African banks are ramping up investment in artificial intelligence even as most of their senior leaders fail to measure whether the technology is delivering a financial return, according to an industry survey of 277 executives across 37 countries.
The findings, published on Monday in a report by African Banker in partnership with banking software provider Backbase, expose a stark accountability gap at the top of the continent’s lenders. While 82 per cent of finance and profit-and-loss leaders said they formally tracked the return on AI investments, the proportion fell to just 50 per cent among C-suite executives. Risk and compliance officers were only marginally ahead, at 48 per cent.
This disconnect is compounded by a broader pattern of spending without measurement. Across all respondents, a third of institutions said they had no formal process for assessing AI’s return on investment. Yet 82 per cent of those non-measurers said they intended to increase AI expenditure over the next 12 months.
The survey captures a sector that has moved rapidly from experimentation to deployment, buoyed by optimism: 87 per cent of respondents said they were positive or very positive about AI’s role over the next two years, and 83 per cent were likely or very likely to increase investment. But the numbers also suggest that much of this expansion rests on conviction rather than evidence.
“The return is real, but the discipline to measure it is not consistently present,” the report concludes. “For Africa’s banking sector, measurement is the differentiator.”
The data points to a divide between institutions that have embedded accountability into their AI programmes and those that have not. Among banks that did track AI returns, 85.1 per cent reported that results met or exceeded their original projections. The most frequently cited high-impact use case was fraud detection and transaction monitoring, followed by credit scoring for thin-file customers using mobile money data. Conversational AI and chatbots, though the most widely deployed application, ranked only third in perceived impact.
The governance gap at executive level has drawn sharp commentary. “The C-suite is the ultimate investment decision-maker, yet it is half as likely to track AI ROI as the finance function,” the report states. “This represents one of the sector’s most critical blind spots.”
Structural factors help explain why accountability remains patchy. Legacy integration was identified as the single greatest barrier to scaling AI, cited by just over half of respondents both as an internal obstacle and as a sector-wide impediment. But many executives appeared to underestimate the severity of the problem within their own institutions. While 48.5 per cent of respondents described their legacy systems as highly or fully capable of supporting AI, legacy integration simultaneously topped the list of barriers — creating what the report calls a “dangerous blind spot”.
The tension is especially acute for pan-African banks. Despite their operational breadth and sophistication, only 54.2 per cent of pan-African respondents said they measured AI returns, compared with 90.9 per cent of regional banks. The report attributes the shortfall to multi-jurisdictional fragmentation and inconsistent internal reporting, warning that scale can become “a liability for accountability”.
External partnerships appear to enforce a discipline that many in-house efforts lack. Institutions that used third-party AI vendors reported measuring ROI at a rate of 71.7 per cent, against just 31 per cent among those building entirely in-house.
Currency pressures add a further layer of urgency to the measurement question. The Nigerian naira lost more than 40 per cent of its value against the dollar between 2023 and 2025, while the Kenyan shilling hit multi-decade lows. Dollar-denominated cloud and software costs have risen in tandem, squeezing margins and forcing boards to scrutinise technology budgets more closely. The report notes that the sector’s recent improvement in cost-to-income ratios has come largely from revenue growth, not efficiency gains — leaving banks exposed if they cannot demonstrate that AI is bending the cost curve.
Data localisation rules are compounding the complexity. Central banks in Nigeria, South Africa, Kenya and Egypt have each imposed distinct mandates on how customer financial data is stored and processed, constraining the cross-border data flows on which AI models often depend. The survey identifies data privacy as the second-most-cited internal obstacle, close behind legacy integration.
The findings arrive as African lenders enjoy strong headline profitability — average return on equity reached 19 per cent in 2024 and 17 per cent in 2025 — but face a longer-term efficiency problem. McKinsey has noted that African banks’ cost-to-asset ratio runs at roughly double the global average. AI is widely seen as the most credible lever for closing that gap, but only if its deployment is tethered to measurable outcomes.
“African banking has rarely looked stronger on paper,” the report observes. “But the technology banks are being asked to adopt is the one that punishes inefficient architecture the hardest.”
The survey categorised banks into three tiers: early adopters with departmental-level deployment (45.5 per cent of respondents), an early majority piloting AI across teams (28 per cent), and innovators pursuing organisation-wide transformation (26.5 per cent). Among innovators, 85.5 per cent measured AI returns, reinforcing the report’s central argument that maturity and measurement reinforce each other.
Aymen Daoud, regional vice-president for Africa at Backbase, argues that the deeper constraint is architectural rather than algorithmic. “Africa’s banks don’t have an AI problem. They have an architecture problem,” he wrote in a commentary accompanying the survey. “Autonomous agents are a third actor, and they do not tolerate the isolation that humans do.” His point is that fragmented back-office systems will prevent lenders from realising the gains promised by more advanced forms of AI.
The report stops short of predicting a pullback in spending. But it makes clear that the current trajectory — expanding budgets without tracking returns — is unsustainable. For banks that have built the discipline to measure what they deploy, AI is already delivering. For the rest, measurement itself has become the bottleneck.