The Green Sheet Online Edition
May 5, 2025 • 25:05:02
Insider's report on payments
AI set to change the shopping experience
You know agentic AI has gone mainstream when Walmart enters the fray. The retail behemoth is exploring how to adapt as consumers begin using AI agents to handle their shopping, the Wall Street Journal reported.
Walmart is not alone. In April Chat GTP posted to X that it was "experimenting with making shopping simpler and faster to find, compare and buy products." AI has already changed the way consumers research products. Chat GTP, said in its X post that it would provide consumers with "improved product results, visual product details, pricing and reviews."
Chat GTP is the fifth most visited site on the web. Only Google, Youtube, Facebook and Instagram record more visitors. It's only a matter of time before AI, specifically agentic AI, plays a role in the way consumers pay at the checkout. Agentic AI is already being used to address fraud.
Agentic AI, refers to systems and models that act autonomously with the objective of meeting the specific goals and objectives of users. Agentic AI "is one of the most important developments in AI right now, and it's more than hype," Mark Sundt, chief technology officer at Stax Payments, stated in an email exchange. For example, he noted, agentic AI can be used to support interchange optimization. AI agents can assess how a terminal is configured and recommend changes that reduce fees or improve approval rates.
Fraud controls
With agentic AI, AI bots can be used to solve problems dynamically with task-focused agents that collaborate and communicate. Simple as well as complex problems. In this way, agentic AI differs from generative AI models like Chat GPT. Agentic AI is focused on making decisions rather than creating content.
Financial research firm Moody's mentioned in a recent report that unlike "traditional models, which often require manual recalibration, agentic AI continuously refines its predictions based on new data."
Also, unlike generative AI models, agentic AI models have no need for human prompts; they are set to particular goals or objectives, such as maximizing sales, and carrying out complex sequences of activities. "It's about small, task-focused agents that collaborate, communicate and solve problems dynamically," Sundt said.
Sundt offered the example of a merchant having problems with a terminal. To solve the problem, he stated, one AI agent might check the firmware, another might diagnose the network and a third could verify account status, "all in the background, no ticket required." Stax uses agentic AI extensively but keeps humans in the loop, he added.
A key focus of agentic AI today is merchant onboarding and risk vetting. "We're training agents to analyze signals like business longevity, website presence, and banking history to detect potential fraud before onboarding begins," Sundt wrote. But they don't make final decisions. Every decision flows through human validation before action is taken, he added.
As with every advancement in technology, there is a downside. While the power of agentic AI is undeniable, it also introduces new risks, notably fraud risks. "AI can be weaponized just as easily as it can be used for defense," Sundt said, adding that Stax is "laser focused" on using agentic AI to bolster fraud detection, notably during merchant onboarding and settlement review.
Sundt mentioned, for example, monitoring merchants for refund patterns, like refunding transactions to different cards or batch manipulation. "These aren't one-off issues – they're systemic risks we can now catch faster using AI," he noted. While agentic AI creates new possibilities, it's not a replacement for people; it's a framework, Sundt explained. "When used thoughtfully, it gives us the ability to solve very targeted problems, move faster, and surface risks earlier," he said. It's not about cutting corners. "It's about aligning the technology with what really matters in our ecosystem: safety, speed and reliability."
Moody's also addressed this issue. "At Moody's, we strongly believe that AI is not a replacement for human expertise but a catalyst for more informed, efficient and resilient financial decision-making. By embracing AI as a tool for augmentation rather than substitution, institutions can unlock new opportunities to enhance risk management," the firm wrote.
Payments: the next frontier
With the containment of fraud front and center, the next big step for agentic AI is payments. This is no one-off pie in the sky idea. Dominant card brands Visa and Mastercard and payments powerhouse PayPal all have agentic payment schemes in the works.
Visa revealed in April powerful AI advancements that help consumers find and buy online with AI. "Historically, Visa has used AI to protect consumers, harnessing it to combat fraud," Visa CEO Ryan McInerney said during an April product drop. "Now we will also enable AI to empower consumers, fundamentally shifting digital commerce to make it more personal, more relevant and more delightful."
McInerney added that "For any AI commerce use case to take hold, the payment is a critical enabler of success. If there is no payment, there is no commerce. That's the expertise and trust that Visa brings."
Visa is collaborating with AI platforms and brands that consumers work with regularly, including the AI company Anthropic, Microsoft, Mistral AI, OpenAI, Perplexity (an AI-powered search engine), Stripe and Samsung.
"We see tremendous potential for the role AI agents will play in commerce from streamlining 'regular' transaction-driven tasks such as ordering groceries, to more sophisticated search and decision-making like securing that hard-to-get restaurant reservation or concert ticket," said Jack Forestell, Visa's chief product officer and strategy officer.
"This," he added, "will be a transformative change, bringing more magic and convenience to the consumer experience and creating a new world that will forever change how we shop and buy."
Also in April, Mastercard introduced Mastercard Agent Pay, which builds upon proven tokenization capabilities that already power global commerce solutions like mobile contactless payments, secure card-on-file, recurring payments and subscriptions. "This helps unlock an agentic commerce future where consumers and businesses can transact with trust, security and control," Mastercard said in a press release.
The card company added that it will work with acquirers and checkout players like Braintree and Checkout.com to enhance existing tokenization capabilities that are being used with merchants to deliver safe, transparent agentic payments.
By identifying and validating a customer using Mastercard's tokenization technology, a retailer could offer a meaningful and consistent shopping experience, layering on relevant personalized benefits, such as recommended products, rewards, discounts and free delivery.
Mastercard is working with Microsoft to integrate Microsoft's leading AI technologies, including Microsoft Azure Open AI service and Mastercard Copilot Studio, with Mastercard payment solutions
"Mastercard is transforming the way the world pays for the better by anticipating consumer needs on the horizon," said Jorn Lambert, chief product officer at Mastercard. "Recognizing the seismic implications of this evolution, we are keen to collaborate with industry players to advance the advance standards for agentic payments, such as applying model context protocol to secure remote commerce. This lays the foundation for scale and build trust in agentic commerce."
Consumers will have complete control over what the agent is allowed to purchase on their behalf, Mastercard noted.
Both Mastercard and Visa are using tokenization technology. This enables payments to be initiated through conversational interfaces and conducted at the millions of merchants of all sizes supporting online commerce. Mastercard, in a press release, said it will protect against fraud and support consumer disputes with "best-in-class" cyber, security, and authentication capabilities to protect merchants and consumers from bad actors, end-to-end.
PayPal agent toolkits
PayPal is taking a different approach than Visa and Mastercard. It introduced PayPal Agent Toolkit to help developers seamlessly integrate PayPal's comprehensive suite of APIs – including those for managing payments, invoices, disputes, shipment tracking, catalog, subscriptions, reporting and insights – into various AI frameworks.
The toolkit is a library designed to simplify the integration of PayPal's core commerce functionalities into AI agent workflows. The idea, PayPal stated, is to empower businesses with their own agents "facilitating seamless connections to PayPal services for process workflows and task generation for use cases." These include:
- Order management and shipment tracking: Developers could design a customer service agent to use the toolkit to process an order and pay for it using PayPal with buyer authentication. For example, an agent could create an order in PayPal when a user clearly confirms the purchase through a conversational interface, and authenticates the purchase by logging into their PayPal account.
- Intelligent invoice handling: For example, an AI assistant could generate a PayPal invoice based on details of a service provided and email it to the client, PayPal explained in a briefing paper.
- Streamlined subscription management: Agentic AI can be used to manage an entire subscription lifecycle, from creating new products, subscription plans and processing recurring payments through PayPal.
The consultancy McKinsey & Co. recently opined that AI has the potential to be as transformative as the steam engine was to the Industrial Revolution. But we are only at the very beginning. While nearly all companies invest in AI, just 1 percent believe they are at maturity. Will agentic AI in commerce change this outlook? It remains to be seen. 
Patti Murphy is senior editor at Green Sheet and president of ProScribes Ink, www.proscribes.net. You can also follow her blog, Today in Payments, at Todayinpayments.com, and her podcast, Merchant Sales Podcast, co-hosted with James Shepherd at www.ccsalespro.com/
Notice to readers: These are archived articles. Contact information, links and other details may be out of date. We regret any inconvenience.
