By: Karl Fischer - CTO at Obsidian Systems

Africa’s AI journey has changed sharply over the past two years. The clearest sign is financial. Organisations are now setting aside real budgets for AI rather than treating it as a side experiment.

Early adoption often started with inexpensive tools or small proofs of concept. As usage spread into daily work, the costs became harder to ignore. Licences, cloud consumption, integration work, governance, and internal support all require funding.

Leaders are now asking where AI can deliver useful results and how to keep spending predictable as adoption expands. That is a healthier discussion than chasing every new model because it brings AI into the same commercial discipline as any other enterprise technology.

Africa can build differently

Africa is increasingly described as an AI innovation hub. I believe one of the continent’s strongest advantages comes from its history of leapfrogging older infrastructure models. Mobile adoption grew quickly across many African markets because fixed infrastructure was limited. The result is a digital economy that developed differently from Europe or North America, with mobile devices serving as the main computing platform for much of the population.

That experience creates a strong opportunity for edge AI. African users are highly distributed, while connectivity and infrastructure quality can vary widely. Systems that process information closer to the user or device can be more practical than services that rely entirely on distant cloud infrastructure.

Solutions built for these conditions have to work with limited bandwidth and older devices, often in places where access to infrastructure is uneven. That pressure can produce technology that is useful across other emerging markets as well.

Infrastructure cannot remain an afterthought

The opportunity will be limited if the continent does not invest in enough computing capacity. AI workloads demand far more processing power than many traditional applications, and I do not believe Africa is preparing quickly enough for the level of adoption now being discussed.

Local data centres and high-performance computing will be important, while reliable connectivity remains essential. In some markets, regulation still slows the use of proven technologies that could improve access.

Sovereign AI is difficult to discuss seriously when African organisations cannot operate critical workloads in environments they control. This does not mean every workload has to remain on the continent. It means organisations need a genuine choice about where their data is processed and how their systems are secured.

The origin of a model is only part of the story

There is a tendency to assume that Africa needs to build every foundation model from scratch. I do not think that is the most useful starting point. The more important issue is whether organisations can inspect the technology and adapt it to local needs without depending on a single vendor’s permission. Licensing has a direct effect on that freedom. A vendor-controlled model can change in price or access, while usage terms can also shift with little warning.

Open-source models and openly licensed datasets give local teams more room to adapt systems to African languages and operating conditions. They also create a stronger base for long-term services because the organisation has more control over deployment and future development.

For me, AI sovereignty is about practical independence. African organisations should be able to decide how the technology is used and where the data goes.

Where the value is becoming visible

The strongest adoption is currently appearing in environments with high volumes of customer interaction or technical support. AI can help teams find information faster and reduce the routine work that absorbs skilled people’s time.

The next wave is likely to emerge from admin-heavy parts of the enterprise. Many organisations still rely on employees to move information between systems or check standard documents. These processes are often repetitive, which makes them suitable for carefully designed automation.

AI can deliver immediate value in this kind of work, provided the process is clear and a person remains accountable for the outcome. The aim should be to remove avoidable effort without weakening oversight.

Africa’s AI opportunity will come from building for local operating conditions and investing in the capacity adoption requires. It will also depend on retaining enough control to adapt as the technology changes.

That is a more durable path than chasing every new feature. It is also the path most likely to produce AI that works for African organisations rather than around them.