AI in Africa: Leap-frog or lock-in

Africa has often pinned its hopes on the promise of leapfrogging stages of development.

Mobile phones would bypass landlines. Fintech would bypass banks. Renewable energy would bypass grids.

Now comes artificial intelligence, and with it, the boldest promise yet: that Africa can skip industrialisation entirely and jump straight into a knowledge economy. No factories, no heavy industry — just code, data, and innovation.

As a coder and data analyst that should make me feel comfortable, but history suggests something more uncomfortable. Technology does not erase structure. It amplifies it.

Because development has never been about technology alone. It has always been about structure — who produces, who owns, and who captures value. And on that front, AI risks entrenching the very patterns it claims to disrupt.

The promise: AI as a shortcut to development

At first glance, the case is compelling. AI lowers the cost of knowledge. It automates expertise. It allows countries with limited infrastructure to plug directly into global systems. In agriculture, AI can optimise yields. In healthcare, it can compensate for shortages of doctors. In education, it can scale personalised learning across entire populations.

In theory, AI could allow African economies to bypass the slow accumulation of industrial capabilities and move straight into high-value services. For countries with weak infrastructure, AI looks like a miracle — a way to compress decades of development into a few years.

This matters because development has always been about capabilities — the ability to produce complex goods and services. Countries that diversify into more sophisticated activities grow faster and are more resilient.

AI appears to offer a shortcut to those capabilities. But access is not the same as capability, and that is an important difference.

The reality: Development is path-dependent

Economic transformation does not happen in a vacuum. It’s deeply shaped by what already exists — skills, industries, institutions. Research shows that regions diversify into new activities that are closely related to their existing capabilities. The further they move from what they already know, the higher the risk of failure.

This is the core problem for AI in Africa. Because AI does not automatically create capabilities: it builds on what already exists. For example, Kenya’s use of AI tools in agriculture built  on existing advantages, including mobile phone coverage and higher education institutions, to increase productivity and value, including among smallholder farmers.

Countries with strong education systems, data infrastructure, and industrial ecosystems will use AI to move into more complex activities. Those without these advantages economies risk being locked into low-value roles.

Instead of leapfrogging, AI may reinforce the same structural divide. Because advanced economies design and control AI systems. Developing economies consume and implement them. So the global divisions of power and wealth remains intact, just digitised.

The labour market shock: Automation without absorption

The greatest challenge AI presents to Africa is not technological. It is the impact on labour.

Africa’s development model has always depended on structural transformation: moving workers from low-productivity employment such as agriculture into higher-productivity manufacturing and services.

AI threatens to disrupt that pathway. Many of the occupations that AI automates are routine, middle-skill jobs — precisely the kinds of jobs that historically absorbed labour during development.

Meanwhile, the jobs AI creates are often high-skilled, complex occupations requiring advanced education and specialised capabilities.

And herein lies the trap. Complex occupations are not easily adopted. They depend on existing high level skill bases and institutional capacity, and these are concentrated in already advanced regions .

This creates a dual risk. One is that it produces automation without industrialisation. The other is that innovation is not accompanied by inclusion.

So if AI is not introduced carefully, Africa could face a future in which productivity rises, but employment does not.

The platform problem: Who owns the value?

Even where AI is adopted, another question emerges: who captures the value?

AI systems are not neutral tools. They are embedded in platforms — owned, trained, and controlled elsewhere. This matters because development is not just about using technology. It is about capturing value from it for the local economy.

Without domestic capabilities, developing economies will not reap the benefits of AI. Data generated will flow outward, profits will accrue elsewhere, and local businesses will remain dependent. Consumers of AI, not creators.

This mirrors existing patterns in global value chains, where countries exporting raw materials may capture little of the final value. AI risks becoming the digital equivalent of the resource curse — high usage, low ownership.

The inequality engine: AI and the geography of development

AI reshapes inequality, within countries as well as between them. Within countries, the benefits of technological change tend to concentrate in already advanced regions — those with better education, infrastructure, and connectivity.

AI accelerates this dynamic. Urban elites, tech hubs, and globally connected firms move ahead. Rural areas and informal economies fall further behind. In the AI revolution,  exclusion from it is as much a constraint as lack of literacy or numeracy was in previous eras.

The policy illusion: Technology without strategy

Perhaps the biggest danger is not AI itself, but misperceptions of how it can be utilised. There is an illusion that AI can substitute for development strategy, and across Africa, there is a growing focus on adopting AI strategies, innovation hubs, digital roadmaps.

However, adoption without transformation does not produce sustainable development.

Development economists have increasingly viewed structural transformation as the central mechanism through which countries can achieve sustained growth. They grow by deliberately building capabilities, diversifying production, and moving into higher-value activities. Economies that fail to do so remain vulnerable — dependent on a narrow set of low-value sectors.

AI does not bypass this process. Without industrial policy, education reform, and increased institutional capacity, AI becomes just another imported technology layered onto a weak economic structure.

And weak structures do not produce strong outcomes. Without the necessary policy, adopting AI is consumption, not economic transformation.  Digital tools are more likely to reinforce low-value specialisation and act as a roadblock to structural change.

The opportunity: AI as a tool, not a strategy

None of this means Africa should reject AI, but it does mean the conversation needs to change. AI is not a shortcut to development.  Used well, it can enhance the performance of existing industries and support gradual diversification. Over time, with the right investment, it can build new capabilities.

Used poorly, it’s more likely to  entrench dependency, displace workers – especially those in the all-important emerging intermediary jobs – and deepen inequalities.

The difference lies in how it is embedded into broader economic transformation. The introduction of AI must be used as an opportunity to build capabilities. This means:

• Investing in the workforce, in skills training, not just in software.

• Building local data ecosystems, not just importing tools from developed countries.

• Embedding AI within broader industrial strategies, not treating it as the core industrial strategy.

  
  

Conclusion: Leapfrog or lock-In?

AI offers speed. In the right circumstances it offers acceleration along a pre-determined  trajectory. But development is all about defining the direction, and creating the preconditions.

The question is not whether Africa adopts AI, but whether AI helps Africa travel in the right direction up the ladder of complexity — or locks African economies further into underdevelopment.

Because in the end, the future will not be determined by algorithms, but by who builds them, who owns them, and who benefits from them.