Two AI stories came out of the continent this week that on paper are both good news. Read them next to each other, though, and I started noticing a pattern I don't love.
Nigeria's Communications Commission ran its fourth Hackathon Live Show in Abuja, twenty innovators, one weekend, theme was "Technology Without Barriers." What they built was genuinely impressive: AI speech-to-text tools, real-time sign language translation, voice-enabled interfaces, the kind of stuff that could make a real difference for people with disabilities trying to use digital services. But buried in the coverage was the actual punchline: a lot of the platforms these tools need to plug into aren't accessibility-ready themselves. NITDA has guidelines requiring web accessibility standards. Plenty of government and private platforms just don't meet them. So you've got a brilliant sign-language translation tool with nowhere fully compliant to point it at.
Same week, on the other side of the continent, Mozambique launched AMIA, the Mozambican Association of Artificial Intelligence. This is the opposite kind of effort entirely, a coordinating body pulling government, companies, universities and civil society into one room to figure out AI policy together. It's the third step in a deliberate sequence, national strategy in 2025, a National AI Commission in March, now this broader coalition. That's a genuinely thoughtful, staged approach to governance, way better than a lot of countries that just wake up one day and pass a law.
Here's what's bugging me. These are two different layers of the same continent's AI moment happening completely independently of each other. One is builders, moving fast, hackathon-speed, producing real working tools in a weekend. The other is governance, moving deliberately, institution-speed, building coordinating bodies and commissions over the course of a year and a half. Both are legitimate and necessary. But nothing in either story suggests these two tracks are actually talking to each other, and I think that's exactly the setup that produces the NCC hackathon's own problem: brilliant tools with no ready infrastructure to land on.
Think about what AMIA is actually for, coordinating government, industry, academia and civil society on AI policy. If that coordination is happening at the level of frameworks and strategy documents, and meanwhile the actual builders, the twenty people who spent a weekend in Abuja building things regular people could use tomorrow, are operating in a completely separate lane that never intersects with the policy conversation, you end up with exactly what we're seeing. Good policy thinking that never gets grounded in what builders are actually running into on the ground. Good tools that never get the platform-level standards enforcement that would let them actually work.
I don't think the fix here is complicated, and that's what makes it frustrating. It's not that Africa needs more strategy or more hackathons, both of those are working fine on their own terms. It's that whoever's running these AI coordination bodies, AMIA and whatever comes next in other countries, needs the actual builders in that room, not as a token gesture but as the people telling them what's breaking in practice. NITDA's accessibility guidelines existing on paper while platforms ignore them isn't a hackathon problem, it's an enforcement and coordination problem, and that's precisely the kind of thing a body like AMIA is supposed to exist to fix, if it's actually talking to the people building things instead of just the people writing policy about things.
Africa has real momentum on both fronts right now, genuine builder energy and genuinely thoughtful governance instincts. I just don't want to look back in two years and find out we built two impressive, well-funded, well-intentioned things that never once had a real conversation with each other.