However you look at it, the future of AI and its effect on humanity is daunting. No one, including the leaders at OpenAI and Anthropic, has been able to predict with certainty the extent to which it will alter our society. But so far, we know this: it will make our work faster, some of us will have to learn new skills, and those who do not understand how AI works risk condemning themselves to damnation.
In Africa, adoption and tech savviness have been slow, and this is already a concern. But more concerning is the region’s participation as a key player in producing innovative AI products that can compete globally. Yet it is not too late to begin to look past chatbots for the tech that would define the future.
Justin Irabor is an AI product engineer and one of the brightest builders from the region in his generation. He has spent the last few years thinking about and researching new use cases for agents. Kunle Akinola has spent the last few years investing in tech companies on the continent. He is an investment professional who, among other things, spends his time thinking about AI. David Oks has made a name for himself on Substack for his commentary on the current state of innovation. His viral essay “How funerals keep Africa poor” has become one of my recommended reads on African development.
In this written roundtable, they shared their ideas for how Africa can be better prepared to make the most of AI. The conversation is below:
1. Without itemising all our problems, do you think AI can turn things around for Africa?
Justin: AI is mostly inert technology that becomes reactive to the intent of its orchestrator, so I avoid talking like it’s a genie we can rub and poof, out comes solutions. The African problem (to the extent that we can even articulate it) isn’t one that was bottlenecked by intelligence, so it’s not going to be solved by having mechanised data modelling on tap.
It is entirely possible, however, that the existential threat of foreign parties controlling AI will force Africa to seriously consider the implications of its own sovereignty today, but one can only hope.
Kunle: Absolutely, and I think Justin undersells it slightly by framing AI as inert. Yes, it’s not a genie, but capital doesn’t flow to genies; it flows to productivity, and that’s exactly what AI unlocks. Whether Africa is a net consumer or a part-producer, I believe it will transform banking, access to credit, manufacturing, education, agriculture, defence, and healthcare, both directly and indirectly. We’re already seeing it: scoring algorithms in fintech, healthcare administration and research, production lines, education. If adoption keeps compounding, the productivity gains could be astronomical.
And here’s the part that makes me more optimistic than most developed markets: we have little of the job-displacement fear that paralyses the West. Our companies are expanding, not contracting, so there’s enormous room to absorb AI as augmentation rather than replacement. Coming off a low economic base, that same technology can start closing real demand gaps in critical industries. Also, where I think the impact could be greatest is public governance, but that’s precisely where technology hits its ceiling, because none of it matters without the will to deploy it.
David: I agree with Justin that Africa wasn’t bottlenecked by intelligence: most of the problems that Africa faces have pretty obvious solutions—“here’s how to develop a population’s human capital,” “here’s how to run an electric grid,” “here’s how to manage a currency”—but the difficult part is actually getting them done in very difficult political conditions, which isn’t really an intelligence problem. So I don’t think that AI itself will fix Africa’s problems.
However, I do believe that AI will make the world much richer, whether over the course of years or decades: I do think that we’re entering the world of robotic factories and autonomous scientific discoveries. That will make Africa much richer in the process—more efficiency, production, and abundance are ultimately good for everyone—without really solving the specific problems that African countries deal with. (I’d argue that the consumer surplus from wide diffusion of AI models, have made people effectively richer than they are today: I can now learn about things much more cheaply than I could before!) The main thing African countries should worry about, in my view, is making sure that they’re in a position to benefit from this diffusion—I sometimes worry that regulation and state intervention, for example, could slow or halt this process.
2. We are still in the thick of the AI revolution, but already it seems like the continent of Africa is swimming against the tide. Is there any course correction that can turn Africa into a key player?
Justin: Africa is too sprawling for me to even wrap my head around what a course correction might look like. There are too many moving parts, various misaligned players and interested parties, and Africa is highly responsive to/affected by foreign interests (think US and China) who are motivated to not let Africa operate in a vacuum as far as the new Cold War (this is what I call the frontier lab wars) is concerned.
So, yeah, it’s difficult to imagine Africa’s hand in this game of intelligence poker, because Africa is too constrained and manipulable. I wish it weren’t true, but it is.
I’d rather speak of Nigeria. A similarly opaque base, but at least it’s smaller (enough to reason about), and it’s also familiar to me. For Nigeria, the playbook seems simple to me that I might as well spell it out cleanly:
We need to actively put money in the hands of the best people building and experimenting with these technologies, and
We need a list of problems that could theoretically be solved (or at least ameliorated) by the current AI/ML technologies today, and
We need a way for stakeholders across disciplines to share knowledge about their fields and collaborate.
The government should only provide the means for the above to happen, in my opinion. The government is the only instrument we have for ensuring the efficiency and flywheel for inter-disciplinary collaboration. This means the following:
Sourcing for and providing an abundance of local compute
Incentivising the private sector players to care about communal breakthroughs
Organising referendums and theses (see: America’s Federalist Papers, precursor to the Constitution) for crafting a clear vision for the nation, planned in decades and measured in bi-annual intervals.
We need a national psychosis (pardon my French), a fervour we haven’t seen in this country in such a long time. A crystal-clear vision for what the future may look like if we harness this technology as a collective to potentially lift millions of people out of ignorance, poverty, disease and starvation is enough to fill even the most disillusioned with, ahem, renewed hope.
Kunle: I guess my response is, are we really swimming against the tide, or are we just being Africans about it? The continent has never been a key player in any major technology advancement, so why would AI be different? The way I like to frame the AI value chain as an investor is, borrowing from Jensen Huang, the AI value chain runs on energy, chips, infrastructure, models, and applications. We are not competitive in any of those stacks. Can we play a meaningful part in some? Absolutely. Can it be at a globally recognisable scale? Definitely not, at least not in the next foreseeable decade or two.
So I’d rather we focus on what actually matters to us right now, which is economic development as a whole. Africa’s challenges and aspirations are fundamentally different from those of the US or China. Where do I see low-hanging fruit? First, in the “application” layer. We need to adopt AI in ways that translate into long-term, consistent double-digit economic growth. Second, in hybrid sovereign compute, where African companies with highly sensitive data outsource the reasoning to foreign models but keep the underlying data within local boundaries. This hybrid mechanism can and should lead to a gradual development of other parts of the value chain, over time. We’re already seeing early signs of this with the recent Central Bank of Nigeria (CBN) guideline requiring housing transaction data to sit in local data centres.
David: I don’t think there’s really a way for Africa to be very important to the AI stack (similar with India, Latin America, and most of the world). As Kunle said: the crucial inputs are energy, chips, models, and applications. I don’t see a way in which Africa can really be world-competitive in any of those, save perhaps local applications. (There’s also, perhaps, the prospect of building massive solar panels in particular parts of the continent, like the Namibian desert, to power massive data centres; but that would require a lot of things that aren’t really on the offer today.) That’s fine—it’s not catastrophic not to be at the centre of a technological transformation, and I do think a rising tide lifts all boats—but it’s important to be realistic about what Africa can offer and what it can’t.
3. Sometimes it feels like we are speaking about AI in a less exciting way, with a heavy focus on using ChatGPT to pursue mediocre writing careers. Any revolutionary use of AI, aside from Terra Industries, has certainly eluded me. How can Africa avoid becoming merely a consumer of AI developed elsewhere?
Justin: This is one of those moments in history where being a consumer of the technology isn’t that big a deal, at least to me. Only a few people in the world can innovate on this technology, and they’re locked in billion-dollar contracts at frontier labs.
This technology is like fire: we shouldn’t care about making our own fire, but about harnessing it to bring light, heat, and energy to our homes. On this, I agree: thinking of the technology purely as “ChatGPT” is a rube’s perspective, yet I’m not mad at it. I don’t expect the average person to know about this technology beyond a chat interface. The builders should, and do (case in point: me). The technology as it exists today, however, is both expensive and punishing of error, which means if we are to gain true mastery here, this is the sort of thing labs, accelerators, and grounds for experimentation are built for. If we cannot properly articulate what we want to solve with technology, we won’t be able to see the contours of the technology either.
Because the technology is expensive to deploy and experiment with, builders in Nigeria have to follow the path of least resistance (ie, use it in ways that have been proven to yield revenue), which is why you may be disillusioned by the case studies you see in the wild. Yet, if you peer deeper, you’ll see people deploying the technology (even today) to impressive ends.
(A small, potentially contestable claim: because the technology is pricey, Nigerian or African builders in the AI space may be averse to building for the continent, instead choosing to build locally but for an international audience. A form of brain drain.)
Kunle: I agree with Justin here. Being a net consumer isn’t a bad thing in itself, as long as we’re adopting AI to genuinely improve local economies across the continent. As I said earlier, the way we avoid becoming a mere consumer is by owning the application layer, deploying AI in ways that translate into sustained double-digit growth, and by pushing hybrid sovereign compute so that even where we outsource the reasoning to foreign models, the underlying data stays within local boundaries.
I’d add one thing. The interviewer is right that Terra Industries stands out, but notice why it stands out. It’s vertically integrated deep tech built for a real, high-stakes African problem, and it attracted top-tier global capital precisely because of that. That’s the pattern worth internalising. The revolutionary African AI stories won’t come from generic consumer tools; they’ll come from narrow, high-value verticals where local context is the moat and the problem is expensive enough to be worth solving. My job as an allocator is to fund more of those, not to mourn the fact that most people only see AI as ChatGPT.
David: I agree with Justin and Kunle that being a net consumer of AI probably isn’t a very big problem. Pursuing indigenous production of AI seems to me like a fool’s errand for almost all of the world’s countries: it is simply extremely complex and capital-intensive. For better or worse, all of that is best left to the U.S., China, South Korea, Taiwan, and a small number of other places.
I think African countries in particular would be smarter to pursue lower-hanging fruit. There are plenty of local and domestic opportunities in the Terra vein; this will become particularly true in the years to come as robotics becomes more of a practical opportunity, and once we start seeing AI-driven scientific breakthroughs. These are the types of opportunities that African entrepreneurs are best-suited to pursue.
4. Is the chatbot race over, considering ChatGPT and Claude’s dominance? Would you bet your money on a new African chatbot, and why?
Justin: Hah, I wasn’t even aware there was a chatbot race. Chatbots are mere interfaces wrapping your model of choice with the context required to do the job. In that regard, different chatbots may be suited for different jobs due to substrate/context lock-in. Nothing’s going to beat Google’s Gemini at parsing context across my Google, Gmail, location data, YouTube, and browser histories, for example, because no one has the ability to natively stitch that data together like Google will.
And while Claude Code may be superior to OpenAI at code-related tasks this month, my IDE and codebase are portable contexts, so I can always switch between either provider whenever one or the other records a benchmaxxing breakthrough :)
Seen in this light, then, the context is what matters when you have hit a baseline of model competence, so if we want African chatbot dominance, we must ask ourselves: where is the African context?
I suspect there is value in framing this as narrow-domain chatbots (though why chatbot specifically?) within African contexts (education, healthcare, legal, logistics and commerce, etc).
Kunle: Easy answer. No, I wouldn’t. Why not? Because I’ve yet to hear a convincing value proposition for one. What exactly would an African chatbot do better, and how would it realistically stay competitive against models that are improving every few weeks? I’ve stumbled on more than a few “Africa-focused chatbots” already, and honestly, they’re easy passes for me. We don’t need to reinvent this wheel. The smarter play is integrating outsourced models into existing processes and applications to drive real efficiency gains across key industries.
Justin makes the right point that context is the moat, not the interface. I just take it one step further. If context is what matters, then the winning African companies won’t be the ones building a chatbot at all. They’ll be the ones embedding these models deep into workflows in banking, health, logistics and commerce, where the proprietary data and distribution already sit. That’s a far better place to put capital than another chat wrapper.
David: I agree with Kunle. I don’t think there will be widely-adopted Africa-specific chatbots—there might be LLMs that manage to beat ChatGPT and Claude in, say, particular languages that aren’t covered much in their corpuses, but I’d be very surprised if they manage to beat the major chatbots in terms of real-life utility—accomplishing things for you, getting things done—simply because it’ll be exceptionally difficult for local companies to match the resources that Anthropic, OpenAI, and so on can muster. Selling locally-specific data to the major AI companies (a model pioneered by India’s Karna) to give them valuable local context seems like a better opportunity to me.
5. What kind of infrastructure—digital, physical, or regulatory—is most urgently needed for an AI economy?
Justin: We need compute. Lots of it. We also need data, and we’ve barely begun to do the work to collect reams of this stuff. We’ve lost historical data, and are currently losing real-time data.
Kunle: Honestly, the truthful answer is all of it. We need compute, power, data and sensible regulation, and Justin is right that compute and data are the binding constraints. But I try to be realistic about what an African government or private sector can actually move on in the near term, so I’d focus on the light-touch, high-leverage wins first.
On the physical side, none of this works without reliable power, so energy is the real bottleneck beneath the compute conversation. That’s a long, capital-heavy fix. In the meantime, the realistic move is regulatory: get data-residency and cloud rules right so that African companies can safely use outsourced compute while keeping sensitive data local, the hybrid sovereign model I mentioned earlier. That’s cheap, fast, and unlocks adoption without waiting a decade for domestic data centres. On the digital side, the quiet priority is data itself. Justin’s point that we’re losing historical and real-time data is the one that should worry us most, because you can rent compute tomorrow, but you can’t recover data you never captured. So if I had to sequence it: fix the regulatory rails now, treat data collection as urgent, and build toward power and compute over the longer horizon.
David: I think the best thing that can be done for African countries in this context is simply “all the stuff that they should’ve been doing anyways.” That means better regulations, more investment in human capital, and better basic infrastructure. I’d actually disagree with Justin and Kunle and put compute relatively low on the list for now. Compute is really expensive and will only get more expensive; I think it’d be a huge waste of money for relatively poor countries to send tons of money to American, Korean, and Taiwanese companies for high-performance computing equipment. Focus on the low-hanging fruit! The things that make sense in the AI era aren’t so different from the things that made sense before.
6. There is a version of events where we outsource our thinking to AI, and we become duller for it. Do you ever worry about what this might portend for the region?
Justin: How much worse do you reckon it can get? It might be my own cognitive bias on display here, but Africa-wide, this isn’t a problem I take very seriously. We are poised to become much smarter, on average, than dumber from interacting with this technology. For one thing, if Africans are using AI chatbots en masse, it’d mean a lot of us can read, or at least understand what the LLM is saying. That sounds like a good thing. Cognitive offloading suggests the presence of cognition a priori.
Kunle: I echo Justin’s point completely here
David: Yes, agreed with Justin and Kunle—I think a world where ordinary Africans can interact with AI chatbots in huge numbers is a good world! I’m not really worried about cognitive degradation due to AI use in poor countries. Yes, it’ll be something that we’ll have to deal with, just like we’re dealing with the consequences of phone addiction; but I think it’s worth the trade.
7. Are there any actual AI products in development on the continent that actually excite you, and why?
Justin: None that I can think of for now, but this isn’t an indictment on the continent. I really haven’t been paying close attention, as I’m currently studying and building (tough to pay attention to the ambient preoccupations of others in this mode). I am, however, interested specifically in edtech and healthcare AI startups, and I’m working on something that may be interesting in the healthcare space.
Kunle: Yes, I’ve seen a few, but they’re largely at idea stage. Most are chatbot-like, trying to wrap things in an African context, and honestly I think they have a bleak future. What I’ve seen more of, and what actually encourages me, is AI being integrated into existing products and tech. That’s where I think the real value is, and where we can score quick wins: the application layer, not standalone products.
David: Yes, none that I can think of—I don’t follow the space too closely, but I haven’t really been impressed by the ideas that African entrepreneurs have come up with when it comes to AI. There are some interesting ideas surrounding robotics and autonomy, but those seem far-off to me. There definitely does seem to be a major opportunity around AI edtech—people haven’t paid enough attention to what “free brilliant tutors for everyone would mean”—but I’m not sure if that’s better served by the major AI companies or by local startups build on top of them.
8. If you had to advise an African government about making the most of AI, what would you say to them?
Justin: I’d say “give your innovators everything they need to attempt to augment the state’s infrastructure with technology, and get the hell out of their way. Fund safety and implementation researchers. Increase allocations to schools. God be with thee.”
Kunle: First, put regulation in place that actively forces the development of the local industry, the way we saw with the CBN guideline on keeping housing transaction data in local data centres. Smart rules like that create demand for local infrastructure and capacity almost overnight, without government having to build or fund everything itself. Second, drive good governance that can unlock massive investment in power, at least for starters. Nothing in the AI stack runs without electricity, so fixing the enabling environment for energy investment is the highest-leverage thing a government can do. Get the incentives right, protect our data, unlock power, and let the private sector do the rest.
David: Deregulate, deregulate, deregulate. Get the basics right—focus on human capital, infrastructure, fundamental state capacity—and let private individuals do the rest.
9: What do you think we should pay attention to that we are not currently focused on enough?
Justin: I would have said “we should pay attention to the mass knowledge transfer, via RLHF, of human capital from Africa to the West, encoded permanently in trained models”, but that ship’s sailed. Africans were paid peanuts to train the intelligence that has now made them less useful in the global workforce, but that’s the way of the world. What we need to pay attention to is the value of African human capital. We need thinkers and visionaries who can learn how to harness it for our own benefit, otherwise, the pillaging will continue until morale improves.
We have a sprawling youth population, and we don’t know what to do with them. The West did, though.
Kunle: Honestly, we don’t pay attention to a lot of things because of the huge misalignment across the fabric of our societies. Government, governance, people, capital, systems, all of it pulling in different directions.
But seriously now, here’s my real point. We don’t need to be sad about losing a race we were never in. We just need to keep finding ways to use this unprecedented technology to develop local industries. And that responsibility doesn’t sit with government alone. It rests on all of us: as individuals, as employees, as entrepreneurs, as capital allocators, and yes, as government too. Everyone has a part to play, and waiting for someone else to move first is how we stay behind.
David: A true AGI world is a world where “what’s in the training corpuses of the leading models” matters immensely. Cultures and societies that aren’t reflected in the training weights will be systematically disadvantaged relative to the cultures and societies that are reflected in them. So try to get Africa into the weights.
10: What skills should every young African start learning today, regardless of their profession?
Justin: We all need to learn how the world really works.
Kunle: Learn beyond borders. The single most valuable skill now is the ability to think, work and compete globally, because the tools finally let you. Refuse to be confined by where you happen to sit in it.
David: How to learn. The world we’re entering demands flexibility and adaptation; luckily, it’s easier to learn than ever. All the knowledge of the world is at your fingers. You just have to be able to learn!





