Machine Learning promises to revolutionise financial inclusion in South Africa, but new research reveals a complex picture of benefits and risks. This opinion piece examines whether ML is truly democratising finance or simply replacing old gatekeepers with new algorithmic ones
Image: Supplied.
Machine Learning (ML) is being positioned as a game-changer in financial services, but the latest research from Experian, conducted by Forrester Consulting, suggests that the reality is more nuanced than the tech industry would have us believe.
While ML is making inroads across financial institutions and telcos in South Africa and abroad, its promise of inclusion, efficiency, and profitability is still entangled in practical, ethical, and regulatory complexities.
The report’s headline takeaway is that ML is helping South African organisations reach thin-file and underbanked consumers, those who’ve historically been excluded from formal credit systems. By using alternative data sources, ML models can assess eligibility more accurately than traditional scorecards.
That sounds progressive, especially in a country where access to credit is often skewed by legacy systems and rigid criteria. But we need to ask: what kind of data is being used, and who controls it?
To be clear, Machine Learning refers to a branch of artificial intelligence that enables systems to learn from data patterns and improve their performance over time without being explicitly programmed. As Arthur Samuel, one of the pioneers in the field, famously defined it, “Machine Learning is the field of study that gives computers the ability to learn without being explicitly programmed.” In the context of financial services, ML models are trained on vast datasets to predict credit risk, detect fraud, and automate decision-making.
80% of ML adopters in South Africa say the technology helps widen access to financial services. That’s encouraging, but it also raises questions about consent, transparency, and fairness. If someone’s mobile usage, social media activity or location data is being used to determine their creditworthiness, are they aware of it? And more importantly, can they opt out?
The report also states that 71% of respondents see improved profitability through better risk prediction and reduced bad debt.
Automation is another area where ML is making its mark. 70% of users cite improved operational efficiency, and nearly 80% believe that most financing decisions will be fully automated within five years.
Generative AI (GenAI) is also entering the fray, with 77% of respondents saying it can reduce the time and effort required to develop credit risk models. More than 60% believe GenAI will streamline regulatory documentation and improve collaboration between risk and compliance teams. That’s promising, but it also introduces new risks. GenAI is still in its infancy, and its outputs can be unpredictable, biased, or difficult to audit. In a sector where compliance and accountability are non-negotiable, that’s not a minor concern.
Despite the optimism, the report acknowledges that many organisations remain hesitant. Cost, regulatory uncertainty, and lack of internal expertise are major barriers to ML adoption. Two-thirds of non-adopters believe the cost outweighs the benefits, and more than half admit they don’t fully understand the value ML can bring.
It suggests that ML is still seen as a high-risk, high-cost investment, one that’s not yet accessible to smaller players or those without deep pockets. Concerns around explainability and compliance also persist. 56% of non-adopters worry about model transparency, and 62% fear regulatory misalignment.
These are valid concerns, especially in South Africa’s complex financial landscape. Legacy IT systems and fragmented data infrastructure make ML deployment difficult, and the risk of unintended consequences, such as algorithmic bias or data breaches, is real.
The report notes that many concerns stem from misconceptions and that modern ML models can be both explainable and compliant. It also points to third-party platforms that can help bridge skills and infrastructure gaps.
Ferdie Pieterse, CEO of Experian South Africa, says ML enables lenders to grow responsibly and support social progress. Mariana Pinheiro, CEO of Experian EMEA & APAC, adds that ML is unlocking access to financial services for millions.
In South Africa, where inequality is entrenched and financial literacy remains low, the stakes are high. ML has the potential to transform the sector, but only if it’s deployed with care, transparency, and accountability. Otherwise, we risk replacing one set of gatekeepers with another and leaving the most vulnerable behind.
* Maleke is the editor of Personal Finance.
PERSONAL FINANCE