The standard framing of Africa's technology talent problem is a pipeline problem: not enough graduates, so build more schools. I think that framing is wrong, or at least badly incomplete, and it sends money to the wrong place.
The shortage that actually constrains companies here is not junior developers. It is people who already know how an organisation works and can now put AI to work inside it. That is a different person, and you reach them differently.
Graduates are not the bottleneck
A new graduate needs two to three years before they are net positive on real client work. That is true everywhere and it is nobody's fault. But it means training a graduate is an investment with a long payback and a high attrition risk, and it does nothing for the company that needs the capability this quarter.
Meanwhile there is a much larger group being overlooked: the operations manager who has run a supply chain for nine years, the accountant who knows exactly which reconciliation eats three days a month, the agency producer who has shipped a hundred campaigns. These people already have the thing that cannot be taught in a classroom — context. What they lack is a specific, learnable set of tools.
Teaching tools to someone with context takes weeks. Teaching context to someone with tools takes years.
What certification has to mean
The word "certification" has been devalued by programmes that certify attendance. A certificate that means you sat through the material is worth roughly what it costs to print.
What we do at BUILD AI Academy is certify on shipped output. You do not finish by passing a quiz. You finish because you built something that works against a real brief, and someone who does this for a living reviewed it. The distinction sounds pedantic until you are the employer on the other side deciding whether the certificate tells you anything.
This is also why I think an academy run by an operator behaves differently from one run by an institution. We are not guessing at what the market needs — the Studio is downstream of us, and it tells us within weeks when the curriculum has drifted from what clients actually pay for. That feedback loop is the whole product. An academy without one is teaching last year's tools with great sincerity.
The uncomfortable implication
If you accept this framing, a lot of well-intentioned training spend is going to the wrong cohort, and a lot of it is measuring the wrong thing. Headcount trained is not an outcome. Completion rate is not an outcome. The outcome is whether someone's work changed, and whether somebody paid for the difference.
That is harder to report and harder to fund. It is also the only version of this that compounds. A professional who becomes materially more effective goes back into an organisation that now has evidence AI is worth taking seriously — and that organisation becomes the next customer, and the one that sends four more people.
Train the people who are already in the room. The pipeline takes care of itself after that.