As global investment in Artificial Intelligence (AI) accelerates, Africa is still working to define its own approach. The continent faces a complex situation, balancing strong demographic potential with persistent structural constraints. High connectivity costs, limited access to computing power and a shortage of high-quality local data continue to slow the adoption of advanced technologies.

Yet these constraints have also driven an important trend: the creation of low-cost, highly optimized solutions built for resource-limited environments, which are increasingly able to compete with international models.

Celina Lee, CEO of Zindi, the leading pan-African data science competition platform, notes that regular practice, project-based learning and strong talent networks can speed up the growth of this emerging sector.

In this interview with Ecofin Agency, conducted during the Digital Transformation Summit held in Cotonou in mid-November, she outlines the key drivers for developing the African AI market, including improving data systems, building skills and strengthening public support. Lee also explains why the continent does not need to replicate large-scale models like ChatGPT to innovate effectively.

Ecofin Agency: Africa’s digital economy is growing, but adoption is still limited by the high cost of data, smartphones and restricted access to credit. Based on what you observe on Zindi, to what extent do these barriers limit the development of AI talent?

Celina Lee: Today, we have nearly 100,000 users on Zindi from across the continent. They take part in competitions built around real-world problems and develop AI solutions. However, the cost of connectivity and access to computing power are real obstacles.

For example, we recently launched a competition using video footage from roundabouts to analyze driving behavior. Even after compression, the dataset was several hundred gigabytes. For many young Africans, downloading that amount of data is simply too expensive. Data is costly relative to the average income, and having access to a sufficiently powerful machine is just as difficult.

That said, this situation has an unexpected effect: it pushes young people to develop highly innovative and efficient approaches. In a competition we ran with the Mexican government to map informal settlements using satellite images, the models proposed by African participants were far more lightweight and efficient than the ones initially used. While others relied on large amounts of computing power, African data scientists found clever solutions that required far fewer resources.

Ecofin Agency: Have you been able to measure Zindi's impact on employability and access to global opportunities?

Celina Lee: We published a report in partnership with the government of Kenya that analyzed the profiles and career paths of our users in the country. It showed that 18 percent of users secured professional opportunities directly through Zindi. We also found that participating in at least four competitions significantly improves a user's chances of being recruited. Across Africa, our surveys indicate that 85 percent of users say they have acquired new skills that supported their careers or professional advancement. This supports the idea that regular practice and exposure to real problems accelerate skills development.

Ecofin Agency: Many African companies are still hesitant to adopt AI because of limited digital maturity and fragmented infrastructure. What practical use cases can create value for SMEs, especially in agriculture, finance or logistics?

Celina Lee: What I observe is that you always need to start with the business problem, not the technology. In some cases, the right solution will involve AI, and in others it will not. We need to move away from the belief that AI should be applied everywhere. For businesses with low digital maturity, a first step might simply be helping employees use tools like ChatGPT to improve their efficiency in daily tasks such as drafting, analysis, marketing and financial reporting. 

The next step is to use solutions that are already integrated into existing software and accessible through standard licenses. It is only at a later stage that SMEs can consider integrating AI into the core of their business model. What strikes me about African businesses is their pragmatism. Economic constraints are significant, so companies adopt technology only when it delivers direct and measurable value. We should not push them to adopt AI for its own sake. Instead, we should support their gradual progress based on real needs.

Ecofin Agency: Africa faces a chronic shortage of quality local data, which limits innovation. What would it take to build an ethical, shared and large-scale data ecosystem? And who should take the lead: governments, the private sector or regional organizations?

Celina Lee: At Zindi, we have organized more than 500 competitions, each of which relies on local datasets, mostly African. This shows that much of the data a company needs often comes from within the company itself. A business can train an AI model on its own historical information, for example customer interactions, without relying on external data. On the other hand, when we talk about general-purpose models such as large language models, the logic is different. They require very large volumes of diverse data, similar to what exists across the internet. 

Few African languages currently have a large enough digital corpus to support this. To build a coherent African data ecosystem, structured collaboration will be needed. Governments can establish ethical frameworks, businesses can generate and share sector-specific data, and regional bodies can harmonize standards and ensure interoperability. A common misconception linked to the rise of ChatGPT is that effective models require enormous databases. 

In reality, AI is not limited to large general-purpose models. An SME can develop a much simpler model that performs a narrow set of tasks very efficiently. In fact, we are seeing a significant rise in Small Language Models on Zindi. These are compact models optimized for specific uses that require far less data and computing power. This trend is particularly promising for Africa, as it supports innovation without huge infrastructure requirements and aligns with the continent’s economic realities. Africa does not need to replicate ChatGPT. Smaller, specialized models can be far more effective.

Ecofin Agency: Several governments now want to create sovereign African AI models. Is this economically realistic, or is there a risk that these projects become symbolic exercises with little practical value or end up as white elephants?

Celina Lee: I do not think it is purely symbolic. It is possible, as long as the approach is pragmatic. Take Nigeria, for example. They recently launched N ATLAS, a multilingual model covering five Nigerian languages, and they made it open source. They did not train it from scratch on a massive dataset. They started with an existing open model, such as Llama, and fine tuned it using smaller, high quality local data. This path is realistic, and it works. There is a spectrum between the very large foundational models that are not feasible to build in Africa right now and the highly efficient small specialized models. The challenge is finding the right position on that spectrum. 

For instance, there are discussions about creating a foundational Swahili model because the language has a much larger digital corpus than most others. Some countries, including Benin, have begun collecting voice data to develop models adapted to local realities. These initiatives show that the work has already started. What matters is that these projects remain connected to real applications such as linguistic inclusion, public services, agriculture and education. If they address concrete needs, they will be useful and sustainable.

Ecofin Agency: Zindi plays a major role in talent development, but Africa still risks losing its best specialists. How can the continent retain its top data scientists?

Celina Lee: First, we have to be realistic. If a young person receives a great opportunity, they should take it. We cannot stop them or blame them. The issue is not that they leave. The problem is that the local market does not offer enough opportunities. Governments need to invest in the entire ecosystem. They should support startups, which are often the first entry point for young professionals, encourage research and development and create programs that train and integrate recent graduates. The private sector is often reluctant to take risks, whether by investing in AI or training young professionals who are still beginners. If governments create a clear pathway into employment, young people will stay. They want to stay. They want to build their futures at home. There simply needs to be a market for them.

Ecofin Agency: If the continent could offer reliable connectivity, affordable devices, quality data and basic digital skills, what could Africa accomplish in the next ten years?

Celina Lee: The gains would be enormous. The continent is full of energy, creativity and innovation. Young Africans are highly motivated. They solve real problems, often in difficult conditions, and they are just waiting for an environment that allows them to go further. With reliable connectivity, affordable tools and steady access to electricity, the potential would be unlimited. We have seen it in Kenya and Nigeria. An ecosystem can emerge very quickly when it is supported. I believe AI will follow the same path. And what is particularly exciting is that Africa can create applications that would not exist elsewhere because its realities are different. The continent could contribute something unique to the world.

Ecofin Agency: Which sectors should be prioritized for AI deployment over the next five years?

Celina Lee: On Zindi, agriculture stands out clearly. It is the sector where the impact can be the strongest and the fastest, whether through yield forecasting, disease detection, irrigation optimization or weather analysis. Climate change is also central. It is not a sector in itself, but it affects every industry and is an area where AI can provide tremendous support. After that come traditional commercial uses such as finance, retail, logistics and manufacturing. Wherever AI helps reduce costs or increase productivity, businesses will adopt it.

Interview by Fiacre E. Kakpo