producing more but about moving people into more productive work.

Lewis’s framework explained East Asia’s structural transformation, from South Korea to China. But he couldn’t have foreseen what happens when factories, offices and industries expand without proportionally absorbing labour. Artificial intelligence (AI) forces us to confront that question. For Africa, with its growing working-age population, this may become a key economic issue of the century.

The Lewis model was a theory of labour absorption in which workers shifted from low-productivity agriculture to higher-productivity industry. As surplus labour decreased, wages rose and workers benefited more from productivity gains, a phenomenon known as the Lewis turning point.

AI introduces a new possibility: productivity growth without a proportional increase in employment. Machines have displaced human labour since the Industrial Revolution. What makes AI different is its reach into cognitive work. Software can now write code, analyze documents, translate languages, prepare accounts and perform parts of professional jobs once considered relatively secure. This does not mean humans will simply disappear from production. AI can augment workers, create new occupations and increase demand elsewhere. But it changes the economic calculation.

A human worker needs a salary, housing, healthcare, and social protection. Artificial labour needs electricity, computing infrastructure, and software. As machine intelligence becomes cheaper and more capable, employers face a new question: should the next unit of production come from a worker or from a unit of compute? That question was largely absent from Lewis’s world. 

The traditional development ladder was relatively clear: agriculture to manufacturing, then to services, and eventually to sophisticated knowledge industries. Each stage created new opportunities for human labour. But that ladder is changing just as Africa needs to climb it.

Many developing countries face premature deindustrialization, with manufacturing employment peaking at lower income levels than in advanced economies. Automation and global competition have shortened the manufacturing pathway. Now, another risk emerges: premature de-professionalization. Finance, software, accounting, consulting and business-process outsourcing seemed protected by their reliance on cognitive skills. Generative AI is now entering these sectors.

For Africa, the question isn’t just about adopting AI but about whether it can create ladders of economic mobility rather than remove rungs. AI is reshaping the factors of production; labour now often means human-machine collaboration, which can be empowering.  

But capital is changing too. Productive assets increasingly include algorithms, foundation models, proprietary datasets, semiconductor capacity and computational infrastructure. These assets can scale globally at remarkable speed, concentrating economic value among those who control compute, data, and models.

Africa has long relied on an economic model that exports low-value raw materials and imports high-value finished products. In the digital realm, we must avoid recreating this pattern by exporting data and talent while importing algorithms and cloud services, which capture most of the value. Otherwise, we simply replace mineral dependency with algorithmic dependency.

AI sovereignty matters because it means having the capacity to make meaningful choices, such as accessing compute, governing data, developing local applications, training people and evaluating imported systems, rather than necessarily building a frontier AI model. Sovereignty is about agency, not isolation.

If AI weakens the historical link between productivity and mass employment, our institutions must adapt accordingly. First, governments should distinguish between labour-augmenting and labour-displacing innovations. Productivity is essential, but it cannot be our sole measure. We must also ask: Who captures the gains? Who gets the jobs? Who owns the technology?

Second, social protection must evolve because industrial-era systems relied on stable employment for income and security. If AI weakens that link, countries need adaptable social protection, lifelong learning and support during technological shifts. The key is that technological progress shouldn’t produce economically disposable citizens.

Third, taxation should follow value creation. Governments can’t rely solely on taxing human labour if digital and automated systems generate more economic value. This isn’t about a “robot tax” but about modernizing taxation to reflect production shifts so that societies share fairly in the value created by digital and algorithmic capital.

Lewis’s turning point was optimistic. As surplus labour vanished, workers gained bargaining power, wages rose, and productivity improved living standards. AI could be a different turning point when it becomes capable and cheap enough that adding compute surpasses human labour in many activities. For aging societies with shrinking workforces, that could solve a major problem.

Africa faces a different challenge: a growing population that needs productive opportunities, not labour shortages. We can’t simply adopt an AI economic model designed for labour shortages. Instead, Africa needs a model that uses AI to boost productivity while expanding human skills, entrepreneurship and participation. Lewis saw development as more than growth; it includes structural transformation. While that view is correct, the structure is changing again. In the AI era, we must ensure meaningful jobs remain available for people to transition into.