AI Isn't Fate. But in Africa, We're Still Acting Like It Is.
Who Gets to Decide What AI Optimizes in Africa?
That was the quiet but radical thread running through Maximilian Kasy’s talk at Oxford yesterday, on his new book The Means of Prediction: How AI Really Works (and Who Benefits).
Sitting in the room, I kept asking myself: What does this mean for Africa, and for how we invest?
Kasy’s core argument is disarmingly simple. AI is optimisation. Someone picks what gets maximised, what actions the system can take, and what data it learns from. Whoever controls data and compute controls what gets optimised; whether that’s ad clicks, worker pay on gig platforms, who gets a job interview, who gets targeted for deportation, or who gets bombed. Those aren’t hypotheticals. Every one of those examples is already operational somewhere in the world.
Two things hit me hard as an African investor.
First: if we don’t own the means of prediction, we don’t own the futures they create. When African data is scraped into frontier models built and hosted elsewhere, decisions about African citizens, consumers, and workers will increasingly be made by systems whose objectives we didn’t set and whose governance we have no influence over. That’s not an abstract concern; it’s the current default.
Second: democratic control of AI objectives is a design choice, not an afterthought. Kasy’s argument isn’t that we need a global UN-style AI body. It’s far more practical; a city can regulate how a ride-hailing algorithm assigns wages, a hospital can govern how patient data is used, and a university can set rules for AI in admissions. Governance can be built from the ground up, domain by domain.
So what should African investors actually do with this?
Back local control of data and infrastructure. Companies building African data platforms, privacy-respecting data cooperatives, and regional compute infrastructure deserve serious capital attention. “Sovereign data” shouldn’t only be a government talking point; it should be an investable thesis.
Fund AI that optimises for African priorities. Models tuned for agriculture, logistics, financial inclusion, and public health, with objective functions aligned to local outcomes, not engagement metrics designed for other markets.
Treat governance as a product requirement, not a compliance checkbox. The question worth asking every portfolio company: who gets to define what your AI optimises, and how are the people on the receiving end of that algorithm represented in that decision?
Lean into smart regulation as a competitive advantage. Clear rules around data governance and transparency can give local founders a structural edge. Trustworthy AI, built to African regulatory standards, becomes exportable.
Africa has arrived late to most technological revolutions and ended up paying rent to foreign owners of the infrastructure that mattered. AI gives us a narrower window than usual to do it differently.
If AI is just optimisation, the real question is: who decides what gets optimised in Africa?
Our capital is one of the few levers we directly control. We should use it accordingly.