SHARE
By Dr Mark Nasila, Chief Data and Analytics Officer, FNB AI Strategy
The recently launched Evident Insights 2026 AI Index for Banks, which assessed 25 institutions across the Middle East and Africa, suggests that artificial intelligence in banking has reached the point where it can be usefully benchmarked and compared. The Index measures enterprise capability across talent depth, leadership commitment, transparency in reporting, and innovation pipelines.
South African banks performed strongly across the cohort, reflecting several years of investment, planning, and organisational change. This signals the sector’s maturation, but not its end state. And, while the debate around AI is often framed around its risks and benefits, the more important question is whether institutions can convert AI capability into medium-term enterprise value. In FNB’s view, five considerations determine whether AI creates lasting value, or remains a series of disconnected experiments.
The first is whether AI is treated as a technology strategy or a people strategy. In implementations that have delivered durable value, the important factor has been reskilling teams, building literacy across the organisation, and ensuring that the capacity AI unlocks is reinvested in higher-order work rather than absorbed into the productivity trap, where efficiency improves but strategic outcomes remain unchanged. Interestingly, subject-matter expertise becomes more important under these conditions: somebody has to understand the process well enough to accurately judge where AI brings benefits.
The second consideration is whether the institution has a clear strategic frame for the value AI is meant to create. Without it, AI conversations stay general and resources get committed without a clear goal. Effective programmes start by asking which processes, experiences or decisions are inefficient, and which of those can be improved with AI.
AI must then be evaluated as one possible instrument, not the only one. In some cases, it’s the right solution. In others, a simpler intervention does the job.
The third is the analysis, treatment, and disclosure of risk. Reputational damage and workforce disruption are persistent risks, while regulators and banks are increasingly focused on AI sovereignty. This includes who controls the models, where the data lives, what the training set contained, and how dependent an institution becomes on a single provider or platform.
An institution reliant on a single external platform is exposed to that provider’s decisions on access, pricing and terms, none of which it controls. This calls for discipline in knowing where dependencies sit and what it would cost to move. The fourth consideration is governance. Closing the gap between ambition and execution requires clear accountability across data teams, risk partners, legal teams, and the business owners who the processes AI is expected to improve.
A common failure mode is ambiguity. AI becomes everyone’s responsibility and therefore no one’s. A working governance model identifies opportunities, evaluates risks in detail, assigns ownership, and measures outcomes against a baseline. Without that, AI becomes a vague, generalised and unquantifiable capability.
The fifth is sequencing. There is a strong temptation, particularly for customer-facing banks, to lead with customer-facing AI. Chatbots are visible and generative interfaces demo well and they often attract early attention. But for most institutions, the higher-value short-term opportunities sit inside the organisation.
At FNB, AI is already delivering measurable impact at scale – from significantly improving throughput in legal review processes, to transforming fraud and anti-money laundering capabilities. These interventions protect both the bank and its customers from billions of rands in potential losses each year. These use cases are valuable and measurable, and they build the institutional capability that eventually makes customer-facing AI safer to deploy.
There is no shortcut to AI value at enterprise scale. Institutions that perform well in benchmarks such as the Evident Index do so because they have invested in the conditions that make AI effective: deep, multi-disciplinary talent bases; leadership that engages substantively with the technology; rigorous governance; and a clear structure for determining where AI delivers value aligned to strategy.
That investment has to be rolled out methodically over years.
Institutions that treat AI as a series of glamourous experiments will find their expenditure compounding without the payoff. The technology is rarely what separates the banks that realise enterprise value. The discipline to deploy it well is much harder to acquire, and much harder to copy.
INFO SUPPLIED.