Africa’s AI race will be won by businesses solving problems rather than chasing hype

Artificial intelligence may be the latest technology to arrive wrapped in breathless promises, but African businesses are already discovering that deploying it is less about flashy demonstrations and more about data, governance and finding problems worth solving.

That was the thread running through a panel discussion titled AI beyond the hype: Deploying AI across African industries, facilitated by broadcaster Bongani Bingwa at the Standard Bank Africa Unlocked conference in Cape Town last week.

Bingwa framed the discussion around a practical question: where is AI already working, and how can African companies use the continent’s mobile-first infrastructure, underserved markets and habit of necessity-driven invention to compete?

“The discussion has to confront critical risks around dependency on non-African platforms and models,” Bingwa said, pointing to the need for “African data, African models and African applications”.

From experiments to everyday business

Cathy Muraga, managing director of the Microsoft Africa Development Centre, said AI adoption was moving fastest in industries where there was a yawning gap between the number of people needing a service and those available to provide it.

Education, healthcare and agriculture were obvious examples because there are too few teachers, health workers and agricultural extension officers to serve growing populations.

Fintech companies also have an advantage because many are built as digital businesses from the start, with clearer systems for collecting and managing data.

But Muraga warned businesses against rushing onto a technology platform without considering what would happen if they wanted to move later.

“These platforms are great, but it is very important for organisations to figure out: how do we get in, but not get locked in?” she said.

Companies needed to retain access to their data, develop internal skills and understand the contracts governing the technology they were using. Otherwise, an apparently convenient AI partnership could become an expensive technological trap.

AI deployment also had to be driven from the top. Muraga said leaders could not simply instruct employees to adopt AI while remaining detached from the learning process themselves.

“Leadership must work,” she said. Executives need the humility to ask younger or more technically skilled employees to show them how the tools worked, and then model the behaviour they expect from the rest of the organisation.

The boring work comes first

Satish Babu, principal engineer at Standard Bank, pointed out that banks have used conventional AI for years in areas such as risk assessment and decision-making. Generative AI is now being tested for its ability to read documents, support employees and speed up the development of services.

The difficulty is moving that demonstration into a regulated production environment where it could be used safely and at scale.

“The tool stuff is easy,” Babu said. “The boring stuff is extremely important.”

That means getting the data into usable shape, making consistent technology choices and teaching employees how to use AI properly.

Read more

Integrating AI in your business is a double-edged sword

July 4, 2025

Standard Bank’s experience also showed why governance should not be treated as a bureaucratic brake on invention.

“Governance is the way that you need to go faster so you don’t crash,” Babu said.

An AI system could produce wrong answers, use confidential information improperly or make decisions that could not be explained to customers or regulators. Banks operating across several African jurisdictions also have to contend with countries moving at different speeds on AI and data regulation.

Babu said governance should therefore be built into the technology platform rather than added to every individual project afterwards. Clear ownership was equally important. Standard Bank had developed a use case that improved a particular decision by about 90%, but it stalled when business leaders passed responsibility for it between them.

Without somebody accountable for an AI system from the beginning, he said, it was unlikely to reach production or scale.

Build what outsiders cannot easily copy

Nkemdilim Uwaje-Begho, chief executive of Future Software, argued that Africa’s strongest AI opportunities lie in solving problems rooted in the continent’s languages, markets and institutional knowledge.

She cited a company developing a voice agent that could communicate in Nigerian and other African languages, which is in use in telehealth and customer service.

“Building things that are hard to replicate, building things where no one else has that data, are what we need to focus on,” Uwaje-Begho said.

This was particularly valuable where African languages were poorly documented and therefore largely absent from global AI training data.

The same principle applies inside companies. Businesses should identify information, expertise or operating knowledge that competitors did not possess and organise it into reliable, verified datasets.

Buying a generic AI tool would not, on its own, produce a competitive advantage.

Uwaje-Begho said companies could use AI defensively to make existing operations more efficient, extend their business models by offering additional services, or rethink their businesses entirely. The last option demanded changes to hiring, management and organisational design rather than another project handed to the IT department.

“Giving everyone access and saying, ‘just play around with it’, may not lead to the best results because you just burn a lot of money,” she warned.

Necessity is driving adoption

Yacob Berhane, chief operating officer and head of growth at Quill and chief executive of Pariti, said start-ups often have more freedom to experiment than large, regulated businesses.

AI is particularly useful in high-volume, low-margin industries because it allows companies to serve more customers without increasing costs at the same rate.

He described using an open-source model to help assess workers across thousands of placements and dozens of countries. Instead of relying solely on CVs, the model used whether a candidate had passed probation as a signal that the person could do the job.

Berhane said companies were still learning how to combine different models and AI agents, manage costs and check the work produced by automated systems.

AI could increase the amount of work an employee completed, but it also increased the burden of checking that work.

“The beauty of AI is that it increases your surface area for work, but it also increases your cognitive task for reviewing,” Berhane said.

His advice to executives was to experiment personally, start with repetitive work and find technically curious people who could help others move past the intimidation factor.

Capital still determines who gets to build

AI does not exist outside Africa’s broader economic constraints. It requires electricity, computing infrastructure, skilled workers and patient capital.

Leslie Maasdorp, chief executive of British International Investment, noted that development finance institutions are increasingly seeking to use their funding to draw African pension funds, insurers and sovereign wealth funds into productive investments.

“We are now seeing ourselves as principally being de-risking machines, as being catalytic,” he told delegates at the conference.

Rather than measuring success only by the money they invested directly, development financiers would increasingly measure how much domestic capital their investments unlocked.

“This is Africa building on its own financial strength,” Maasdorp said.

That capital will be needed if African businesses are to own more of the infrastructure, intellectual property and data underpinning the AI economy.

Mohammed Dewji, chief executive of Tanzanian conglomerate MeTL Group, offered a wider warning about Africa’s habit of exporting raw materials and importing finished products.

“Every time we export raw materials without processing them, we also export jobs, skills, technology and economic value,” he said.

The same logic applies to data. Africa risks supplying the raw material used to train global systems while importing the finished technology at prices and on terms set elsewhere.

Dewji’s prescription was to invest in people, skills, technology and knowledge transfer.

“Africa should not simply supply raw materials that power the world, but rather process, manufacture and capture more of the value they create,” he said.

For AI, the race is therefore not simply about adoption. It is about whether African companies become permanent customers of other countries’ technology or build products, models and businesses grounded in their own markets. DM