The Green Sheet Online Edition
July 7, 2026 • 26:07:01
Insider's report on payments
Agentic commerce: not quite ready for prime time
Glance, the consumer technology company of AI company InMobi, in late June said it was launching a first-to-market agentic commerce capability for Samsung smart televisions in the United States. The integration, built on Glance's proprietary AI platform runs natively on Samsung's Tizen OS, "turning the living room screen into a generative-AI-powered, interactive shopping experience," the companies said in a press release.
The capability is available on all Samsung television models released since 2020 and can be accessed from the apps tab on those TV sets.
"Samsung has always believed the television is more than just an entertainment device – it's the centerpiece of the home," said Maya Harris, vice president for business development and strategic partnerships at Samsung. "Glance's agentic shopping experience brings that belief to life, meeting consumers in moments of inspiration with personalization and intelligence that only this screen can deliver."
The news caught my eye and harkened me back to an interview I conducted in the early 1980s with a consumer EFT expert at the Federal Reserve Board. ATMs were relatively new at the time, and some pioneering banks (like the bank known then as Chase Manhattan) were experimenting with at-home banking, which leveraged television screens and telephone lines to allow consumers to perform routine banking tasks like balance inquiries and bill-pay. (Consumer-grade desktop PCs were still a few years off.)
I had asked this expert what he thought the future held for EFT. And his response, in retrospect, was eerily prescient. I no longer recall his exact words, but what he said was something to the effect of "we're going to get to the point where every kind of transaction can occur on your television set, with the exception of dispensing cash."
Announcements come in rapid fire
The Glance-Samsung announcement was one of many involving agentic commerce to hit the wires in recent weeks. Agentic commerce—shopping powered by AI agents—"represents a seismic shift in the marketplace," the consultancy McKinsey & Co. observed recently.
McKinsey's research suggests that by 2030 the business-to-consumer retail market could see up to $1 trillion in revenues generated from agentic commerce; global projections run between $3 trillion and $5 trillion.
The investment bank Morgan Stanley has a more modest prediction; it expects $385 billion in U.S. ecommerce spending will be tied to agentic commerce by 2030. Still a sizable sum. Regardless of the total spending expected, there will be, McKinsey noted, both benefits and risks for the commerce ecosystem, with all types of businesses needing to adapt to navigate the challenges of trust, risk and innovation.
In early July, the financial technology company Nuvei, said it had completed a proof-of-concept with Visa, Arvato Systems and the fashion brand Kings and Priests, where the merchant's AI agent initiated a product purchase on a shopper's behalf and paid inside the agent, with no hand-off to a separate payment flow.
Multiple issuers across Europe took part in completing agentic payments settled live on Visa rails using a tokenized Visa credential within Visa Intelligent Commerce, the card company's AI protocol. And everything was governed by shopper-set guardrails, including spending caps and approved purchase categories, Nuvei explained in a July 2 press release.
"Agentic commerce is the next evolution of digital commerce, with AI not just finding products but initiating purchases," said Phil Fayer, chair and CEO of Nuvei. "This proof of concept starts inside a merchant's own experience and points to where payments are heading: a layer that lets any agent, on any protocol, make a payment."
Also in early July, BBVA, headquartered in Madrid, and Visa, revealed they had completed an AI-initiated transaction for a cardholder as part of a program Visa calls Visa Agentic Ready.
To support the strong customer authentication (SCA) requirements of the European Union, the transaction, in addition to Visa Intelligent Commerce, used payment passkeys, a biometric authentication method that enables consumers to authorize online payments securely without passwords or SMS codes.
"This milestone within Visa's Agentic Ready program confirms that agentic payments can function in a real-world purchasing environment while fully maintaining cardholder consent, issuer oversight and compliance with existing protections," the two companies stated in a press release.
'Very early stages'
So where does all this activity place agentic commerce on a developmental scale? "I think it's very early stages," said Jonathan Razi, founder of Findustry AI. "And right now I think it has more builders than users."
Razi, the Harvard-educated attorney and founder of CardX, a financial technology platform that automates credit card surcharging for merchants in compliance with card-brand rules and state laws, has now set his sights on what AI can do to support card payments.
Razi sees agentic commerce as a two-part automation process: product discovery and autonomous execution of the payment. He expects the most value for agents that can discover the best products for the best prices within the parameters set by the customer. "There's already so much friction that's been taken out of the checkout," he said. "There is not really much more that's achievable through AI."
"Agentic commerce is one opportunity in payments, but it's just one, and there are so many operating processes that can be remade with AI to take out all that manual work," Razi said in a recent interview. "And it's a whole new source of savings for the merchant." This is also a money-making opportunity for acquirers and their sales partners, he stated. Think in terms of fraud controls, reporting, disputes and other labor-intensive processes.
AI beyond payments
Razi, who sold CardX to Stax Payments in 2021, is building Findustry to leverage generative AI for labor-intensive areas of payments. For example, if AI agents are put to work executing payments on behalf of consumers, it's a safe bet that disputes will ensue, and disputes are already on the rise with the boom in ecommerce, Razi said.
Referencing data compiled by Visa, Razi noted that disputes rose 35 percent between 2019 and 2025. Hence the need for AI solutions that automate chargeback workflows.
Richard Crone, CEO of Crone Consulting LLC, offered a similar assessment. "For merchants, the next logical step is broader than payments," he said, adding it's about answering questions like should I reorder this SKU, change pricing, adjust labor, launch a campaign, change tender mix, draw credit, analyze disputes?
The real opportunities for merchants, and the answers to these questions, Crone said, will accrue from generative AI applications. (GenAI uses machine learning algorithms to generate text, images or other desired output.)
Crone pointed to the recent release of ChatGPT Finances, which can securely connect the chat bot to an individual's or business's bank, credit card and investment accounts. In March of this year, Intuit, the company behind Quickbooks and other popular financial apps, said it had integrated with Chat GTP.
But while most businesses are using Generative AI, they aren't all using ChatGTP, Crone added. For example, they might be integrating their business management software with Anthropic's Claude or Google Gemini. So, it's important that acquirers and processors have the technology in place to integrate with an array of AI engines, he stated.
"That means ISOs, payfacs and processors are not selling payment processing anymore; they are selling access to business reasoning and true 'enterprise resource planning [ERP]' made possible with bring your own AI," Crone said.
What are consumers saying and doing?
So what do consumers think about agentic commerce? The payment processor Worldpay recently queried 8,000 consumers across seven countries, compiling the results in a tome titled The agentic commerce report.
Globally, around 40 percent of consumers in the study said they are open to letting AI agents do their shopping; only 27 percent rejected the idea outright. Enthusiasm is highest in China, Worldpay reported, while the United States mirrors the global perspective.
Not surprisingly, age is the biggest dividing line, with younger consumers more inclined (50 percent, globally) to let AI agents do their shopping. Gender also plays a role, with men in western markets (like the U.S., U.K., France and Australia) more willing to use AI agents, while women lead interest in Brazil and Singapore.
Despite their interest, many consumers expressed concerns that could temper adoption of agentic commerce. Five concerns topped the list, being cited by more than half those surveyed. These were:
- Identity theft (cited by 55 percent of surveyed consumers)
- Incorrect purchases (55 percent)
- Unauthorized purchases (54 percent)
- Fraud (53 percent)
- Losing financial control (51 percent)
Across markets, consumers want reassurances that they can trust the process—the ability to intervene if things go awry. Fraud protections (cited by 54 percent), the ability to cancel transactions within 24 hours (50 percent), and the ability to review purchases before completion top the list of trust builders.
Even so, consumers are nowhere near ready to offload all their shopping onto AI agents. Within five years, consumers expect only about 10 percent of their shopping to be done using AI agents. In-store shopping is still preferred by a majority of consumers (37 percent), especially older consumers. "While AI and automation will enhance convenience and efficiency, physical retail will remain the cornerstone of commerce for years to come," Worldpay concluded.
Patti Murphy is senior editor at The Green Sheet, president of ProScribes Ink (www.proscribes.net) and self-described payments maven of the fourth estate. Her Today in Payments reports are a regular feature of the Merchant Sales Podcast.
Notice to readers: These are archived articles. Contact information, links and other details may be out of date. We regret any inconvenience.
