By Roger B. Jantio*
Africa has no shortage of institutions with problems artificial intelligence can help solve. Banks want to detect fraud and improve underwriting. Telecom operators want more efficient networks and better customer analytics. Insurers need faster claims processing. Hospitals, utilities, universities, mining companies and large industrial groups are all confronting problems for which young technology companies are developing increasingly sophisticated solutions.
Nor is the supply side empty. Africa’s AI startup base is expanding rapidly. TechCabal Insights recently mapped 207 AI startups across 17 African countries, almost twice the number in its comparable 2022 dataset. The more interesting question is therefore no longer simply whether Africa can produce AI companies. It is whether its markets are becoming easier for those companies to enter, compete in and scale.
The buyers are there. The needs are there. The technology increasingly is too. What often remains difficult is the exchange between them. Africa may not have a buyer shortage. It may have a transaction-cost problem.
The hidden cost of buying AI
The cost of an AI solution is not limited to the vendor’s price. A large institution considering a young supplier must find it, understand its technology, establish whether it works, assess cybersecurity and regulatory risks, determine how it will interact with existing systems, secure internal approvals, negotiate access to data, agree contractual responsibilities and decide who will be accountable if something goes wrong. A young company can therefore offer the cheaper product and still be the more expensive transaction.
Economist Ronald Coase made transaction costs central to understanding why seemingly beneficial exchanges do not always happen. Markets themselves are not costless. Searching, evaluating, negotiating and monitoring all consume resources. If the cost and risk of completing an exchange become too high relative to its expected benefit, a rational buyer may simply decide not to transact.
That helps explain an apparent paradox in emerging AI markets. A bank or telecom company can recognize that a startup has promising technology and still conclude that purchasing it is too cumbersome or risky.
The startup pays for the friction as well. Months of bespoke diligence, demonstrations, security reviews and custom integration consume scarce management time and capital. Worse, the next prospective customer may require much of the process to begin again.
What looks like a startup sales problem can therefore be a transaction-cost problem shared by buyer and seller. Some sophisticated corporations have recognized that this friction imposes costs on them, not merely on startups.
BMW’s Startup Garage is an instructive example. BMW describes it as its “venture client” unit. Instead of investing in startups, the company identifies technologies that may solve problems inside its business, provides a structured pathway for testing them and, when they perform, enables the companies to become suppliers and longer-term partners. BMW says it has assessed more than 6,000 startups, conducted more than 280 pilot projects, and that over half of its program alumni have secured paid follow-on projects.
The logic is commercial, not charitable. BMW wants access to innovations that can improve its products, processes and competitiveness. It has therefore reduced the organizational cost of dealing with unfamiliar young suppliers instead of asking every business unit to rediscover from scratch how to work with one.
The next accelerator may sit inside the buyer
That model deserves more attention in Africa. For years, African technology ecosystems have built accelerators, incubators, demo days, competitions and venture programs designed to make startups more ready for markets. Perhaps the next African AI accelerator needs to sit inside the buyer.
A bank could maintain a structured channel through which emerging technologies in fraud detection, underwriting or customer service are evaluated. A telecom operator could create a repeatable route for testing tools in network optimization or customer analytics. Utilities, hospital groups, universities, mining companies, insurers and large retailers could build similar mechanisms around the problems most important to them.
These should not be charity programs. They should not guarantee contracts, favor weak local products or oblige companies to purchase technology merely because it is African.
Quite the opposite. Their purpose should be to make good innovation easier to evaluate while keeping the standard for winning the business high.
Everett Rogers’ classic work on the diffusion of innovations helps explain why this matters. Innovations spread more rapidly when users can see their relative advantage, when they fit existing needs and systems, when they can be tried before full adoption, and when their results are visible to others. Simon & Schuster
A sophisticated first buyer can therefore do more than create revenue for one startup. By testing the product, integrating it into a demanding operating environment and demonstrating measurable results, that buyer creates information that did not previously exist. The second buyer no longer encounters an entirely unknown technology. There is a reference implementation, operational evidence and a customer whose diligence carries credibility.
The first buyer can, in effect, become a disseminator of trust. That matters because the first contract creates revenue, but a market begins to form when the experience of the first buyer makes the second transaction easier.
Buyer-side accelerators could help make that process repeatable. They can make value easier to measure, compatibility easier to test, quality easier to verify and successful results easier to observe. Standardized diligence, clearer contracting procedures and recognised testing processes can prevent every new commercial conversation from returning to zero.
The governing principle should be straightforward: Lower the barriers to the transaction, not the standards for the product.
This is also an investability issue. An AI company whose every sale requires months of bespoke evaluation and integration will struggle to scale even if its technology is excellent. When the cost of transacting falls, successful pilots are more likely to become contracts, reference customers can help generate subsequent customers, revenues become more repeatable and companies become easier to finance.
Africa has spent years strengthening the supply side of innovation. That work should continue. But functioning innovation ecosystems also require sophisticated buyers—and mechanisms through which those buyers can engage unfamiliar suppliers efficiently. African AI companies do not need protection from demanding customers. They need more opportunities to prove themselves to them.
From prototype to procurement, the next great African AI accelerator may be built not for the startup, but for the buyer.
*Roger B. Jantio is an AI investor and merchant banker focused on artificial intelligence, development finance, emerging markets and strategic capital. He is CEO of Sterling Merchant Finance Ltd., a Washington-based merchant bank active across Africa for more than three decades, and General Partner of its affiliated investment funds.