Advice challenges mount as FSPs accelerate AI adoption
Artificial intelligence (AI) is set to reshape South Africa’s financial services sector in ways few appreciate, introducing a range of unforeseen risks. As financial services firms start experimenting with AI in client servicing and product design, financial advisers will have to go far beyond simply reassuring clients who have ‘bet’ the proverbial farm on a crypto asset or gone all in on a US-listed chipmaker.
Consumers in an AI Age
The potential impact of the unfolding AI Age on consumer outcomes is not lost on the country’s joint financial sector regulators. In Artificial Intelligence in the South African Financial Sector, the Financial Sector Conduct Authority (FSCA) and the Prudential Authority (PA) offer a comprehensive overview of AI adoption, machine learning (ML) and generative AI (Gen AI) within South Africa’s financial institutions. The report was informed by a survey of banks, insurers and investment managers carried out in 2024.
The foundational finding is that there has been a steady increase in AI usage, led by banks and payment providers. “Banks are at the forefront, with 52% of banking institutions actively employing AI, followed by payment providers at 50%,” notes the FSCA media release accompanying the report launch. The conduct authority mentioned that most survey respondents were investing R1 million or less in the technology, though bank respondents anticipated investing over R20 million in AI tech during 2024.
Your writer was quite taken aback by the “most institutions plan to invest less than R1 million in AI” finding, and the R20 million estimate thrown out for banks seems very light too. But drilling into the report confirmed these numbers. It stated: “Of the 21 banks that have adopted AI, 45% planned to invest more than R30 million over 2024; in contrast, of the relatively few insurers and investment providers that have already adopted AI, planned investment for 2024 was comparatively lower, with 62% of investment providers and 41% of insurers [saying they] would spend less than R1 million.”
Seven learnings from the AI report
The FSCA mentions seven learnings from the report, and financial services providers (FSPs) would do well to reframe their AI strategy with these points in mind. “Together, these learnings reflect a shift towards a more inclusive, transparent and resilient financial system that leverages AI responsibly to benefit both institutions and consumers,” the authority wrote.
First and foremost, institutions were encouraged to adopt recognised explainability methods to ensure AI-driven decisions are understandable and auditable. Explainability refers to the tools and techniques that show how an AI model arrives at a particular outcome, allowing humans to interrogate the inputs, logic and weightings behind the decision. The report found that one-in-five financial institutions were not using any form of explainability method, raising concern around the governance of traditional AI technologies in finance. Per the survey, banks are doing far better in this area than insurers and investment managers.
“Financial institutions should consider comprehensive governance structures, including strong data governance, model risk management and board-level oversight, to ensure ethical and effective AI deployment,” wrote the FSCA in its second learning. According to the report, governance frameworks “are essential for ensuring that AI systems are used responsibly, transparently and ethically.” Firms were encouraged to develop governance frameworks that span accountability and delegation; data governance; board oversight and ethical guidelines; risk management; and model development, validation and ongoing monitoring.
GDPR and POPIA compliance
At this early stage, the joint authorities are concentrating on consumer protection and data privacy. “The FSCA and PA would like to collaborate closely with the Information Regulator to ensure alignment with the Protection of Personal Information Act (POPIA), particularly in relation to data privacy and consumer protection,” the FSCA wrote. According to the report, participants prefer principle-based over rules-based regulation to drive these outcomes. To reframe slightly, consumer outcomes should not be negatively impacted by the adoption of AI or other technology.
The fourth learning was somewhat counterintuitive unless viewed through a regulatory lens. The FSCA suggested that AI use cases should be prioritised so that their benefits are maximised and associated risks effectively managed. “High-impact use cases include fraud detection and prevention, risk management, regulatory compliance, credit scoring and lending and consumer protection,” noted the report. But this narrowing of AI applications is unlikely to be well received by financial institutions that want to leverage the technology across their businesses to achieve competitive advantage.
Ethical standards and oversight featured prominently throughout the report. Under this learning, the FSCA called for “the development of sector-wide guidance for ethical, fair and responsible AI … alongside enhanced oversight to mitigate bias, inaccuracies and consumer harm.” The good news is this level of oversight need not result in a slew of new regulations. South Africa’s financial conduct regulation, exemplified by the soon-to-be enacted Conduct of Financial Services (COFI) Bill, is already based on six Treating Customers Fairly (TCF) principles.
The sixth learning is for efficient and effective disclosure when AI is used in consumer-impacting decisions. Inadequate disclosure was singled out as a key risk in the AI-in-finance context. “Limited disclosure of AI usage in finance can lead to misinformed decisions due to a lack of transparency about how AI influences financial products and services,” the report noted. “This can result in unexpected outcomes, such as unanticipated fees, biased lending practices or privacy concerns.”
Your writer agrees but wonders how financial institutions will ever be able to explain the utilisation of AI in financial products or services, let alone get consumers to pay attention to or understand them. To illustrate, choose any one of the applications that you use on a daily basis, perhaps LinkedIn or Microsoft or WhatsApp. Now read the terms and conditions of the product. Best guess, it will take you well over an hour to get through the multi-page disclaimers and a lifetime to understand how that service applies the various technologies it refers to.
Awareness and education
And that brings us to the seventh and final learning, which centres on AI and digital literacy. “Promoting consumer awareness and education will be critical to ensure that individuals understand how AI affects their financial decisions and rights,” the FSCA wrote. Agree, in part. The question becomes how a country with South Africa’s dismal English and maths literacy can achieve the dual goal of affordable and widespread access to financial products and services with full customer understanding of the technology that enables it.
The FSCA and PA have made it clear that AI’s future in the financial sector will depend as much on responsible deployment as on technical progress. For those in advice and distribution, the focus should now shift to understanding how these systems shape advice processes and influence client-centric decision-making. The task is to identify and mitigate the risks that AI introduces into advice while ensuring that client outcomes remain front and centre as the technology becomes embedded in financial product and services.
Writer’s thoughts:
The accelerated deployment of AI in finance will test how well advisers can balance innovation with fair client outcomes. You know your own processes, but how confident are you in the AI practices of the product suppliers you rely on? Please comment below, interact with us on X at @fanews_online or email us your thoughts [email protected].
Comments