Agentic commerce is one of those buzzwords that has been floating around fintech circles for a while now. You might have heard it in conference keynotes or read it in trend reports. But for all the hype, one major player has kept its cards close to its chest. That player is JPMorgan Payments. Prashant Sharma, the firm’s executive director of biometrics and identity solutions, recently spoke with American Banker to break the silence. He offered a rare glimpse into how the banking giant views this emerging space and, more importantly, how liability is shifting in the age of large language models.

Let’s be honest. When a company as big as JPMorgan stays quiet on a topic, it usually means they are watching closely. Sharma confirmed as much. He explained that agentic commerce, which essentially refers to autonomous agents or AI systems making purchasing decisions on behalf of humans, is not a distant fantasy. It is already taking shape. But the reason JPMorgan has been ‘pretty quiet’ is that the implications around responsibility and risk are still being untangled.

The Shift in Liability That Keeps Bankers Awake

Here is the crux of the matter. Traditional commerce is simple in terms of liability. A human initiates a transaction. That human is responsible. If something goes wrong, say a stolen card or a disputed charge, the rules are clear. But what happens when an AI agent, powered by a large language model, decides to buy something? Who is on the hook? The human who gave the agent general instructions? The developer who trained the model? The bank that processed the payment?

Sharma pointed out that this is a fundamental shift. It changes everything from fraud detection to authentication. In a world where machines act on our behalf, the concept of ‘consent’ becomes murky. Biometrics and identity solutions are no longer just about verifying that you are you. They must now verify that your AI agent is authorized to act within certain boundaries. It is a headache, but also an opportunity for innovation.

Why Large Language Models Change the Game

Large language models are not just smarter chatbots. They are decision engines that can negotiate, compare prices, and even execute payments. Imagine a future where your virtual assistant books a flight, pays for it, and arranges travel insurance, all without you clicking a single button. That is agentic commerce in action. But it also introduces a new class of risk. What if the model misinterprets a request? What if it is tricked by a malicious prompt?

Banks are understandably cautious. They have spent decades building systems around the assumption that a human is the final decision maker. Giving that power to software feels like a leap of faith. Sharma compared it to the early days of online banking, when many were terrified of typing their credit card number into a website. We got over that fear because security caught up. The same will happen here, but the transition will be messy.

How Virtual Cards Fit Into the Puzzle

This is where virtual card technology steps into the spotlight. If AI agents are going to spend money on our behalf, we need tools that provide control without sacrificing convenience. For example, you might give your agent a virtual card with a strict limit, a merchant category restriction, or an expiration date tied to a single transaction. This is not a theoretical concept. It is already being used by businesses and individuals who want to automate payments without exposing their main accounts.

For those looking to experiment safely in this space, VCCWave offers a trusted and free virtual card generator service. It allows you to create cards that are perfect for testing AI driven purchases or managing subscriptions. Think of it as a sandbox for the agentic economy. You can set your own parameters and see how the technology behaves without risking your primary funds. It is a simple, effective way to take control before the banks figure out the liability puzzle.

What This Means for Merchants and Fintechs

For merchants, agentic commerce could be a double edged sword. On one hand, AI agents could streamline purchases and reduce cart abandonment. On the other, they might introduce new friction at checkout. How do you verify that an agent is legitimate? How do you handle returns when the buyer was an algorithm? These questions will need answers, and quickly.

Fintech startups have a chance to lead here. While large incumbents like JPMorgan move slowly, smaller players can build specialized solutions for agent identity verification, transaction limits, and real time monitoring. The key is to focus on the user experience. Adults do not want to micromanage their AI. They want to set it and trust it. That trust will only come from robust systems that are transparent about what is happening under the hood.

A Word on Biometrics and the Human Element

Sharma emphasized that biometrics will play a massive role. Voice recognition, facial scans, and even behavioral patterns could be used to confirm that an agentic transaction aligns with the user’s intent. But here is the funny thing. We are talking about creating digital fingerprints for machines. It sounds like science fiction, but it is already in testing. The goal is to make the invisible visible. To ensure that when an AI spends money, there is an unbroken chain of accountability that leads back to a person.

And yes, there will be hiccups. There will be stories of AI agents buying 1000 paper clips by mistake or booking a hotel in the wrong city. Those moments will test our patience. But they will also drive the industry to build smarter safeguards.

Looking Ahead: The Quiet Giant Will Speak Soon

JPMorgan’s silence on agentic commerce has not been a sign of disinterest. It has been a sign of caution. They are waiting for the landscape to mature before making their move. That is smart. But the silence will not last forever. Once the liability frameworks are clearer, expect a flood of products and partnerships from the bank. Agentic commerce is not a fad. It is the next logical step in a world that is becoming more automated every day.

The big lesson here is that convenience and responsibility must walk hand in hand. If we get the balance right, agentic commerce could free up human time for more creative and meaningful tasks. If we get it wrong, it could become a regulatory nightmare. For now, the quiet observers are the ones you should watch. They are the ones building the safety nets before the circus arrives.