Project by Projects

About a year ago, I wrote a piece after a series of conversations with African technologists, policymakers, designers, researchers, community builders, and people thinking seriously about what AI might mean for the continent. I was trying to make sense of something I could see happening across different countries and sectors, even though the individual efforts often appeared disconnected from one another.

The question I asked then was: What if our seemingly scattered efforts are actually the emergent nodes of a continental system we have not recognised yet?

I returned to that piece recently because so much has happened in a relatively short period of time. Reading it again, I would not ask exactly the same question today. Not because the original argument was wrong, but because some of the things that felt emergent then are considerably easier to see now.

There are more national AI strategies. There are more African researchers, founders, policy actors, language initiatives, public-interest technologists, infrastructure conversations, investment announcements, governance debates, and locally relevant applications. Questions about compute, data, energy, digital public infrastructure, public-sector capacity, language, regulation, and technological sovereignty have become the meat of the conversation today.

So perhaps we have moved beyond asking whether the nodes exist.

The more important question now might be: what would it take for those nodes to begin functioning as a system?

You see, a collection of impressive initiatives is not automatically an ecosystem, and an ecosystem is not automatically a functioning system. A fully functional system requires relationships, a steady flow of information, resources, capital, knowledge, people, infrastructure, and authority between its different parts. It requires us to understand where the dependencies sit, where the bottlenecks are forming, and where something important is being built in one part of the continent without enough visibility or connection to people working on an adjacent problem somewhere else.

That is the question I am interested in today.

When I wrote the original piece, one of the things that excited me was the growing confidence with which African actors were challenging the assumption that AI governance frameworks should simply be imported from elsewhere and adapted afterwards. There was a strong insistence that local priorities, political realities, institutional arrangements, languages, social conditions, and development ambitions should shape governance from the beginning.

A year later, that argument is clearly established. The African Union has a Continental Artificial Intelligence Strategy, more countries have developed national strategies and policy frameworks, and the conversation is now moving beyond whether Africa should have a distinctive voice in AI governance.

That is important progress. But it has also exposed the difference between designing a strategy and having the institutional machinery required to make that strategy work.

Once a country has an AI strategy, a new set of questions come up. 

• Which institution actually owns implementation? 
• How do ministries responsible for technology, education, trade, labour, health, agriculture, finance, justice, data protection, and national security coordinate when AI cuts across all of them? 
• Do regulators have the technical capacity to evaluate the systems entering their jurisdictions? 
• Do public institutions know how to procure AI responsibly? 
• What kinds of public datasets exist, and are they reliable enough to support the systems governments hope to build? 
• What happens when a successful pilot needs to become part of a functioning public service?

The answers to these questions will determine whether African AI strategies become meaningful. And by extension, fully functional.

A year ago, I was interested in whether governance could be reimagined from the ground up. I still am. But today I would add: Can our institutions actually operationalise the governance architectures we are now creating? This seems to be the place where the most strategic conversations have moved.

Last year, I cited figures suggesting thousands of AI companies were operating across the continent. The more I have read since, the more cautious I have become about numbers like these, largely because different datasets are counting very different things.

Some are counting companies using AI somewhere in their product. Others are looking specifically at businesses where AI is central to the company’s value proposition. Others measure only funded startups. Some include established technology firms, while others look only at early-stage ventures.

The numbers therefore tell us less than they initially appear to, unless we ask what exactly is being counted. Which brings me to another strategic question that Africa may need to confront more directly: What kind of AI economy are we actually building?

Are we primarily developing African companies that apply models and infrastructure created elsewhere? Are we building foundational technologies ourselves? Are we creating highly contextual applications for agriculture, healthcare, finance, education, public services, climate resilience, and language? Are we developing the datasets, compute, research capacity, and technical infrastructure required to shape the technology at a deeper layer?

The answer is probably that all of these things are happening at once, but unevenly. There is a difference between being very good at applying AI and having meaningful influence over the models, data and capital that determine how AI develops globally. Applied innovation can create enormous value, for sure and in fact, Africa’s ability to build for difficult operating environments may become one of our greatest contributions to global technology. However, contextual ingenuity should not become the go-to excuse to become permanently dependent on infrastructure controlled elsewhere.

Today, the conversation needs to hold two ambitions at once: 

1. Becoming exceptionally good at building AI that responds to African realities. 
2. Increasing African ownership of the foundational structures on which that innovation depends.

My biggest learning yet since had I originally described what I could see as a constellation of AI explorers, builders and skeptics, governance minds, and community operators. I still think those categories are useful, but today I would add another group: infrastructure stewards.

These are the people and organizations thinking about compute, cloud infrastructure, data centres, electricity, connectivity, public datasets, cybersecurity, interoperability, procurement, digital public infrastructure, and the physical systems that sit beneath the technologies we usually experience as software.

AI is often discussed as though it exists somewhere abstractly “in the cloud.” But of course it does not. It requires many things from servers to energy, to skills and institutions that can procure and govern it. And so when all of those things exist somewhere, and are owned by someone, they will be distributed unevenly. Once you start looking at AI through that lens, many conversations that previously seemed separate begin to collapse into one another.

The AI question becomes an energy question because compute requires reliable electricity. It becomes a financing question because digital infrastructure is capital intensive. It becomes a public-sector question because governments hold significant datasets and increasingly procure AI-enabled systems. It becomes a regional integration question because many African markets may not individually have the scale required to justify building every layer of infrastructure independently. It becomes a sovereignty question because control over infrastructure affects who can build, who can access systems, who sets prices, and whose rules apply.

It is starting to get uncomfortable to talk about African AI sovereignty that focuses primarily on models and regulation while treating infrastructure as background noise. If an African country develops an excellent AI strategy but does not have adequate compute, reliable electricity, accessible high-quality data, technical capability, or institutions able to procure and govern these systems, what exactly is the strategy going to deliver?

To be clear, this question is not an argument against ambition. It is an argument for systems thinking.

We need to keep in view, the relationships between AI policy, energy policy, industrial policy, education, public data, telecommunications, research funding, procurement, digital infrastructure, and regional cooperation. Otherwise, we risk designing strategies around only the most visible part of the system.

One part of my original piece that I feel even more strongly about today is language.

I had argued then that African language, context, and lived knowledge should not be treated as localisation work that happens after a technology has already been designed. They should be part of the foundation from which the technology is built.

Over the past year, there has been visible movement in that direction. More open datasets are being developed. African-language benchmarks are becoming more sophisticated. Speech resources, language-identification tools, and locally relevant datasets are being built across multiple initiatives. Work that once sat primarily in research communities is gradually becoming part of a wider infrastructure conversation.

That is encouraging, and yet I ask. Once these resources exist, who maintains them? Who funds the unglamorous work of continuously improving them? Who decides how they are governed? How are they connected to public-interest applications? How do researchers and builders across different countries know what already exists so that scarce resources are not repeatedly used to rebuild similar foundations? How do we make sure languages spoken by smaller populations do not remain invisible simply because they are commercially less attractive?

Language is not merely a user-interface problem. It is social and shapes who can interact with technology and whose knowledge becomes digitally legible, and therefore what kinds of information eventually find their way into the systems in education, healthcare, agriculture, finance, public services, and civic life.

We no longer need only to say that African languages must be represented. We need to start asking what kind of institutional, financial, and technical backbone will ensure that the language resources now being built become durable parts of the ecosystem rather than another collection of short-lived projects.

Capital has also returned more visibly to African technology after several difficult years in global venture markets. That is a positive signal, and now we need to move towards a healthy system. Among other things, funding remains concentrated because the most visible companies and institutions remain clustered within a relatively small number of countries and cities.

There is nothing inherently wrong with clusters. In fact, clusters are often how innovation ecosystems develop. Lagos, Nairobi, Cape Town, Cairo, Kigali, Accra, Tunis, Casablanca, Dakar, and other emerging centres can all become stronger and create spillover effects far beyond their immediate geographies.

The question I am asking from a strategic standpoint is what happens between the clusters? Do they become increasingly powerful islands, or do they become nodes in a system where research, infrastructure, talent, policy learning, capital, market access, public-interest knowledge, and technical capability move more easily across borders?

This is where I think Africa’s next AI challenge is. Strong nodes within a system  do not automatically translate into resilience. It will be determined by the quality of the relationships between those nodes. A brilliant AI research community that has little connection to policymakers will have limited influence on governance. A startup ecosystem without reliable infrastructure will struggle to scale. Regulators without access to technical knowledge will struggle to oversee rapidly changing technologies. Public institutions may sit on valuable data without the systems required to make it usable. Communities may experience the consequences of AI-enabled systems without meaningful access to the places where those systems are designed or governed.

So if you come up with me and take a bird’s eye view of things, you will see that the biggest gap may not be the absence of activity. It may be the weakness of the connective tissue.

At the end of the piece I wrote last year, I made a fairly simple ask: Let us document it. Align it. Map it. Make it actionable.

I still want that.

But the map I imagine today would be much more ambitious.

I would want to see the researchers and universities, the startups and established companies, the policy labs and regulators, the civil society actors and public-interest technologists. But I would also want the data centres, compute resources, cloud infrastructure, energy systems, public datasets, language resources, digital public infrastructure, venture capital, philanthropic capital, government procurement systems, regional institutions, skills pipelines, standards bodies, and the communities where AI-enabled systems are actually being deployed.

More importantly, I would want to see the relationships between them.

Who is talking to whom? Who funds whom? Who learns from whom? Which infrastructure is shared? Which knowledge travels? Where are the duplicated efforts? Where are valuable resources sitting almost invisible to the wider ecosystem? Where is a policymaker struggling with a question that an African researcher somewhere else has already spent five years studying? Where is a startup rebuilding something that could potentially become shared infrastructure? Where is a community experiencing the consequences of a system without being connected to the institutions governing it?

A systems map does not merely show us who exists. It helps us see the flows, dependencies, gaps, concentrations, and opportunities between them.

That is the map I want now.

A year later, I am less interested in whether Africa is “ahead” or “behind”

Finally, one sentence from my original article that I would definitely change. I wrote that the true narrative was that Africa was not lagging.

At the time I wrote that, too much technology discourse was positioning Africa as a place where innovation eventually arrives, rather than a place where knowledge, experimentation, governance thinking, and technology are created.

I still reject that framing.

But I no longer think the answer is simply to replace a deficit narrative with a triumphalist one.

The reality is far more interesting.

Africa can be leading in contextual innovation while remaining severely constrained in compute capacity. African scholars and policymakers can produce sophisticated thinking about inclusive AI governance while regulators still lack the resources needed to implement those ideas. We can build impressive applications while depending heavily on models, cloud infrastructure, and capital controlled elsewhere. We can have extraordinary technical talent while continuing to lose some of that talent to ecosystems that offer greater access to research infrastructure, funding, and professional opportunity. We can make meaningful progress on African-language AI while thousands of languages remain barely represented in digital systems.

All of these things can be true at once.

And perhaps strategic maturity begins when we stop needing the story to be either one of failure or one of triumph.

This brings me to the final question in this piece: What is the system telling us about where we are, and what needs to be built next?

A year ago, I wondered whether Africa’s scattered AI efforts were actually the emerging nodes of a continental system we had not yet learned to see. Today, I think many of those nodes are much more visible.

We have researchers. We have builders. We have policy thinkers and regulators. We have community actors. We have growing language infrastructure. We have investment. We have national strategies. We have increasingly sophisticated applications. We have growing conversations about compute, energy, public data, digital infrastructure, and sovereignty.

But having the pieces is not the same as having the system.

Perhaps the work now is not simply to create more nodes. Perhaps it is to intentionally build the relationships, infrastructure, information flows, capital pathways, governance mechanisms, shared resources, and trust that allow those nodes to function as part of something larger.

Last year, I wanted us to see the constellation.

This year, I am asking what it would take to make it a system.

And perhaps that is where the most interesting work is now.

Still Listening. Still Learning. Now Asking.