Everyone will remember the hype before the pandemic about the so-called fourth industrial revolution. 4IR was everywhere. While we were distracted by a global shutdown, the 4IR has not gone anywhere. In fact, all eyes are trained on the single-most disruptive and powerful tool of the era: artificial intelligence. Comparisons to the industrial revolution are understandable, but perhaps a more precise comparison for AI is: dynamite.
Dynamite comes in small packages, but it is an exceptionally powerful tool. It revolutionised mining. Naturally, it needs to be handled carefully and respected because it carries inherent risk. AI has many of those traits: It can blow up productivity like you’ve never seen before, but it must be deployed responsibly and with discipline.
AI will not suddenly revolutionise payments. It already has.
There’s a misconception about AI in the payments industry. Speculators and commentators alike predict that generative AI will suddenly revolutionise how money moves. Reality is far more grounded. AI in payments is not futuristic. It is not something we are all excited to see unfold. Rather, it is two-decade old news. The industry has successfully leveraged machine learning and predictive analytics for 20-odd years to monitor transaction volumes and to pinpoint fraud.
The core care transaction flow is mature. It is well-defined and well-regulated. There is no ambiguity in how money transfers. There is no need for large language models (LLMs) or autonomous agents to augment the underlying card rails.
Take a moment to think about the evolution of online shopping authentication. We moved on from the clunky days of entering a 3D Secure v1 PIN for every transaction to 3D Secure v2. When you purchase something online today, an issuing bank sends transaction profiles to an AI risk agent. This is already happening. If you order groceries from the same place for consecutive days, the system will approve the transaction seamlessly. Hop on a plane to another country and try to place an order. You will be prompted for verification.
These highly sophisticated machine learning models are non-deterministic, and they occasionally misfire, such as flagging a routine monthly rental payment on the banking app as potential fraud. It happens to all of us. This is why human statisticians constantly retain the models, feeding them data to refine their accuracy over time.
Where LLMs introduce severe turbulence is in the environment surrounding the transaction flow, not inside it. Cybersecurity risks have multiplied exponentially. System infiltration, impersonation, and deepfakes have become so accessible that virtually anyone, from anywhere in the world with a computer, can launch sophisticated attacks.
Core defensive parameters such as firewalls and PCI DSS certifications remain robust. What has changed is the sheer volume of incoming threats. Like dynamite, it has exploded. Security operators have no choice today but to deploy AI to fight AI.
The true epicentre of AI disruption
The real frontier of AI disruption is in front-end commerce and consumer behavior. We are moving very quickly to a world where agentic commerce is pervasive. This is where personal AI agents on platforms such as Google Gemini or ChatGPT will handle end-to-end commerce tasks, such as booking a weekend getaway to another city and executing payments independently using tokenised Visa or Mastercard credentials.
It’s not a matter of “will this become mainstream”? It is coming, and merchants have no choice but to optimise their digital footprints for LLMs: clear Markdown data and Model Context Protocol (MCP) links for faster, deterministic searches.
Apps are already reshaping spending habits. On-demand delivery naturally encourages smaller, more frequent baskets, which drastically reduces impulse buys like chocolates in the checkout aisles. Meanwhile, retailers are pivoting toward dark stores stocked exclusively with fast-moving consumer goods to maximise margin efficiency. Dark stores are warehouses that are closed to the public and used exclusively to fulfil online shopping orders.
As is patently clear, the core card payment rails remain rock solid. They do not need generative AI to function better. Instead, AI’s deepest impact remains confined to risk management, cybersecurity defence and a reshaping and reimagining of the front-end retail journey.
Consumers are currently navigating a fine line between ultimate digital convenience and security fatigue. And so, while the backbone of the payments industry is somewhat isolated, AI is and will change how we all work, learn, play, and shop.
