AI agents are Financial transactions are already taking place on behalf of the people, and South African banks were not created for this. This was the central warning at Nedbank's Innovation Day in Johannesburg last week, where banking and technology leaders said the industry's biggest blind spot is not AI itself, but who, or what, is on the other side of the transaction.
Every bank account, every credit product, and every fraud-detection system in the country was designed for a human customer: a person with a face, a salary, and a credit history. That notion is now under pressure.
david kerriganThe problem goes much deeper, said Michael Schwartz, an author and technology analyst who lectures at Stanford's Continuing Studies Program and consults for Mastercard and Innovation. “Everything they built was based on the basic assumption that they were serving human customers,” he said.
Kerrigan told TechCentral that today, someone can open an app, instruct an AI assistant to book flights, compare insurance options and complete payments, and never have to personally authorize each step. “The agent does it. The bank processes it. But the bank's system wasn't designed to know the difference.”
The first problem is identification. Banks verify humans using faces, fingerprints and ID numbers. An AI agent has none of this.
Chipo MushwanaNedbank's executive for payments and technology said a new approach was needed. “You need a different framework to assess the identity of an agent,” she said. “When was it built? Who owns it? How long should it last?”
Certification for Agents
Mastercard and Visa are already building authentication systems for agents, he said — work that matters because credit is the next complexity. The lending bank needs to assess the ability to repay, which always means looking at a person's income. The entrepreneur running multiple revenue-generating AI agents doesn't fit that model.
“Maybe we need to make decisions that are a reflection of someone's ability to pay at a given time,” Mushwana said, “and not necessarily tied to traditional income cycles.”
Reading: AI models have scared the world's biggest banks
Nedbank says it is already using agents internally to speed up its credit-scoring system. The next step, Mushwana said, is to build systems that can recognize when an agent is on the other end.
A big question is whether banks remain relevant at all as agents take over a greater share of transactions.
“Banks always worried about being the top of the wallet, the first card you use,” Kerrigan said. “But AI agents will check for each purchase you make to see what is the best payment method for it.”
Agents will have accurate pricing and product information and will route payments to the option that will yield the most savings. A retailer may offer a discount for account-to-account payments by card; A human will never notice, but the agent will always notice.
Mushwana said banks would not be completely bypassed, as payments still needed to be validated, traceable and reversible. “You need a function that validates agents, validates the transaction, reports it as true,” she said. “The responsibility is borne by someone.”
But the game has changed. The question now is not which bank app is easiest to use, but which bank infrastructure agents will choose to use.
South African banking is highly regulated. The Financial Intelligence Center Act – better known as FICA – requires identity verification; The National Credit Act regulates lending. None of that law contemplated a non-human transacting party.
This gap is not unique to South Africa, Kerrigan said. “The law they have is barely there to capture AI, and it wasn't ready for agentic AI then.”
'Industry-wide growth'
sicko thomasNedbank's group managing executive for personal and private banking acknowledged the issue. “As more autonomous capabilities emerge, the regulatory conversation will evolve, particularly around liability, consent, and accountability,” he told TechCentral. “At this stage this is not a bank-specific issue, but an industry-wide development.”
On liability, Kerrigan said the law currently defaults to the human. “In most cases, the responsibility lies with the human, not the people who created the AI.” He pointed to work by one of its customers, Mastercard, as an early effort to bridge that gap.
Reading: Investec's reverse AI bet: people over machines
Card networks have really made agentic payments a priority. Mastercard launched Agent Pay in 2025, which issues “agent tokens” that link card credentials to a specific AI agent, and in March added Verifiable Intent, an open, cryptographically signed record that links the cardholder's identity and instructions to what the agent ultimately does, so intent can be proven if a payment is disputed. Visa established a rival approach in its Trusted Agent protocol, and the two have since joined a broader industry effort to standardize how an agent proves who authorized it.
Thomas said Nedbank is generating 48 million personalized interactions using machine learning, has seen a 70% year-on-year increase in AI-driven activity and has more than 2,000 AI agents in internal use. “Our ability to respond is not about catching up,” he said. “It's about accelerating from an established base.”
Mushwana was more direct about the actual trial. “We all have access to the same capacity. We all have access to the same technology,” he said. “It's just, is it valuable to our customers? Are they willing to pay for it?”
Reading: Nedbank, Jumo bet on AI lending for the underbanked
The banks that answer that question first will not only win customers. They will decide how AI agents spend everyone else's money. – © 2026 NewsCentral Media