A young man came to see me last month. He had finished three certificates — two on machine learning, one on prompt engineering. He had built a chatbot, deployed it, and was looking for the next step. He wanted to know how to "get into AI."
I asked him what problem he was trying to solve.
Long pause.
That pause is the conversation Nigeria is not having with its young people. Everyone is telling them to learn AI. Nobody is sitting them down and telling them what to use it for.
It is not his fault
The advice he had been given — by influencers, by bootcamps, by his own ambition — was technically correct. Learn the tool. Build a project. Apply for a remote job. Earn in dollars. It is a real path. Many have walked it.
But it is also upstream advice. It assumes that once you know the tool, you will figure out what to point it at. In a market where there are more problems than there are people thinking deliberately about them, that assumption is dangerous. It produces a generation of capable hands holding tools they do not know what to do with.
There is a Yoruba reminder I keep coming back to: ọwọ́ ẹni la fi ń tún ìwà ara ẹni ṣe — we use our own hands to repair our own character. Translated into this conversation: the hand matters more than the hammer.
" We keep upgrading the tool and ignoring the hand holding it.
The economy does not reward tool-knowers
The graduates of Nigerian universities — and increasingly, our online bootcamps — are being trained as tool-knowers. Python. React. TensorFlow. Now, prompts. The cycle of "what skill should I learn next" is treated as the central question.
But every market in the world — including the ones that pay in dollars — rewards problem-solvers, not tool-knowers. A junior engineer who knows three frameworks but cannot articulate a real-world problem is just another resume. A young person who has sat with a specific problem in their own community for three years — even one they cannot yet solve — is closer to value than they realise.
The tool can be learned in weeks. The problem-finding instinct takes years.
Where the real problems live
Here is the part nobody says out loud. Nigerian youth have access to a frontier of problems the global tech industry literally cannot see.
They are walking through markets where traders keep records on paper. They are watching teachers grade three hundred students by hand. They are sitting in clinics where patient files are stacked in cardboard boxes. They are arguing with okada riders about meter-less pricing. They are translating between banking apps and grandparents who never learned to read English on a screen.
Every one of those moments is a problem someone, somewhere, will eventually be paid to solve. The question is who.
If the future of African technology is going to be built by anyone, it should probably be built by the people who have lived the friction firsthand. Not by an engineer in San Francisco reading about it.
A framework, if you want one
When a young engineer asks me now how to "get into AI," I try not to answer the question they asked. I answer the one they should have asked.
1. Find a problem you have personally lived. Not one you read about. Not one you saw on a pitch deck. One that has interrupted your own week, the way it interrupts other people's weeks. If you cannot name a real problem from inside your own life, you are not ready to build for someone else's.
2. Map who is suffering and why. A problem is not a complaint. It is a chain of decisions, incentives, and constraints. Who suffers when this thing breaks? Why does it stay broken? What have people already tried? You do not need a research budget for this. You need a notebook and curiosity.
3. Decide whether AI is actually the right tool. Most problems are solved by simpler things — better forms, clearer processes, a phone tree, a spreadsheet. AI is sometimes the right answer. Often, it is not. The bias toward AI as a solution is a marketing problem dressed up as a technical one. Do not solve it by adding to the noise.
4. Build the smallest version that proves the idea. Not the impressive version. Not the demo-day version. The smallest version that shows the problem can be addressed at all. If you can solve ten percent of it for ten people, you are further than ninety percent of the engineers who built nothing because they were waiting for the perfect dataset.
What this actually changes
The young man I mentioned at the start went back home. A few weeks later he texted me. He had been spending two hours every morning with his uncle — a furniture maker who had been losing track of inventory and customer orders for years.
He had not written a single line of code yet.
He was learning the problem.
That is the start of every meaningful piece of software I have ever seen built. Not the certificate. Not the framework. The patient, unglamorous, sometimes embarrassing process of sitting with a real problem until you understand it well enough to know what to do.
What I want for them
I want the next generation of Nigerian builders to stop being recruited as international tool-users and start being recognised as problem-finders for a continent the rest of the world is only beginning to look at seriously.
That work is harder than learning a framework. It is also more durable. The tools will keep changing. The problems will outlast them. The people who learn to see the problems clearly will keep building no matter what tool the world is currently excited about.
" The tools will keep changing. The problems will outlast them.
The next time someone tells a young person to learn AI, ask a second question with them. Learn it to do what?
That second question is the one that matters.