In Africa and emerging markets, the story of digital banking over the past few years has been one of relentless product launches. New mobile apps. New BNPL features. New wallet integrations. Capital has flowed into building many fintech platforms that have posted promising numbers in the African payments sector.
Across 25 African countries, 64 billion instant payment transactions worth nearly $2 trillion were processed in 2024, a 26 per cent average annual rise since 2020, according to the AfricaNenda Foundation. The same report, compiled in consortium with the World Bank and the United Nations Economic Commission for Africa, notes that Nigeria is the only country on the continent with a payment system rated ‘Mature’ on the global inclusivity index. Volume, reach, and adoption have all moved quickly.
Unit economics and, by extension, profitability have, however, not been as pronounced. And this is an important dimension of digital transformation, as without achieving profitability, the case for building lasting businesses is weakened. As funding cycles tighten, the institutions that will survive and lead the next decade are those that build infrastructure capable of bearingthat build infrastructure capable of bearing the weight of scale. Thus, building sustainable growth in digital finance goes beyond launching a new product. It begins with having a modern core.
The Gap Between Growth and Scale
There is a whole world of difference between growing and scaling. Put simply, growth means more customers, more transactions, more products. Scaling means doing all of that without a proportional increase in cost, risk, or operational fragility. Most digital banks have figured out how to grow. Far fewer have figured out scale.
One reason for this is a structural constraint. Many institutions built their digital layer on top of legacy cores that were never designed for real-time analytics, AI integration, or multi-channel orchestration. The result is a patchwork of middleware, manual workarounds, and fragmented data; the kind of architecture that works at 10,000 customers but begins to show signs of fracture at 100,000.
As Abdul Sulaiman, Regional Head – Africa, Oradian, puts it, “The bottleneck is almost never the front-end product. It is the core infrastructure that either enables or limits everything built on top of it.” This is what makes the current moment so consequential for fintech founders and regulators alike.
Regulators are aware of these issues and are now institutionalising frameworks to address them.
Nonetheless, the first wave of digital banking proved that mobile-first products could reach underserved populations. The second wave — the one we are in now — will determine which institutions can serve those populations profitably, safely, and at genuine scale.
Why AI and Data Only Work When the Foundation Does
The global financial sector is not short on AI ambition. Financial institutions invested more than $35 billion in AI in a single year, with projections pointing to $97 billion by 2027, according to the World Economic Forum.
Yet, on a sombre note, most AI projects fail. Not because the algorithms are wrong, but because the data underneath them is incomplete, inconsistent, or inaccessible. You cannot build a fraud-detection model from fragmented transaction records. It is difficult to run credit scoring for thin-file borrowers without a clean, continuous view of their financial behaviour. Personalising a customer’s banking experience becomes a pipe dream if your core system cannot tell you what that customer actually did last Tuesday. A recent white paper on digital banking by Oradian reports that 95% of AI projects fail to deliver on their promises, contributing to 40% of failed business initiatives, due to poor data foundations.
“Banks keep asking which AI use case to start with. The real question is whether their data infrastructure can support any AI use case at all. That is the conversation we need to have first.”
— Rodney Trivangalo, VP of Marketing, Oradian
There are other considerations to keep in mind. A 2023 global bank survey by EY found that the most widely cited benefits of generative AI in banking were increased productivity through the automation of sales-related activities, enhanced existing technological capabilities, and accelerated broader innovation.
Yet the same report points to a gap between ambition and readiness. Despite the promise of AI, 37% of bankers expressed limited confidence in their internal capabilities, including infrastructure, controls and talent required to implement GenAI use cases effectively. Concerns around data privacy, security, accuracy and reliability also remain prominent, underscoring the operational and governance challenges that continue to shape adoption.
Regardless, many forward-looking institutions are getting it right by treating data consolidation as a strategic priority rather than an IT task. They established secure, real-time replicas of their core banking data. This enabled environments where analytics and AI model development could happen without touching the live production system. They also built pipelines before models. And as a result, when they did deploy AI for collections prioritisation, customer onboarding, or fraud detection, it worked.
What Regulators Are Watching
For regulators across Africa and emerging markets, the rise of AI-powered banking brings its own nuances. The promise of faster credit decisions, broader financial inclusion, and more efficient compliance is welcome. However, the risk of algorithmic bias, opaque decision-making, data misuse, and institutions deploying tools they do not fully understand also exists.
The regulatory posture emerging from Nigeria’s CBN to Kenya’s CBK reflects sensible pragmatism: innovation is welcome, but it must be explainable, auditable, and fair. In March 2026, the CBN issued baseline standards for automated AML solutions that explicitly incorporate AI and machine learning into Nigerian financial regulation. Banks have 18 months to comply. An AI system that denies a loan must be able to explain its decision. A fraud detection model must not disproportionately flag customers from particular regions or demographics. A chatbot handling sensitive financial queries must have a fallback to a human when it matters.
These demands are not unreasonable. However, they require a level of institutional infrastructure that many fintechs and digital banks have not yet built. Explainability requires documentation. Auditability requires logs. Fairness monitoring requires data. All roads lead back to the same place: the core.
Institutions that want to work with regulators need to demonstrate that their AI is well-governed, that their data is clean, and that they have the organisational maturity to own the models they deploy, rather than just buy them from vendors. That means having named owners for AI projects, involving risk and compliance teams from day one, and maintaining audit trails that align with global best practices.
The AI Opportunity Is Real
The African Development Bank projects $1 trillion in additional GDP by 2035 if AI is adopted with the right enablers in place. In Ethiopia, businesses like Kifiya have already used AI to disburse $100m in loans to small businesses. The technology works.
What does not work is deploying AI on a broken data layer. Oradian’s whitepaper is precise on this point: success with AI in digital banking requires reliable access to production-grade data, a way to work with that data off the live core system, and a consistent pipeline from raw data to AI-driven decisions. Each condition is non-negotiable. Fail on any one of them and the model, however sophisticated, will produce unreliable outputs.
What the next chapter requires is the institutional maturity to build profitably, to govern AI responsibly, and to treat core infrastructure as the enabling condition for everything else. Those who do will not merely survive the current moment. They will define what comes after it.
For the full framework on AI-ready banking infrastructure, including a readiness checklist and use-case prioritisation matrix, read Oradian’s guide: The Digital-First Bank’s Guide to AI in 2026.
