Ask a founder in Lagos or Nairobi whether their team uses AI tools and the answer is almost always yes. Ask what changed as a result and the answers get much more specific, and much less breathless, than the sector's marketing would lead you to expect.
The consistent report is that code assistants meaningfully speed up routine work and do very little for the hard parts. Teams describe faster boilerplate, faster tests, and no change at all to the architectural decisions that actually determine whether a product survives contact with users.
The pricing problem nobody abroad mentions
Seat-based pricing set in dollars lands differently on a payroll denominated in naira, shillings or CFA francs. A subscription that reads as trivial to a company in Berlin can be a real line item for a team of eight in Accra, and several founders described rationing access — senior engineers get licences, juniors do not — which inverts who benefits most from the tooling.
Data and language
The second recurring complaint is performance on local context. Models handle English and French well and handle everything else on the continent poorly, which limits what can be built for users who do not operate in those languages. Founders working on voice, translation or customer support in local languages describe doing substantial work themselves to make general-purpose models usable.
That work is valuable and largely uncompensated. It is also, several pointed out, being fed back into products they pay to use.