AI Won't Save African IT — Operators Will
The model is not the moat. The discipline of running things in Lagos, Nairobi, and Accra is the moat.
There is a particular kind of optimism circulating in African tech right now that worries me. It goes like this: the new generation of foundation models will collapse the cost of software, automate the tedious parts of operations, and let the continent leapfrog a generation of enterprise IT we never built. Some version of this argument is correct. Most versions of it are dangerously wrong about which problem actually needs solving.
African IT does not have a model problem. It has an operating problem. The bottleneck on a Lagos SMB is not 'we couldn't generate code fast enough'. It is that the warehouse manager and the accountant disagree about what is in stock and there is no system of record either of them trusts. The bottleneck on a Nairobi mid-cap is not 'we needed a smarter chatbot'. It is that the four core business systems do not share a customer ID, so every report has to be reconciled by hand. The bottleneck on an Accra broadcaster is not 'AI editing'. It is that the live ingest fails every third Sunday and nobody is on-call to fix it.
None of these problems are solved by adding a model. They are solved by an operator — a human who owns the outcome end to end, who writes the runbook, who makes the unsexy decision to standardise on one accounting platform, who insists on a single source of truth for the customer record, who builds the on-call rotation. AI is a useful tool inside that operator's toolkit. It is not a replacement for the operator.
I think the next decade of value creation in African enterprise IT will go disproportionately to the people who can do the boring work — process design, data hygiene, change management, observability, vendor management — in environments where the network is unreliable, the talent pool is uneven, and the regulatory ground is shifting. The AI layer will be a multiplier on that work. It will not be a substitute for it.
If you are a founder reading this and choosing between hiring 'an AI engineer' and hiring an operator who has shipped two ERP rollouts in Nigeria and is willing to live in your warehouse for a month — hire the operator. The model will still be there in six months. The operator with that calibre of context will not.