If AI prompts can build your product, will African VCs still invest?

Artificial intelligence (AI) has reshaped how the world works, and African AI startups, venture capital and tech ecosystems are no exception. The technology is evolving so fast that many investors are still working out how to evaluate AI startups, let alone how to fund them.

Over the past six months, this column has spoken to African investors across regions, sectors, and funding stages about where AI creates genuine value, where it falls short, and what merits institutional capital.

They are Babacar Seck, founder of Askya Investment Partners; Eloho Omame, partner at TLcom Capital; Emmanuel Adegboye, head of Madica; Magnifique Ishimwe, fund manager at the Development Bank of Rwanda; Husein Merchant, investment officer at Village Capital; and Benjamin Toulouze, head of corporate venture capital at AXIAN Investment.

When we asked them what makes an AI startup fundable, they mostly described something an AI model itself cannot supply. Anyone can build on the same foundational models now, so investors are paying for what sits around them: voice products that work in Wolof and isiZulu, software tied so deeply into a customer’s systems that leaving is painful, licences that take years to win, and revenue.

Voice is where they see the clearest opening. Adegboye says the opportunity lies in building on top of foundational models with local data, and above all in voice. He points to a man who speaks only Wolof: what is the easiest way for him to take part in an economy increasingly run by software?

Seck makes the same point about South Africa, where census data shows only 8.7% of citizens speak English at home.

Building software has also never been cheaper, and the number of AI startups has climbed as costs have fallen. TechCabal Insights counted 207 African AI startups across 17 countries in 2025, roughly double the 104 it tracked in 2022. Most of the newest ones work in software development, business intelligence, finance and education, the areas where off-the-shelf models lowered the barrier most.

When code is cheap, investors look for what rivals cannot copy. Eloho Omame of TLcom Capital compares this moment to the arrival of mobile technology in Africa. Once it becomes easy to build the thing, she says, companies defend themselves in other ways: brand, network effects, and the cost to a customer of pulling out software that runs their business.

The following responses are drawn from a series of interviews TechCabal conducted with investors between April and September 2026. They have been curated by theme and edited for length and clarity.

Babacar Seck. Image Source: Askya

Why context is an advantage

Local context decides whether an AI tool works in Africa. Founders who have it can build for specific problems, as Pewbeam AI has done with software that projects Bible verses during live church services, and as YarnGPT has done with a speech model for Nigerian-accented English, Yoruba, Igbo and Hausa. They can also make something as viral as Lagos Life.

Of the 207 African AI startups TechCabal Insights counted across 17 countries in 2025, two-thirds are still early stage and only eight have reached maturity. When that many companies can plug into the same foundation models, saying a product is AI-powered proves very little.

That changes what investors ask founders. If AI keeps making software cheaper to write, software on its own protects nothing. What matters is everything around it: data nobody else holds, local knowledge, a way to reach customers, regulatory permission, and how deeply the product sits inside the work a customer already does.

Merchant of Village Capital, whose fund has not yet looked at AI as a sector, expects most of the African opportunity to sit at the application layer.

Husein Merchant (Village Capital): My personal opinion is that there is going to be a lot of opportunity on the consumer side. Given that Africa may be more of a consumer of AI technology than a creator of it, I think the investments and solutions will lean that way.

Eloho Omame (TLcom Capital): AI does not make everything cheaper to build. It is not a panacea. Some things are cheaper. Testing is cheaper, so you can experiment far more quickly with LLMs than before. Whenever a technology has driven down the cost of delivering something, the shape of defensibility has changed. A company could no longer rely on a high cost of entry. AI has lowered the cost of entry the way mobile technology lowered it before. When we look at a company, we ask how it builds defensibility if testing or delivery is now commoditised. It might be the brand. It might be network effects or switching costs.

Eloho Omame. Image Source: TLCOM

**Data and language **

Seck splits AI startups in two. One group builds the layers of AI itself. The other applies AI to ordinary businesses, and that is where he sees an opening for Africa.

Speech data for African languages is scarce. A company that collects it, trains on it, and builds a customer base around what it can do with it owns something rivals cannot easily copy. Caantin, a Zambian startup that replaces call-centre staff with AI voice agents for banks and fintechs, told TechCabal in June 2025 it was approaching $1 million in monthly revenue, charging by the volume of calls it handles.

No one has fully cracked it yet.

Babacar Seck (Askya Investment Partners): Wherever you have structured data at scale, you can build small AI models that improve performance. That is exciting in the African context because we are digitising in the age of AI. We are creating structured pools of data at the same time as we apply AI to them, so companies build a competitive advantage much faster as they scale. Language is the clearest case of data that is hard to copy. In South Africa, Census 2022 found that 8.7% of people speak English most often at home. IsiZulu (24.4%), isiXhosa (16.3%), Afrikaans (10.6%) and Sepedi (10%) all rank above it. A product that only works in English is built for a small slice of that market.

Babacar Seck (Askya Investment Partners): One very exciting area is voice AI in African languages. Many startups operate in that space, and none have fully cracked it. Today in Africa, most people do not speak English, French, or Portuguese at home. They speak their mother tongue, the language they are most comfortable in. The minute you have an application that lets companies talk to their customers in the customer’s own language, you bridge a divide. That applies to companies and their customers, and to governments and their citizens.

Emmanuel Adegboye (Madica): I think daily about people in rural areas. What is the easiest way to take part in the global economy if you only speak Wolof? The fastest way is to build platforms that speak Wolof or deliver information in Wolof, so you can do all of that without needing to learn English. That is why I think voice is an important long-term opportunity in AI, and it is why some of our investments have been in voice AI.

Emmanuel Adegboye, head of Madica. Image Source: Madica

Being one with the customer

Omame’s portfolio has shown her a second kind of defence: a company that sits inside a customer’s daily operations. Once its software is wired into thousands of end customers, a rival cannot easily replace it, and the margin that integration earns is protected with it.

Eloho Omame (TLcom Capital): We invested in an API infrastructure business. Its API plugs into its customers’ software and delivers a fintech product. Before, each time the customer needed the product, it made an API call. Now, because of agentic technology, the system prompts the customer when they are about to run out of the resource they lease. Revenue goes up, and the company also knows its revenue is going up. The product also costs less to deliver, because there is an agentic layer on the production side too.

In principle more people can now build what it builds, as cheaply as it can. But once our software is talking to each other and I am delivering value across thousands of your own customers, it is much harder for you to unplug from my system.

I want a clear line of sight to how AI makes the product cheaper to create, and what happens to the margin that creates. The founder can pass it all to the customer or capture it in the business as a higher margin. As an investor, I want to see it captured in the business.

Revenue comes first

Defensibility answers only half the question. The other half is whether the company earns money while investors are still willing to wait, and on this continent the wait has shortened.

An AI company that spends three years on compute before charging anyone fails that test, however good the idea.

Magnifique Ishimwe (Development Bank of Rwanda): Basic technology can always be built. I could put together an app like the one I showed you by the next morning. The fundamental question is whether you can execute.

I have done a couple of AI deals, and sometimes we cannot see when the company reaches its horizon. You can absorb two or three equity rounds, and we still do not see a path to profitability. In the United States you can be an Uber that loses money for six, eight, or ten years. In Africa, every investor asks where your revenue is, and if you do not have it, what value you have in its place.

Think of an AI company that builds a chatbot for students, but nobody is willing to pay for it. You spent a lot on compute, we helped you get credits, you paid expensive engineers, and for three years nothing came out of it. Those deals are misaligned with our instrument, because we are not an equity investor.

Some investors back AI infrastructure

Some investors do fund the deeper layers, usually for strategic reasons. AXIAN, the Madagascar-based group with telecoms and digital banking operations, frames its infrastructure bets around who controls the data.

Benjamin Toulouze (AXIAN Investment): We want to invest in AI infrastructure, and we have already done so. One example is WideBot AI, a proprietary LLM system for Arabic speakers. We invested in Nucleon Security, a Moroccan company that protects devices using zero trust, EDR, and malware tools.

We are convinced about the sovereignty issue. African countries need autonomy and control over their own data. That is one of the biggest challenges in the African market, because it is fragmented across 54 countries.

Emmanuel Adegboye (Madica): There is also an opportunity in hardware and AI. That is the bet we are making with ChipMango and chip design, and we will see how it evolves.

Madica joined ChipMango’s $1.9 million seed round, led by Atlantica Ventures, in September 2026. Even these bets follow the same logic as the rest of the column. WideBot’s edge is Arabic dialect data and a book of more than 350 government and enterprise clients. ChipMango recruits engineers through partnerships with Nigerian universities, trains them and sells chip-design work to global customers. In each case the hard thing to copy is the data, the talent pipeline or the customer list.

The cost of inaction in AI investing

If African investors back only the application layer, the deeper layers of AI will be financed, built and owned elsewhere. Global capital is already moving that way.

Capital is not the only gap. African data centres are chasing customers while the continent says it needs more of them, and universities are training AI engineers who leave for jobs abroad.

Emmanuel Adegboye (Madica): Capital is concentrating in a handful of funds and a handful of startups. Most funding globally is going to AI infrastructure companies, and only a few of those. On top of that, most of the funding coming into Africa comes from outside Africa, which exposes the continent to that risk. If I am a Silicon Valley investor who is not close to the continent, Africa is not top of mind when I invest.

Babacar Seck (Askya Investment Partners): Most data centres are chasing customers, yet we keep saying Africa does not have enough data centres. One gap is that these operators are not plugged into demand. When you speak to the global cloud players, they will tell you their biggest customers in Africa are startups, with some scale-ups spending upwards of $30 million a year on cloud.

We are training a growing number of AI engineers at Carnegie Mellon in Kigali, at the African Institute for Mathematical Sciences, and at some very good Nigerian universities, and many of them find jobs abroad because the companies on the continent are not yet large enough to keep them.

For now, the investors in this column have answered the question they can control. They will pay for the parts of an AI company that grow more valuable as the AI underneath gets cheaper. Whether the continent also builds that underlying layer is a question their capital cannot settle on its own.

True scale demands moving beyond surface-level integrations to robust execution. We’ve filtered the noise out of Moonshot 2026, optimising the conference strictly for high-calibre connections between startup founders, global financial operators, enterprise leaders and individuals rewiring Africa’s technical frameworks. Get 20% off Early Bird tickets for a limited time.