Companies solve problems instead of chasing hype

Artificial intelligence may be a new technology arriving with breathless anticipation, but businesses in Africa are already realizing that deploying artificial intelligence is more about data, governance, and finding problems worth solving than flashy demonstrations.

This was the thread running during a panel discussion titled “AI Beyond the Hype: Bringing AI across Africa’s Industries” moderated by broadcaster Bongani Binwa at the Standard Bank Africa Unlocked Conference in Cape Town last week.

Bingwa framed the discussion around practical questions about where AI is already at work and how African companies can compete by taking advantage of the continent’s mobile-first infrastructure, underserved markets, and habit of inventing on demand.

“The debate needs to confront the significant risks surrounding reliance on non-African platforms and models,” Bingwa said, noting the need for “African data, African models and African applications.”

From experiments to daily work

Kathy Muraga, managing director of Microsoft Africa Development Center, said AI adoption is happening fastest in industries where there is a wide gap between the number of people who need services and the number of people who can serve them.

Education, health care, and agriculture are obvious examples, as there are too few teachers, health workers, and agricultural extension workers to cope with the growing population.

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

But Muraga cautioned companies against jumping into technology platforms without thinking about what would happen if they wanted to migrate later.

“These platforms are great, but it’s really important for organizations to understand how to get in and avoid getting locked in,” she said.

Businesses needed to maintain access to their data, develop internal skills, and understand the contracts governing the technology they were using. Otherwise, seemingly useful AI partnerships can turn into costly technical traps.

AI adoption also needed to be led from the top. Muraga said leaders cannot simply tell their employees to adopt AI while remaining disconnected from the learning process themselves.

“Leadership has to work,” she said. Management should humbly ask younger and more technically proficient employees to teach them how the tools work and model the expected behavior for the rest of the organization.

boring work comes first

Satish Babu, principal engineer at Standard Bank, pointed out that banks have been using traditional AI in areas such as risk assessment and decision-making for many years. Generative AI is currently being tested for its ability to read documents, support employees, and speed up service development.

The challenge is moving that demonstration into a regulated operational environment where it can be used safely and at scale.

“The tools are easy to work with,” Babb says. “Boring is very important.”

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

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

“Governance is a necessary tool for companies to move quickly and not crash,” says Babu.

AI systems can generate incorrect answers, inappropriately use sensitive information, or make decisions that cannot be explained to customers or regulators. Banks operating across multiple African jurisdictions also have to contend with countries moving at different speeds on AI and data regulation.

Governance therefore needs to be built into the technology platform, rather than being added as an afterthought to individual projects, Babu said. Clear ownership was equally important. Standard Bank had developed a use case that improved specific decision-making by about 90%, but got stuck as the responsibility was handed down among business leaders.

If no one is responsible for an AI system from the beginning, it’s unlikely to reach production or scale, he says.

Build something that outsiders can’t easily copy

Nkemdilim Uwaje-Begho, CEO 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 that has developed voice agents that can communicate in Nigerian and other African languages, and is used in telemedicine and customer service.

“Building things that are hard to replicate, building in places where no one has that data is what we need to focus on,” Uwajebego said.

This was particularly valuable where African languages ​​are poorly documented and therefore poorly represented in global AI training data.

The same principle applies within a company. Companies need to identify information, expertise, and operational knowledge that their competitors don’t have and organize it into reliable, verified data sets.

Purchasing a general-purpose AI tool does not, by itself, create a competitive advantage.

Uwajebego said companies could use AI defensively to make existing operations more efficient, extend business models by offering additional services, or completely overhaul their businesses. The last option required changes in hiring, management, and organizational design rather than leaving IT with another project.

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

Necessity drives adoption

Jacob Berhane, chief operating officer and head of growth at Quill and CEO of Parity, said startups often have more freedom to experiment than large, regulated companies.

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

He explained that it uses open source models to help evaluate workers in thousands of placement locations and dozens of countries. Rather than relying solely on resumes, this model used whether a candidate passed a probationary period as a signal of whether they were capable of the job.

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

AI may increase the amount of work employees complete, but it also increases the burden of checking that work.

“The benefit of AI is not only more surface area to work with, but also more cognitive tasks for review,” Verhane says.

His advice to executives was to find technically curious people who can experiment individually, start with iterative tasks, and help others overcome the intimidation factor.

Capital still determines who builds

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

Lesley Maasdorp, chief executive of British International Investment, noted that development finance institutions are increasingly seeking to use their capital to attract African pension funds, insurance companies and sovereign wealth funds into productive investments.

“We now think of ourselves primarily as risk aversion machines and catalysts,” he told conference participants.

Development investors will no longer measure success solely by the amount of money they invest directly, but by how much domestic capital their investments free up.

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

African companies will need that capital to own more of the infrastructure, intellectual property and data that will support the AI ​​economy.

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

“Every time we export raw materials unprocessed, we also export jobs, skills, technology and economic value,” he said.

The same logic applies to data. Africa risks supplying the raw materials used to train the world system while importing finished technology at prices and terms set elsewhere.

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

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

So for AI, competition is not just about competing for adoption. It will depend on whether African companies become permanent customers of other countries’ technologies or build products, models and businesses that are rooted in their own markets. DM