African banks are still pouring money into artificial intelligence. What has changed is that their boards now want to see the receipts.
That is the central finding of The State of AI in African Banking 2026, a report published on 21 July 2026 by the banking-software firm Backbase in partnership with African Banker magazine. Drawing on responses from 277 senior banking executives across 37 African countries, it describes a sector moving out of open-ended experimentation and into what it calls an “accountability phase”, where the question has narrowed to one line: ROI or no AI.
The report should be read with its authorship in mind. Backbase sells an AI banking platform, so a study concluding that African banks need to invest in AI infrastructure is not a disinterested one. Several of its findings, however, are specific enough to be useful regardless.
The measurement gap
The starkest number is about accounting rather than technology. Only 67.1% of the banks surveyed measure the return on their AI spending at all, even as budgets rise across the board. Among the third that have set up formal ROI measurement, 85.1% say the results meet or exceed what they had projected, which suggests the problem is less that AI underperforms than that most institutions cannot yet tell whether it does.
Pressures outside the banks are forcing the issue. The report points to foreign-exchange strain, rising dollar-denominated cloud costs and tightening data-localisation rules as the forces concentrating executive attention on hard returns.
Legacy systems as the ceiling
Ageing core banking systems emerge as the single biggest internal obstacle, named by 50.2% of respondents. The report estimates that, on average, 55.7 cents of every IT dollar an African bank spends goes on maintaining legacy systems rather than building anything new, a drag that also degrades the clean data AI needs to work well.
“African banks don’t have an AI problem, they have an architecture problem,” said Aymen Daoud, Regional Vice President for Africa at Backbase. It is a claim that neatly describes a market for the kind of modern core systems his company and rivals sell, but the underlying constraint is one African bankers themselves rank first.
Where the AI is actually going
Conversational AI, the chatbots and virtual assistants that front customer service, is the sector’s default entry point, cited as a use case by 49% of respondents. The report frames it as the shallow end: more advanced deployments in credit, risk and revenue tools separate what it labels “Innovators” from “Early Adopters”.
The use case respondents rate most impactful is fraud detection and transaction monitoring, followed by credit scoring. That second application matters beyond the balance sheet: alternative credit assessment for “thin-file” customers, those with little formal financial history, is one of the more credible routes to pulling more of the region’s unbanked population into the formal system, a thread that runs through recent African fintech moves such as Interswitch’s core-banking tie-up with Temenos.
Why it matters
Sentiment remains firmly positive: 86.9% of respondents are positive about AI’s role over the next two years and 83.2% expect to increase their investment. The report’s value is in the counterweight it puts against that optimism, an insistence that intent is outrunning the ability to measure whether the money works.
For a continent where the same institutions are simultaneously chasing financial inclusion, the discipline cuts both ways. Spend that cannot be measured is hard to defend to a board, but tools like alternative credit scoring only reach the underserved if banks keep funding them long enough to work. The report’s warning is that the banks best placed to do both will be the ones that fix their plumbing first, and can prove the difference afterwards. That same generational shift is visible elsewhere in African finance, from the under-35s driving earned-wage-access uptake to the steady digitisation of core banking.
