Beyond basic chatbots: How agentic AI Is transforming South African finance and security - IOL

As AI evolves from simple assistants to autonomous agents making complex real-world decisions, South African financial institutions face both massive operational opportunities and sophisticated new fraud threats. Picture: Sumsub

For most people, using artificial intelligence simply means relying on the technology to assist with numbers, polish text documents, or occasionally generate images. The future, however, paints a distinctly different picture as more individuals and organisations begin to embrace and rely on agentic AI.

But what exactly is agentic AI? At its core, it means trusting AI to make complex, consequential decisions on your behalf.

It goes far beyond asking an AI model to draft an email or answer a straightforward query. Instead, agentic AI takes on far more intricate tasks—such as independently managing investment strategies, executing multi-step business operations, or booking end-to-end travel itineraries.

A few weeks ago, I attended an insightful roundtable lunch focused on closing the trust gap with AI agents and exploring how they are actively transforming banking, payments, and business workflows. The event was organised by Sumsub—an AI-powered trust infrastructure provider that unifies identity verification, fraud prevention, transaction monitoring, and risk management workflows to optimise compliance efficiency.

Led by Mick Amelishko, AI Advocate at Sumsub, the discussion unpacked what agentic AI could mean for South Africa as local adoption steadily expands. Hannes Bezuidenhout, VP of Sales for Africa at Sumsub, and Jarryd Jensen, Regional Director for Southern Africa, also shared their industry expertise during the session.

Amelishko emphasised: “The biggest shift moving into the 'agentic' way of working is agency. Last year, AI was just an assistant—it helped check writing or offered basic suggestions. Now, AI can access the external world, read the live internet, interact with other digital products, make independent purchases, register on websites, or interact directly with bank accounts if granted permission.”

Amelishko went on to explain that for centuries, society has operated on the foundational assumption that every transaction or action was executed by a human. Today, that human element is being gradually displaced.

For many, this rapid shift is understandably a cause for concern. However, equipped with proper education and a clear understanding of how to govern agentic AI, these anxieties can be mitigated to a large extent.

Despite these safeguards, the threat of AI-driven fraud remains a rapidly escalating challenge. The situation calls to mind Kyle Reese’s famous warning about the Terminator in the classic 1984 film: “It can't be bargained with. It can't be reasoned with. It doesn't feel pity, or remorse, or fear. And it absolutely will not stop... ever, until you are dead!”

While Reese was describing a relentless cybernetic killer, the quote vividly captures modern anxieties surrounding autonomous digital threats—minus the physical violence, but with equal persistence in exploiting digital vulnerabilities.

Turning to the local landscape, Jensen highlighted how enterprise banking trends are currently unfolding within South Africa.

“Major enterprise adoption of true AI agents in South Africa is still quite low compared to global markets,” noted Jensen. “South Africa boasts a sophisticated market with advanced digital banking and instant payment systems, and consumers are already very comfortable with digital onboarding. However, full agentic adoption is still in its early stages.”

Adding to this perspective, Bezuidenhout observed that while forward-thinking institutions like Discovery Bank are actively pushing into the agentic domain, South Africa as a whole still lags behind international frontrunners like China.

One of the primary friction points surrounding agentic AI centres on accountability: what happens when something goes wrong, and who bears the ultimate blame—the human who issued the original prompt, or the machine that executed the action?

The prevailing consensus at the event was that AI itself cannot be legally blamed for unlawful activities; accountability must always trace back to a human operator or system controller. Nevertheless, novel methods of financial fraud are evolving rapidly, and regulatory frameworks often lag behind these technological shifts.

Illustrating this growing threat, data from Sumsub reveals that while the global identity fraud rate moderated slightly to 2.2% in 2025 (down from 2.6% in 2024), instances of highly sophisticated fraud surged by 180% year-over-year as bad actors shifted away from low-effort scams toward advanced, AI-driven attacks.

Ultimately, as agentic AI shifts from novel experiment to core financial infrastructure in South Africa, balancing automated convenience with vigilant human oversight will be the key to maintaining security and trust.