Image: Unsplash But in SME finance, speed is not the same as understanding. A faster answer can still be the wrong answer if the model behind it doesn’t understand the business in front of it. That is especially true in South Africa, where the real economy remains uneven, relationship-driven and anything but frictionless.
The latest data tells a more interesting story than the usual doom-or-hype binary. Xero’s 2026 State of South African Small Business report found that 84% of small firms are prioritising stability and steady growth this year, even though 80% reported revenue growth and 75% reported a profit increase last year. Stats SA, meanwhile, shows the informal sector still accounted for 21.4% of total employment in Q4 2025. South African SMEs are digitally active but not digitally uniform; formalising but not fully formalised. A business can look noisy on paper without being weak in reality.
So the next competitive edge in SME finance won’t come from “human versus machine”. It will come from knowing which parts of underwriting should be automated and which still need judgement. A customer might have irregular receipts because a large corporate is paying late. A seasonal business may look volatile in one month and exceptionally healthy across the cycle. A founder may run a textbook-untidy set of accounts and still have loyal customers, operational discipline and real growth.
That distinction matters because most of the people walking through our door aren’t scrambling. As our CEO, Brent Geddes, told Moneyweb back in 2019, “almost every person we’re lending to needs the money for growth or working capital, not to plug a hole.” A model trained only to eliminate uncertainty is liable to misread ambition as risk.
“A fast credit decision is not the same thing as an intelligent one.”
And then there is fraud. AI hasn’t only made lenders more powerful; it has made criminals more scalable. SABRIC’s 2024 crime statistics show digital banking fraud incidents rose 86% year-on-year to 97,975 cases, with losses up 74% to R1.888 billion. Banking app fraud accounted for 65.3% of those incidents, and SABRIC says social engineering – including AI-assisted scams – was a major driver. That should bury the fantasy that all friction is bad. In lending, some friction isn’t bureaucracy. It’s diligence.
None of that is an argument against technology. It is an argument for using it properly. The South African Reserve Bank’s review of AI in the local financial sector makes clear that AI in local finance has moved well beyond experimentation. The question is no longer whether lenders should use AI, they should. The real question is whether AI helps a lender understand context better, price risk more intelligently and spot fraud sooner, or whether it just helps them reject applicants faster and with greater confidence.
Globally, the most interesting examples are hybrid stories. Even the lenders held up as digital exemplars – Judo Bank, OakNorth and others – aren’t betting on software alone. They use technology to sharpen human decision-making, not replace it. If anything, digital maturity is making relationship banking more valuable, because the quality of the human layer now determines whether technology becomes an advantage or a black box.
There is another shift worth noting. SME finance is increasingly moving out of the branch and into the workflow. Embedded finance, credit served inside accounting platforms, payment rails and operational software is real and will grow. But invisible finance still needs visible accountability. Someone still has to own the decision when a good business is declined, a fraudulent application slips through, or a viable entrepreneur ends up saddled with the wrong facility.
“Embedded finance may make lending invisible, but accountability still has to stay visible.”
There is also a category point worth making. Pure fintech is fast but cold and nameless – efficient at processing, blind to context. What South African SMEs actually need is something different: what we call a TEAL, a tech-enabled alternative lender. Still relational, but with optimised speed and smarter decision-making.
That is why the phrase “relationship banking” deserves rescuing from nostalgia. It shouldn’t conjure a handshake, a golf day and a long lunch. It should mean something more rigorous: a lender with enough context to distinguish temporary stress from structural weakness, enough conviction to back a sound business others have misread, and enough accountability to explain why capital was or wasn’t extended. Cash flow tells you only part of the story; a lender also has to look at “the quality of the management team”. In a market where data can be patchy, payment cycles can be punishing and informality still sits close to the heart of the economy, that human layer isn’t an emotional extra. It is part of the credit architecture.
Daily Maverick recently noted that South African banks are increasingly chasing “trust and relevance” in the SMME sector, and that is exactly the right debate. The future isn’t branch nostalgia. It is high-tech, high-touch finance: software that reduces waste, data that improves signal, and experienced people who can still read what the numbers don’t say.
Relationship banking wasn’t killed by technology. It was exposed by it. For tech-enabled alternative lenders like Geddes, that means using automation and data-driven underwriting to make credit judgement more efficient, consistent and easier to scale. The firms that survive this next cycle will be the ones that use technology to make judgement better, not optional. DM
About Geddes Capital
Geddes Capital is a South African alternative funder focused on tailored SME finance.