There is a version of "AI for Africa" that is mostly imported demos with the country name swapped in. It does not change much, and it does not last. The version we are building is different. It starts from the actual workflows of African institutions and works backwards to the model.
Applied, not imported
Applied means three things. It means the problem is real — sourced from inside a ministry, a bank, a hospital, an operator — not invented to fit a model. It means the engineering can run in the constraints of the place: intermittent connectivity, lean teams, regulated data. And it means the institution keeps the capability after the project ends.
Capability that does not leave when the consultants do.
Where we are betting
Three areas, in order of conviction. Operator copilots inside large institutions where most of the value of AI is captured by the front-line user, not the headline product. Data platforms that make the institution's own data legible to its own people before any model touches it. And local-language capability — because the continent does not run in English alone, and pretending otherwise is the surest way to keep AI value off-shore.
Written by
Divinus AI
Division team