A Cabinet meeting on Monday, June 8, approved the establishment of a National Artificial Intelligence Agency, which will be Rwanda’s first institution dedicated to AI. Rwanda is trying to do something no African country has quite managed: not just use AI, but shape it. The government has put money into an AI Scaling Hub. The national policy commits most of its attention to training Rwandans to build these systems, not just operate them. The new agency pulls together government, companies, universities and development partners under one roof. On paper, this is exactly what was needed. The problem is that Rwanda is walking into a fight that started long before any of this was possible. ALSO READ: Rwanda’s AI Agency is timely, but delivery will matter most Every major technology of the last two centuries arrived in Africa after someone else had already decided what it was for. The question is whether artificial intelligence will be different. ALSO READ: Inside the newly approved National AI Agency Start in 1891. The first telephone service in Bloemfontein, South Africa, connected the railway office to the municipal building. Not the market. Not the community. The technology came to the continent to serve the people running it, not the people living in it. The telegraph, the telephone, the early internet — each arrived through infrastructure built to carry information out of Africa and instructions into it. Colonial telecommunications connected the colony to Europe. African towns were not connected to each other. When independence came, those networks remained. They pointed the wrong way. ALSO READ: How Rwanda is regulating artificial intelligence Mobile phones changed things, not because of wise planning, but because the old telephone system was built. There were no copper lines to protect, no companies to lobby against competition. The gap was so large that the only way was forward. MTN, Safaricom, M-Pesa — these became African stories precisely because the colonial system failed to lay the groundwork for anything else. The leapfrog that economists celebrate was built on an absence, not a foundation. AI is arriving into a different situation. The infrastructure exists. Rwanda is already inside it. The problem is that it was built without Rwanda's data, Rwanda's languages, or Rwanda's problems anywhere near the centre. Africa has over 2,000 languages. A study published in 2025 found that only 42 appear in any meaningful way across major AI language models. Only Amharic, Swahili, Afrikaans, and Malagasy are handled with any real consistency. More than 98 percent of Africa's languages are, for these systems, invisible. Ikinyarwanda is in that 98 percent. This is not a detail for engineers to sort out later. When an AI system advises a bean farmer in Musanze using data from fields in Iowa — different soil, different rainfall, different pests — that advice will be wrong in ways the farmer may not be able to identify. When a moderation system cannot read Ikinyarwanda, it cannot tell whether a post is dangerous or ordinary conversation. When a hospital uses a diagnostic tool trained on patient data from another continent, the tool's confidence may have nothing to do with accuracy. These are not edge cases. They are what happens every day when systems built elsewhere get deployed across our hills. Rwanda's new agency correctly named what the problem is. That is not a small thing. Most governments are still debating whether AI is relevant to them. Rwanda has already moved past that question and is building the institution to answer it. The agency's job is to grow local capacity — researchers, data, and solutions built around Rwandan realities. The $17 million in the AI Scaling Hub is a solid start. The policy framework is serious. The political will is there in a way it isn't in many comparable countries. Rwanda has been here before, identifying the right direction before the resources quite match the ambition, then closing the gap through persistence. The companies that built the dominant AI systems are not pausing to let anyone catch up. The models going into hospitals, schools, farms and government offices across Africa right now were not built with this continent in mind, and every month they embed deeper into systems that will be hard to replace. The window to shape what gets built, what data gets used, whose problems get centred — that window is open today. It will not stay open. After the 1994 Genocide against the Tutsi, Rwanda spent years building institutions the world said could not exist. The AI Agency is another such institution. The difference is that in 1994, the clock was visible to everyone. This time, it is easier to miss. The machine is already running. Rwanda just decided it would rather be inside it than underneath it. The writer is a professional engineer specialising in process design.
