According to a 2026 report surveying 277 banking executives throughout 37 African nations, 83.2% of establishments plan to extend their AI spending over the subsequent 12 months. Yet almost a 3rd don’t formally observe the return on funding. More strikingly, amongst the lenders that don’t measure returns in any respect, 82% nonetheless plan to increase their budgets anyway. Between dollar-denominated cloud internet hosting payments, legacy core banking integration bottlenecks, and an acute regional deficit of machine studying engineers, African lenders are doubling down on an costly know-how whose final fiscal contribution continues to be an article of religion. In this version of Delve into AI, EMMANUEL NWOSU examines the uncomfortable disconnect between operational pace and audited balance-sheet yield inside Africa’s multi-million-dollar banking automation race.
Loan documentation has traditionally been an administrative bottleneck, even inside well-capitalised monetary establishments. At Standard Bank, a vetting course of that beforehand took roughly an hour now wraps up in about 60 seconds, powered by synthetic intelligence.
The metric, disclosed in the South African lender’s Capital Markets Day submitting in March 2026, applies to its industrial asset finance division. It serves as a flagship case research for the way Africa’s largest financial institution by belongings is operationalising machine intelligence.
Standard Bank reported that its conversational bot handles 65% of all incoming digital buyer queries, whereas over 39,000 workers—roughly 72% of its complete workforce—were regular enterprise users of generative AI by August 2026.
These workflows showcase the instant operational promise of automation: administrative backlogs evaporate into single-minute duties, buyer friction drops, and financial institution workers are freed to give attention to high-touch advisory roles requiring human judgement.
Yet the internet monetary dividend stays murky. Deploying AI throughout regulated monetary environments requires rigorous information governance protocols, costly software program licencing, advanced programs integration, and widespread workforce reskilling. For these outlays to be commercially defensible, banks should show that bottom-line yield outweighs implementation overhead.
Industry information reveals an uncomfortable divergence between capital expenditure and efficiency monitoring. According to The State of AI in African Banking—a 2026 joint report by banking software program supplier Backbase and African Banker, surveying 277 executives throughout 37 African nations—83.2% of establishments plan to extend their AI allocations over the subsequent 12 months. Yet solely 67.1% formally observe the return on funding (ROI) of these deployments.
More strikingly, amongst the third of respondents whose banks don’t observe AI returns in any respect, 82% nonetheless plan to increase their budgets. African banking leaders are aggressively capitalising a know-how whose industrial contribution they’ve but to quantify.
This absence of measurement doesn’t signify failure. Algorithmic programs can compress turnaround occasions, sharpen fraud monitoring, and elevate buyer expertise with out instantly shaving prices from the revenue assertion.
But the distinction between operational effectivity and monetary return turns into tougher to disregard as AI strikes past remoted experiments and into programs that banks use every single day.
However, as AI transitions from experimental sandbox pilots into core banking structure, the distinction between unmonetised operational pace and audited fiscal return turns into unattainable to disregard.
Faster work, unsure monetary returns
Standard Bank has been constructing its AI capabilities on know-how investments remodeled a number of years. In its March investor presentation, the financial institution stated conversational AI was dealing with 65% of digital queries on its private banking app, decreasing the workload related to customer support.
It additionally reported that engineering productiveness had improved by greater than 20% in the earlier yr, whereas AI was starting to speed up work in lending. Its industrial asset finance instance, by which mortgage documentation processing fell from roughly an hour to a minute, underscores the extent of the enchancment attainable in some administrative processes.
The financial institution has additionally expanded the know-how throughout its workforce. In August, it reported greater than 39,000 energetic customers of generative AI, supported by authorized enterprise instruments.
Consider the mortgage documentation course of. Reducing processing time from an hour to a minute frees up the workers answerable for that work. The financial institution might then be capable of course of extra functions with the similar workforce, scale back additional time, or redirect workers to different duties.
Those advantages turn into measurable monetary financial savings solely once they end in bills prevented or extra revenue that may moderately be attributed to the funding.
Standard Bank stated it plans to redirect the time workers save utilizing AI in direction of buyer relationships and work requiring extra advanced selections. This means that the financial institution is wanting past slicing working prices to extracting extra worth from its current workforce. On its personal, it doesn’t specify income era or price financial savings.
Sticking with Standard Bank’s lending enterprise, AI may assist workers course of extra mortgage functions with out a corresponding enhance in workers prices. If used to evaluate debtors’ creditworthiness, it may additionally assist the financial institution make higher lending selections and scale back the danger of defaults. These are two completely different advantages, and Standard Bank has not established whether or not its AI investments have delivered both financially.
The financial institution’s non-performing mortgage ratio stood at 5.6% at the finish of June, down from 6.2% the earlier yr. Credit impairment prices, which replicate anticipated losses from loans, additionally fell by 12% to R7.13 billion ($435.5 million) in the first half of 2026. The lender has not attributed the enchancment to AI utilization.
At Nedbank, one other tier-1 South African lender, the enhancements are additionally measurable.
According to its 2025 integrated report, the financial institution stated it carried out 30 superior AI use circumstances throughout the yr. Its broader programme, combining AI, generative AI, analytics, machine studying, and robotic course of automation, helped save roughly 312,000 employee-hours throughout media and content material creation, doc translation, and private productiveness. Global fee processing occasions additionally fell by about 30%.
Nedbank has utilized predictive fashions to establish prospects extra prone to buy funding merchandise, alongside analytics that assist it goal potential bank card and mortgage prospects.
These makes use of illustrate how banks count on to earn returns from AI in two broad methods: spending much less to supply current companies and incomes extra by serving prospects extra successfully.
The price of creating AI work
AI doesn’t function independently of a financial institution’s current know-how. Models want entry to dependable information, safe computing programs, and the functions via which workers and prospects use them. Banks additionally want processes for testing outputs, controlling entry to delicate data, and monitoring selections made with AI.
Those necessities could make the expense of deploying AI significantly bigger than the price of a software program subscription.
At its 2026 Capital Markets Day, Standard Bank reported complete know-how spending of R23.5 billion ($1.44 billion) in 2025, equal to 27.7% of its working bills. It stated 88% of its legacy servers had been decommissioned and that 71% of computing workloads eligible for migration had moved to the cloud at the time of the presentation.
The financial institution subsequently reported that cloud migration had reached 78% of eligible computing workloads by the finish of June 2026.
Standard Bank stated its know-how investments helped enhance effectivity over a number of years, contributing to a decline in its group cost-to-income ratio from 59.1% to 50.2%. This signifies that for each R100 ($6.09) the financial institution earned in working revenue, it spent about R50 ($3.05) on working bills, in contrast with roughly R59 ($3.6) beforehand.
Standard Bank’s know-how spending has continued to rise alongside its increasing AI operations. The financial institution spent R23.51 billion ($1.43 billion) on know-how in 2025, up 4.9% from R22.4 billion ($1.36 billion) in 2024. In the first half of 2026, its software program, cloud, and technology-related bills rose 6% to R7.16 billion ($436.5 million). The financial institution attributed the enhance partly to increasing its AI capabilities, alongside investments in specialised know-how experience, core banking programs and cybersecurity.
Yet, its total know-how spending grew by solely 2% to R11.83 billion ($721 million) throughout the six months, as decrease amortisation and different know-how prices offset a part of the enhance. Standard Bank has been following what it calls a save-to-invest technique, utilizing efficiencies from earlier know-how investments to fund newer capabilities, together with AI.
The financial institution’s funding in modernising its know-how programs comes amid the backdrop of its AI-readiness ambition; it illustrates the work required to introduce new capabilities into a longtime banking enterprise.
To show this level, African banks face AI integration challenges. In the Backbase and African Banker survey, 50.2% of respondents recognized integrating AI with current know-how as their largest inside impediment to adoption.
Banks might have to attach buyer information, transaction histories, and different data saved throughout programs developed at completely different occasions. A lender introducing AI into credit score evaluation, for instance, wants dependable entry to the monetary data used to judge debtors. Bringing these information collectively can require upgrades to current programs, extra computing infrastructure, and specialised workers to take care of them.
These necessities add to the preliminary and ongoing price of utilizing AI. Where banks buy cloud companies or software program in foreign currency, exchange-rate actions may enhance their bills in native foreign money.
The monetary case for AI will depend on greater than how a lot time a system saves. Banks should weigh these features towards the full price of introducing and working the know-how, together with the infrastructure wanted to assist it.
Why the returns stay tough to measure
The Backbase and African Banker analysis means that some African banks are getting the monetary outcomes they anticipated from AI. Among respondents whose establishments formally measure returns, 85.1% stated their investments had met or exceeded their projections.
The banks, nevertheless, have been measuring efficiency towards their very own targets. The survey doesn’t set up how a lot they invested, the monetary worth generated, or whether or not returns have been calculated persistently throughout establishments.
Consider AI-powered lending. If a financial institution approves extra functions after introducing AI, the extra loans may enhance its curiosity revenue. But the monetary profit would rely upon funding prices, the efficiency of these loans, and whether or not the know-how improved lending selections. Similarly, an AI fraud detection system’s worth lies in the losses it prevents, not merely the variety of suspicious transactions it flags.
The survey additionally discovered that 71.7% of respondents working with third-party AI suppliers measure returns, in contrast with 31% of these creating AI totally in-house, underscoring how banks differ significantly of their potential to deploy proprietary AI—one other price centre.
In June 2026, Evident’s inaugural AI Index for Banks in the Middle East and Africa assessed 25 main banks on AI expertise, innovation, management, and transparency. Standard Bank ranked second and Nedbank fourth, making them the solely two South African banks amongst the 4 regional leaders. Other South African lenders, FirstRand, Capitec, and Absa, rounded off the high ten checklist.
Evident attributed their comparatively sturdy efficiency partly to bigger swimming pools of AI expertise, proximity to analysis universities, and engagement with rising central financial institution regulation.
The focus means that South African banks have constructed stronger AI capabilities than a lot of their regional friends. Across the 25 establishments assessed, Evident discovered that specialist AI growth expertise, relative to workforce dimension, was solely half the degree recorded amongst its benchmark of main international banks.
This scarcity has monetary implications. Banks with out adequate inside experience should both put money into constructing their very own AI capabilities or depend on exterior suppliers, whereas nonetheless paying to combine and keep the know-how.
Some banks outdoors Africa have begun offering a clearer account of the monetary worth they derive from AI. Singapore’s DBS Bank reported roughly S$1 billion ($783 million) in financial worth from information analytics and AI in 2025, up from S$750 million ($587 million) in 2024. By the finish of 2025, it had deployed greater than 2,000 fashions throughout over 430 use circumstances.
DBS has gone past reporting what number of duties AI completes or how a lot time workers save. It has positioned a financial worth on the know-how’s contribution to its enterprise, though it’s unclear whether or not the Singaporean lender is now worthwhile on its AI utilization.
The comparability brings the monetary query dealing with African banks into clearer focus. Standard Bank and Nedbank have demonstrated enhancements in productiveness, whereas many establishments surveyed by Backbase report assembly their AI funding targets. But they supply little indication of how a lot particular person AI programmes price or what they return financially.
Productivity is a key use case for AI in banking, and throughout different non-financial organisations. But the alternative right here is establishing what that productiveness is value.
True scale calls for transferring past surface-level integrations to sturdy execution. We’ve filtered the noise out of Moonshot 2026, optimising the convention strictly for high-calibre connections between startup founders, international monetary operators, enterprise leaders and people rewiring Africa’s technical frameworks. Get 20% off Early Bird tickets for a limited time.
