Buy v Build: How South Africa’s financial institutions can accelerate their AI advantage
For years, the transformation debate inside South African financial institutions (FIs) has centred on a familiar build v buy tension.
But with AI accelerating timelines, the urgency to unpack the modernisation challenge and make adoption decisions has intensified.
While 67% of South African FIs are still in the early stages of their AI journey, cloud-native fintechs aren’t waiting. By building the right foundation, they’ve set themselves up to mature quickly and are demonstrating the advantages of modern infrastructure through rapid innovation and market capture. As a result, the technology gap is widening.
It’s a picture South African banking leaders acknowledge. At our recent roundtable of senior executives from across the sector, 52% described their organisation as still testing or experimenting with AI, and only 16% said they are scaling it widely.
Approximately 64% of technology budgets at traditional institutions are consumed by maintaining legacy systems, leaving little bandwidth for the kind of innovation needed to compete.
And for South Africa’s mid-tier and regional banks, without the technology budgets of larger tier-1 banks the challenge is particularly acute.
Asked for the number one reason their bank hasn’t deployed AI at scale, our roundtable participants put data quality and legacy systems firmly at the top of the poll (39%), ahead of every other factor.
What banks underestimate about building AI in-house
In order to innovate at the pace of the industry today, effective banking AI is not simply a matter of hiring data scientists and plugging in a large language model.
It requires a unified, high-quality data foundation spanning all lines of business; domain-specific model training calibrated to banking workflows and credit decisions; rigorous explainability frameworks to satisfy South Africa’s dual regulatory regime under the Prudential Authority and the FSCA; and ongoing governance infrastructure to ensure models remain accurate, compliant, and auditable.
Each of these represents a significant undertaking on its own. Combine them, and you have an expensive, long-term program that is difficult to manage and maintain.
According to BCG, 74% of companies have yet to unlock tangible value from AI, with only 26% having developed the capabilities needed to move beyond proofs of concept.
In South Africa, where 84% of corporations already struggle to source skilled AI and data science talent, those odds are even more challenging.
Little wonder that nearly one in five of our roundtable leaders (18%) named a lack of AI talent and expertise as their single biggest barrier to scale, or that 59% are prioritising reskilling and AI training programmes to prepare their workforce for the AI era.
There’s also the question of data. Forty-seven per cent of participants at our recent roundtable said their data is siloed across systems, and when asked to name the biggest challenge facing their commercial lending teams, data quality topped the list.
The foundation of effective banking AI isn’t just any data; it’s structured, normalised, operational intelligence drawn from real banking workflows at scale. A bank building in-house trains its models on its data alone.
A platform like nCino built across a network of 1,500+ financial institutions is trained on patterns from thousands of real lending cycles, credit decisions, and portfolio outcomes.
The South African context makes the case even stronger
South Africa’s banking sector faces a set of challenges that compound the build vs buy decision in ways that don’t apply to other markets across EMEA.
For example, South African SMEs contribute around 34% of GDP; yet receive just 13% of total bank credit. The IFC estimates the financing gap at approximately $30 billion. Closing that gap is not simply a commercial opportunity; it is a lever for economic inclusion at national scale.
The problem here is infrastructure. SME lending at scale requires connected, automated origination and underwriting that integrates alternative data sources, reduces cost-to-serve, and makes smaller loan sizes commercially viable. That isn’t a system most banks can build quickly. And for every month spent building it, the opportunity cost mounts.
There’s also the regulatory dimension. South Africa’s exit from the FATF grey list in 2025 was a milestone, but it has also raised the compliance bar. The twin-peaks regulatory framework, evolving AML obligations, Basel III reforms, and the South African Payment System Directive 2025 together mean that a specialist platform, with dedicated regulatory update cycles, is far better placed to handle than a stretched in-house team operating on legacy technology.
That anxiety shapes AI decisions directly too, with 14% of commercial leaders attending our roundtable citing fear of regulatory scrutiny as the primary reason they haven’t yet deployed AI at scale.
Fraud is a further pressure point. With 81% of South African banking leaders estimating annual fraud losses exceeding $5 million per institution, and 75% reporting rising attack volumes, the case for AI-powered detection embedded within operational workflows is compelling. With the right foundation, an FI can automate KYC and KYB verification, monitor portfolios for covenant breaches, and generate audit-ready reporting through AI agents.
The smartest South African FIs are buying the foundation and building the advantage
What the most forward-thinking South African FIs are recognising is that the strategic choice is not between building and buying.
Today’s dynamic market demands a more sophisticated approach, one that harnesses the strengths of both strategies while mitigating their inherent limitations.
Buy proven, compliant foundations for core workflows, lending operations, and embedded AI. Then direct internal development energy toward the capabilities that genuinely drive market-specific differentiation in terms of client relationships, product innovation, and local knowledge.
This is precisely where South African leaders want AI to deliver. At the roundtable, 54% named improving operational efficiency as their organisation’s top AI priority, with enhancing customer experiences (32%) close behind.
The metrics support the advantage that can be realised – 30% productivity improvements within 18-24 months, credit review cycles compressed by 70%, more time to focus on judgment and client service rather than administration, commercial loans delivered in days rather than weeks.
South Africa’s banking sector is at an inflection point. Record earnings provide headroom to invest, and improved standing post-FATF removal creates conditions for greater financial integration and partnership.
Meanwhile, the expectations of an increasingly diverse and demanding customer base are rising fast. And when roundtable participants described the future of commercial lending at their own institutions, their answers read like a specification for a modern platform: agentic AI with humans in the loop, end-to-end integration, relationship managers empowered by data insights, and time-to-money measured in hours rather than weeks.
The South African institutions that will define the next decade are the ones deploying proven intelligence, redirecting their talent toward genuine differentiation, and compounding those advantages faster than FIs building from scratch possibly can.
Loading ...