All transactions, whether eCommerce or in brick-and-mortar settings, happen in a merchant’s environment.
Click a buy button, or present a card at the register, and the conversion from browsing to buying is complete.
But there’s a new wrinkle. It’s a semi-intelligent middleman, so to speak, that abstracts away some of the complexity for merchants’ customers. Artificial intelligence agents will reshape how consumers discover and make purchases, moving beyond what PYMNTS CEO Karen Webster has termed the familiar “search, scroll, click and buy” paradigm to a more intuitive “say it and get it” experience, all in a matter of seconds.
As a result, we’ll see a re-architecture of the entire commerce ecosystem, Rubail Birwadker, senior vice president and global head of growth at Visa, told Webster.
“Merchants spend an enormous amount of money getting noticed and getting their brand out there,” he said as part of the “Built to Buy” series focused on the next frontier of AI. “Traditional things like ads don’t really work, so how do they get their brands differentiated in an agentic world?”
There’s some urgency in answering those questions, given that agents offer an advantage that most commerce marketplaces don’t: context.
Merchants will need to adapt, Birwadker said, adding that by the end of 2024, ChatGPT was doing about the same amount of Google searches as Google itself was doing in 2009, and it took only two years for the model and its users to get to that point.
Visa’s architectural imperative, therefore, is to ensure its global, multimarket, multichannel ecosystem can accommodate this next frontier of consumer interaction and transaction.
The Automagical Experience
Visa’s approach to agentic commerce is grounded in the company’s core mission, which is providing an “automagical experience” linking merchants, consumers and issuers, Birwadker said.
“We have 4.5 billion cardholders, 200 million merchants and 14,500 financial institutions around the world, and when you have a Visa card in your wallet,” no matter where one might be in more than 200 countries around the globe, “the card works exactly the same way,” he said.
Visa has been through decades of shifts, from physical commerce to eCommerce, mobile commerce, app-based commerce and social commerce, he said.
For enterprises, the ability to tailor offerings more precisely to individual consumer preferences represents a leap forward in optimizing the merchant-customer relationship, if the right infrastructure is in place for the agents to make purchases on behalf of the cardholders.
The ideal is a “native” payment experience seamlessly integrated within the AI platform itself, simplifying the consumer journey and driving higher engagement, Birwadker said.
Navigating Concerns: Fraud, Liability and the Early Stages
Such streamlined commerce brings with it a new layer of technical complexity and a range of concerns for various ecosystem participants, from merchants and issuers to the agents themselves.
Merchants are worried about whether they’ll have a playing field comparable to larger competitors; fraud is a consistent concern. Where once eCommerce was about keeping fraudsters out, “the next era is that you’ve … got to let the good [customers] in,” Birwadker said.
The technological challenge, coupled with the need for standardization, becomes paramount as merchants consider how to interact with agents, ensuring that authenticated tokens from platforms like Visa are recognized and integrated to access pricing and inventory.
The industry is still in the nascent stages of integrating agentic AI into commerce. The payments piece of agentic commerce “is not super sophisticated yet, and there are very good reasons for it,” Birwadker said.
There is inherent complexity in payments, which deals with people’s money and requires strict adherence to laws, trust and security, he said. However, the pace of AI adoption has been underestimated, suggesting that while “we are in the very early part of the bell curve on this one,” the acceleration could be “relatively rapid.”
Central to Visa’s approach are specific frameworks and guardrails designed to manage risk and enhance the user experience, while keeping personal information out of the reach of fraudsters. This strategy builds on learnings from past innovations like mobile payments (e.g., Apple Pay and Google Pay), where “once you get past the provisioning stage, generally there is very little to no fraud,” he said. That’s because card usage is linked to a specific app, agent or device, and biometrics narrow the surface area that’s open to attack.
“There are also a number of pain points that we’re addressing” for issuers, Birwadker said, to make sure that authentication protocols are stepped up when appropriate. The agent community and payment processors also share concerns, particularly regarding liability in cases where an agent makes an error.
Birwadker posed a scenario that’s still a work in progress: “I asked for a blue bag, and the agent got me a black bag… How does that work through our dispute process? How does compelling evidence work in that case?”
Looking ahead, a key challenge for agent developers, and an additional area for Visa’s focus, is scalability. Birwadker said there is a common concern from developers, who inform the payments giant that, “‘I have 20% or 25% of my users in the United States. How would this work in Europe? How would this work in India?’”
Against that backdrop, network standards might offer a path toward global operability and consistency.
Ultimately, Visa views agentic commerce as a material channel that will complement existing commerce modalities rather than replace them, just as eCommerce did not eliminate physical commerce, Birwadker said.
“The simplest way to think about our vision of agentic commerce is just going back to basics … the automagical experience is the same, from the promise of the Visa brand to the actual payment experience,” he said.
Source: https://www.pymnts.com/news/artificial-intelligence/
