PhotonPay, the stablecoin-powered operating system for global payment infrastructure, unveiled its dual-rail recurring system.

Designed for emerging agentic commerce use cases, the system enables businesses to manage recurring payments across both fiat and stablecoin rails through a single integration. By abstracting the complexity of underlying payment protocols, it allows teams to focus on building and scaling AI-driven products while handling cross-border payment flows in the background.

The Intelligence Surge vs. The Infrastructure Gap

The digital economy is undergoing a structural shift. According to the 2026 SaaS Management Index¹, spend on AI-native applications has surged 108% in the past year, while large enterprises have scaled their AI SaaS expenditure by an unprecedented 393%. Yet, as the subscription economy approaches a $300 billion global market, the underlying payment infrastructure remains tethered to legacy banking constraints that are increasingly misaligned with the speed of AI.

The friction is most visible in cross-border commerce. Current estimates² indicate that 20% to 40% of subscription churn is caused not by product dissatisfaction, but by passive payment failures. For AI platforms operating globally, traditional credit card rails often trigger a cascade of declines due to rigid risk filters, currency conversion errors, and bank-imposed limits—frictions that are antithetical to the borderless, 24/7 nature of agentic commerce.

The Three Structural Frictions of the AI Era

To enable seamless agentic commerce, businesses must overcome three entrenched infrastructure limitations that current payment gateways fail to address:

1. Systemic Cross-Border Inefficiency

Traditional payment gateways remain optimized for the low-frequency, high-friction transaction models of the past decade. Their risk-scoring engines, designed for human-triggered commerce, often misidentify the high-frequency, low-value patterns typical of AI subscriptions as fraudulent. This results in excessive false positives, creating “invisible churn” that can suppress global conversion rates by up to 25%³.

2. The Stablecoin Continuity Gap

Historically, stablecoin payments have been treated as isolated, manual events rather than continuous financial flows. The absence of a native, programmable recurring billing layer has forced AI enterprises to rely on “one-off” invoices, creating massive friction in user retention. For the high-value, Web3-native segment, this lack of automation isn’t just an inconvenience—it is a break in the economic lifecycle that prevents AI platforms from building stable, predictable revenue streams on-chain.

3. Operational Fragmentation & Reconciliation Overhead

Current market solutions often force enterprises into a “dual-stack” reality: managing fiat through legacy gateways while handling digital assets via isolated non-custodial wallets. This infrastructure fragmentation creates deep operational silos, compelling finance teams to perform manual cross-chain and cross-bank reconciliation. As AI enterprises scale into multiple jurisdictions, this complexity becomes an exponential tax on growth, leading to reporting inaccuracies and liquidity bottlenecks.

Stablecoins as a Structural Component

Stablecoin-native payments address infrastructure frictions at the protocol level. Unlike credit card rails, on-chain transactions operate independently of banking authorizations—eliminating the primary drivers of involuntary churn, such as card expirations and arbitrary issuing-bank declines. For subscription-based enterprises, this transition secures the 2% to 5% ⁴ of monthly revenue typically lost to passive payment failures.

The structural advantages extend beyond reliability to fundamental economic efficiency:

• Cost Optimization: Stablecoin processing achieves high-margin efficiency with average fees of approximately 0.8%, a significant reduction from the 2.9% + $0.30 standard of legacy card networks⁵.

• Unrestricted Global Reach: By enabling any wallet-holder to subscribe, the infrastructure unlocks high-growth markets where traditional card penetration is low, yet Web3-native demand is accelerating.

Product Capabilities: Three Layers, One Protocol

Layer 1: The Consent Layer – Seamless Authorization

The core of the dual-rail experience begins with a singular, on-chain authorization. Once the user completes this initial step, the PhotonPay OS initiates recurring charges automatically, requiring no subsequent wallet re-signing. By mirroring the “set-and-forget” convenience of traditional credit card subscriptions, this layer transforms stablecoins from a fragmented, one-time payment tool into a reliable recurring billing infrastructure.

Layer 2: The Execution Layer – Adaptive Rail Selection

At this layer, the PhotonPay OS neutralizes the friction between diverse business models and the underlying payment rails. It provides the programmatic flexibility required for Agentic Commerce, ensuring that value movement is as dynamic as the AI consumption it supports.

• Adaptive Consumption Models

The engine natively supports fixed-tier SaaS subscriptions, high-frequency API-call billing, and token-based usage. This allows AI enterprises to align their revenue capture directly with real-time compute consumption.

• Autonomous Tier Escalation

Through dynamic tier billing, the OS automatically upgrades plans as usage thresholds are met. By removing manual intervention, PhotonPay ensures uninterrupted service delivery while maximizing lifetime value (LTV).

Layer 3: The Intelligence & Compliance Layer – Unified Reconciliation

The final layer leverages the Dual-Rail architecture to provide a single, compliant interface for global liquidity. It treats fiat and stablecoins as interoperable components of a unified corporate treasury.

• Unified Liquidity Intake

The OS dissolves the boundaries between legacy card networks and on-chain rails. Enterprises can capture value in any form—leveraging optimized fiat authorization rates or the borderless velocity of stablecoin-native settlement—through a single, integrated protocol.

• Unified Compliance Interface

All cross-rail activity is consolidated within a centralized dashboard, establishing a “Single Source of Truth” for global operations. This intelligence layer enables one-click export of audit-ready reports, ensuring adherence to regulatory standards across Hong Kong, the UK, and North American jurisdictions.

The Foundation: Stabilizing the Speed of Commerce

At its core, a stablecoin is value reimagined for the digital age—combining the stability of sovereign reserves with the boundless efficiency of blockchain. By removing geographic friction and the constraints of traditional banking hours, stablecoins synchronize the velocity of capital with the speed of information. This is more than a tool; it is the structural evolution of global finance.

“As commerce evolves towards AI-driven automation, the underlying economic interface must become programmable,” said PhotonPay Founder and CEO Lewison Chen. “At PhotonPay, we are integrating fiat and stablecoins into a single, seamless environment. Our goal is to provide the reliability of traditional finance with the agility of digital assets.”

Data Sources

¹ Zylo (2026): 2026 SaaS Management Index, reporting global SaaS market size of $408 billion in 2025, projected to reach $465 billion in 2026; AI-native application spend surged 108% year-over-year, with large enterprise AI SaaS expenditure up 393%.

² Aurpay (2026): Subscription economy data, estimating global subscription economy approaching $300 billion; 20%–40% of subscription churn attributed to payment failures.

³ Stripe (2025): Global Checkout Infrastructure Report. Analysis of authorization rate decay in high-frequency, cross-border SaaS billing.

⁴ Recurly(2025): State of Subscription Report. Statistical analysis of passive churn caused by credit card expiration and declining bank authorization rates in cross-border commerce.

⁵ Worldpay from FIS (2026): Global Payments Report. Comparative analysis of crypto-settlement efficiency vs. legacy card network Interchange and Scheme fees.

The post Fueling Agentic Commerce with Dual-Rail Recurring Billing appeared first on PaymentsJournal.

In a move to promote interoperable standards for agentic commerce, Google is donating its Agent Payments Protocol (AP2) to the nonprofit industry group the FIDO Alliance, which is working on open frameworks for agentic interactions. While framed as an altruistic step, the decision also helps ensure Google remains deeply embedded in this emerging space.

Google also announced a new version of AP2 with the support for “human not present” payments, enabling agents to autonomously execute transactions based on pre-authorized user instructions—such as buying concert tickets the moment they go on sale. To ensure those actions remain accountable, the company, in collaboration with Mastercard, introduced Verifiable Intent, an AP2-compatible standard that creates a tamper-proof record of user-approved agent activity. This standard is also being contributed to the FIDO Alliance.

These contributions position Google’s technology as a building block for future agentic developments—a move some observers described as characteristic of the company’s long-term strategy. Christopher Miller, Lead Analyst of Emerging Payments at Javelin Strategy & Research, calls the move “classic Google.”

“These ‘donations’ are part of the business plan by incumbents to remain relevant,” Miller said. “Make it free and cement their engagement. ‘You can all now build on this thing we developed.’”

Who Is FIDO?

Founded in 2021, the FIDO Alliance initially set out to replace password-based authentication with phishing-resistant protocols tied to user devices. It’s now expanding its scope through an Agentic Authentication Technical Working Group, focused on specifications for agent-initiated commerce, alongside a Payments Technical Working Group chaired by Mastercard and Visa.

FIDO’s efforts will center on agent authentication, ensuring actions are performed on behalf of legitimate users, as well as trusted delegation frameworks that keep transactions within user-approved parameters.

Google’s Agentic Ambitions

Google has made steadily advancing its role in agentic commerce. In January, it unveiled the Universal Commerce Protocol (UCP), an initiative designed to let AI agents manage the entire shopping journey from discovery to checkout.

Developed with partners including Shopify, Etsy, Wayfair, and Target, UCP features a streamlined checkout experience that allows users to complete purchases directly within Google’s AI Mode.

The post Google Donates Its Agentic Payments Protocol to the FIDO Alliance appeared first on PaymentsJournal.

Not long ago, fraud teams could keep pace by reviewing incidents one by one. That era is ending. Armed with artificial intelligence and cloud-scale infrastructure, today’s cybercriminals operate faster, more broadly, and with far greater sophistication than ever before.

The rise of agentic commerce will only intensify these challenges, in part because it upends a longstanding assumption in fraud prevention: that bot traffic is inherently suspicious. In a world where legitimate transactions may be initiated by AI agents, that distinction becomes far less clear.

In a recent PaymentsJournal podcast, AtData’s Diarmuid Thoma, Head of Fraud and Data Strategy, and Brandt Hoffman, Sales Director, Fraud Services, along with Jennifer Pitt, Senior Fraud Management Analyst at Javelin Strategy & Research, discussed how these shifts are dramatically impacting payments risk.

At the center of this transformation is a simple but growing imperative—organizations must know, with confidence, who (or what) is on the other end of every transaction. Achieving this now requires systems capable of analyzing and contextualizing vast, dynamic data streams in real time.

The Outputs of Scalability

Historically, many fraud attacks were treated as isolated events, leading financial institutions to adopt a reactive, situational approach. However, there are often patterns that emerge when these incidents are viewed collectively. Recognizing and operationalizing those patterns is critical.

“From a law enforcement perspective, I remember a mail theft case that I investigated,” Pitt said. “We conducted a search warrant on the suspect’s home and found bags of open and unopened mail. We also found stacks of paper that contained full personally identifiable information—name, date of birth, Social Security number, next of kin, last known addresses—you name it, he had it.”

“We searched his phone and his computer, and we were able to see that he was connected with several other suspects that we were already investigating,” she said. “What we uncovered was this hierarchical organized crime ring where there ended up being more sophisticated identity theft and other crimes. If we were just looking at one of those players or incidents, we wouldn’t have seen this whole organized crime ring.”

While traditional vectors like mail fraud persist, the digital landscape has allowed bad actors to expand their reach exponentially. Technologies such as AI and cloud computing have supercharged criminal capabilities faster than most organizations can evolve their defenses.

Beyond just deploying generative AI to create more convincing impostor sites and deepfakes, bad actors can now deploy AI agents to autonomously carry out widescale fraud campaigns. For example, agentic AI has been used in a technique where email addresses are rapidly and sequentially created for use in fraudulent activities.

“We see thousands and thousands of them every day, where we see sequential types of emails created and they’re not necessarily in one client,’” Thoma said. “Somebody’s using an email over here to create a bank account and going and buying a pair of sneakers over there.”

“Individually, it looks fine; there’s nothing wrong there,” he said. “At a platform level, we see the cumulative effect. It’s a simplistic example, but that type of behavior is a direct output of the scalability of fraud.”

Distinguishing Malicious Automation

Given agentic AI’s potential to amplify fraud across every channel, the emergence of agentic commerce presents unique challenges for fraud prevention teams.

Many of the open questions around agentic transactions center on authorization. In the conventional e-commerce model, the shopper selects items, completes verification, and explicitly authorizes the purchase. When an AI agent acts as the consumer’s proxy, however, new gray areas emerge.

“What happens in a chargeback scenario?” Thoma said. “The industry hasn’t got all the answers on that. It’ll slowly emerge, but one of the things that won’t change is history. It’s still you buying it. Especially for physical goods, it’s going to your physical location, it’s going to your name, and it’s probably using your e-mail address to confirm all the details. There’s still a lot of information, even in the agentic world, that’s going to be coming through.”

This means that one of the most important considerations for fraud prevention will be the user’s history. Fortunately, this data is already present for many consumers. For example, the organization can confirm the age of an email address, whether it has been actively used, and if there are any red flags associated with it.

This historical data becomes a critical point of continuity as organizations design fraud strategies for agentic commerce.

“It was always, ‘Let’s look at the negative aspects of what this transaction could present,’” Hoffman said. “Now, we have to be cognizant to bring in those positive signals. What are the good signals that we can lean on? What allows us to interpret or infer more quickly? How do we start to identify what it means to be a positive bot, or to be a good transaction along the line?”

A Timeline Event

To act on these signals effectively, teams must start from an accurate baseline. A core lesson from AI is that models are only as strong as the data that feeds them. Just as importantly, that data must remain current, especially as consumers’ digital footprints continue to expand.

“Many still look at data like it’s a credit report, where it’s a static thing that you see in a piece of paper and that’s it,” Thoma said. “It’s not. It’s a timeline event. If you think about when you were 20 to now, you’ve had different addresses, you’ve had different IPs and different devices. Your name may have changed for different reasons, and your email probably changed one or two times.”

“Your profile naturally evolves, so the importance of the data quality and the skill in the overlaying models is to know when that change is abnormal versus normal,” he said.

A practical way to evaluate changes in a user profile is through percentage-based shifts. Significant or rapid deviations across key attributes may indicate potential account compromise.

Similarly, the repeated use of a single element across multiple account creation attempts can signal synthetic identity activity, where bad actors combine real and fabricated information.

“We commonly see that, and its behavior that is distinctly different from somebody who’s just moved addresses,” Thoma said. “Yes, they’ve moved addresses, but a lot of the time when people move, they only move a couple of blocks down. There’s continuity in that profile, where we can still say that even though the profile has changed, it’s still fine.”

“That’s a broad example of how important it is to have that data quality,” he said. “Because if you don’t have fresh data to reference, the timeline to reference back further, you can’t say, ‘This is normal behavior for them or not.’ That’s how important it is.”

Data for the Whole Organization

The growing emphasis on identity verification is driving a widescale shift in how financial institutions approach fraud prevention. Yet opportunities remain to break down data siloes and improve visibility across systems.

“We are seeing some evolution in the ability for payments teams and fraud teams to come together quicker,” Hoffman said. “Payments teams are very focused on the transaction and what it means to bring that revenue in. There still is some hesitation for the fraud teams and the payments teams to merge together.”

“In the most advanced organizations that I work with, those two functions are working hand-in-hand,” he said. “They know exactly what’s going on from a payments perspective and how that affects the flow of fraud.”

The pace and complexity of the threat landscape demand more sophisticated infrastructure. Modern fraud prevention solutions rely on graph-based methods to map relationships between entities—sometimes referred to as fraud topology or halos.

These topology-aware systems can enhance detection accuracy while reducing costly false positives. They also enable organizations to apply the right level of friction within the customer journey, including step-up authentication when warranted.

While designed for fraud prevention, the benefits of these capabilities often extend well beyond risk teams, strengthening decision-making and operational efficiency across the entire organization.

“The data is customer data; it has huge amounts of value,” Thoma said. “You’re seeing their geolocation, behavior, age demographics—all that stuff is extremely important for the business, not just for the fraud team. Everybody thinks that’s a lot of money for fraud prevention, but it becomes very cheap because you’re splitting that into multiple budgets.”

“The marketing team can use it for targeted products, and you can increase conversions,” he said. “It doesn’t have to be fraud data, it’s company data for all divisions of that business to use.”

The post As Fraud and Agentic Risks Mount, Data Provides Continuity appeared first on PaymentsJournal.

American Express is moving to accommodate a new kind of customer—not a person, but an AI agent acting on behalf of a cardholder.

To support this shift, the company has announced plans to enable purchases made by registered AI agents. The Amex Agentic Commerce Experiences (ACE) developer kit is designed to facilitate AI-driven transactions through its closed-loop network.

ACE allows cardmembers to designate which verified AI agents can complete purchases using their cards. Only approved agents are authorized to transact, and the system uses what Amex calls Intent Intelligenceto help ensure that transactions reflect the user’s intended actions.

“American Express’ agentic AI developer tool kit strategy is a wise move into propagating usage,” said Brian Riley, Director of Credit at Javelin Strategy & Research. “It removes barriers for developers to integrate into the American Express ecosystem and simplifies how independents can approach account access. We’ve seen this cooperative strategy work well in other segments, such as business cards and rewards, and expect this will help American Express stay at the top of their game.”

The initiative builds on Amex’s existing infrastructure, which is designed to capture and manage multiple aspects of a transaction within its network.

“This will allow American Express to take advantage of their three-party network, where they own both the cardholder and merchant relationship,” Riley said.

A User-Friendly App

American Express intends to make the experience straightforward for users. In an interview with Digital Commerce 360,Luke Gebb, Executive Vice President and Global Head of Innovation, said the Amex mobile app will provide visibility into where a user’s account credentials are stored.

“Maybe you’ve asked three different agents to do different things for you and those intents are outstanding,” Gebb said. “When purchases come through, match them to the intent so you can see. If you have a problem, you can address it in there.”

Addressing Possible Errors

The company also acknowledges ongoing consumer skepticism around AI-enabled purchases.

As part of ACE, Amex plans to offer Agent Purchase Protection, intended to address disputes if an AI agent makes an error. Details of this protection model are still being developed.

The post Amex Bets on AI Agent Payments appeared first on PaymentsJournal.

Visa is introducing an on-ramp for merchants to tap into its AI-driven Intelligent Commerce platform. The new offering, Intelligent Commerce Connect is designed to let AI agents discover merchants and complete purchases on behalf of users.

The service is currently being piloted with a number of partners, including AWS, with general availability expected to occur by the end of June.

The launch is aimed at positioning Visa for the expected growth of agentic commerce, though projects vary widely and remain speculative. Some forecasts have been ambitious—McKinsey for example, has suggested that the U.S. retail market could reach as much as $1 trillion in agent-driven transactions by 2030. However, such estimates depend on significant advances in the technology and adoption. Other data points suggest a much earlier-stage market, with Precedence Research estimating global agentic commerce at under $8 billion in 2025.

Available for Any Credit Card

According to Visa, the platform includes features such as secure payment initiation, tokenization, spend controls, and authentication for agent-initiated purchases. It’s not limited to Visa transactions; it also integrates APIs from other card networks alongside Visa’s own, and can connect with multiple payment token vault providers, allowing agent platforms to work across different vendors.

When an AI agent initiates a purchase, the system identifies the appropriate payment credential, replaces the card number with a token, and routes the transaction through the relevant network. It also verifies that the agent is acting within the parameters set by the users before passing payment credentials to the merchant.

A Big Leap into Agentic AI

The release builds on Visa’s earlier push into agentic commerce, following last year’s introduction of Visa Intelligent Commerce, which enabled AI agents to use stored Visa credentials at accepting merchants.

“This is a very strategic move and a natural extension of what they’ve built with the Visa Acceptance Platform,” said Don Apgar, Director of Merchant Payments at Javelin Strategy & Research. “This effectively enables every VAP-connected processor and acquirer to offer a turnkey agentic commerce platform that’s interoperable with all of the current standards.”

“It’s important to remember that Visa’s customers in card acceptance aren’t the merchants, but rather the acquirers, processors, and PSPs,” Apgar said. “Visa didn’t bring VAP to market to compete with their customers by servicing merchants directly, but rather to enable their customers to bring new technology to their end-user merchants faster than ever.”

The post Visa Unveils Agentic Commerce Platform for Merchants appeared first on PaymentsJournal.

Stripe is betting that the next evolution of payments won’t be drive by humans, but by autonomous AI agents transacting on blockchain rails.

Blockchain has become a core component of the financial services infrastructure, underpinning everything from stablecoins to artificial intelligence models—and increasingly serving as the foundation for programmable, always-on commerce.

Stripe has been an avid investor across these areas, including the acquisition of stablecoin infrastructure firm Bridge and its expansion into agentic commerce through integrations with buy now, pay later leaders in recent years.

Last year, the payments firm unveiled Tempo, a blockchain effort launched in a collaboration with digital assets firm Paradigm. Tempo was built to facilitate high volumes of payments, and as the blockchain brings its mainnet online, there are reportedly over 100 services integrated.

Stripe highlighted Tempo’s capability to reshape cross-border payouts, payments, and remittances, and the blockchain could even play a role in embedded finance and tokenized deposits. Stablecoins are expected to serve as the workhorses facilitating many of these functions, and Stipe has highlighted its objective to bring “real payment workloads” to digital assets.

Sessions with an Agent

Stablecoins will also likely factor into the operations of the newly launched Machine Payments Protocol (MPP) an open agentic commerce standard designed to provide the infrastructure for AI agents to transact autonomously.

Stripe and OpenAI first unveiled plans for the protocol last year after partnering to bring direct payments to ChatGPT. While the protocol runs on Tempo’s blockchain, MPP was designed to integrate with other payments rails, including digital wallets and cryptocurrencies.

One of the key features of the protocol is that it supports “sessions,” where after funds and instructions are determined upfront, allowing agents to carry out multiple transactions with no further interaction.

Sorting the Shared Language

While the session capability is notable, there are still lingering doubts about consumer and business appetite for agentic commerce. Data from Coinbase’s agentic commerce protocol suggests that most transactions on the platform still consist of pilots and trials.

Still, this hasn’t stopped leading payments firms from developing their own agentic commerce protocols, including platforms from Google, Visa, Klarna, and others. Because these protocols are intended to function as a shared language for agentic commerce, the increasingly fragmented landscape could create challenges for merchants, financial institutions, and consumers seeking to develop cohesive strategies.

The post Stripe Launches Tempo Blockchain to Power AI-Driven Payments appeared first on PaymentsJournal.

In the past, banks and businesses could build rapport by delighting customers over several interactions. That window has largely disappeared amid the impersonal nature of today’s digital ecosystem—and the growing sophistication of fraud.

The surge in fraud and money laundering has prompted many experts to advocate for a return to a zero-trust framework, where every party must be verified before a transaction proceeds. That mandate will only grow more complex as agentic commerce gains traction and AI agents—and their intentions—must also be validated.

In a recent PaymentsJournal podcast, FinScan’s Chris Ostrowski, Head of Product Management, and Kieran Holland, Global Head of Solutions Engineering, along with Christopher Miller, Lead Emerging Payments Analyst at Javelin Strategy & Research, discussed how these factors have placed a premium on trust.

There are tangible ways organizations can build trust in a real-time, agentic environment. Increasingly, however, those efforts must take place long before a transaction is ever executed.

Accelerating Social Change

Many artificial intelligence enhancements have been implemented behind the scenes, from workflow optimization to cybersecurity. While customer-facing tools like chatbots have been successful, asking consumers to entrust shopping and payments to AI agents requires a far greater leap of faith.

That leap comes at a time when many consumers are experiencing a crisis of confidence. Fraud attempts have become both relentless and highly convincing—and too many individuals have fallen victim.

“I always give the example of what I would say to any member of my family who says, ‘I’ve received an e-mail offering me this deal or a massive bargain,’” Holland said. “If someone came up to you in the street and said, ‘I’m a Nigerian prince who wants to give you $5,000 if you could cash that for me,’ would you trust them?”

“There’s still that social change needed, because when something is not face-to-face, I have to have certain controls and mechanisms to make me feel confident,” he said. “Maybe that change will eventually become ingrained; maybe it just won’t. Maybe us humans need a certain amount of confidence that we used to get from face-to-face interactions.”

To rebuild confidence in a digital-first environment, organizations must establish effective risk controls around payments. That task has grown more complicated amid the rapid expansion of payment types, now spanning cards, crypto, and real-time payment rails.

This proliferation has elevated payments orchestration platforms to the forefront. These platforms not only operate across multiple payments rails, but also enable businesses to intelligently route transactions to optimize authorization rates, timing, and cost.

Such optimization is no longer just a matter of efficiency. It’s foundational to establishing trust before a transaction ever occurs. It’s also a prerequisite for agentic commerce to scale meaningfully.

“With those true agentic payments, you’re trusting that individual to act on your behalf with that vendor, potentially for the first time, or even a network of vendors,” Ostrowski said.

“You have to trust through interaction, but also within access and being able to facilitate enabling the right credentialing and set of controls within it. So you don’t have your agentic AI go out and buy you 10,000 rolls of toilet paper because it was more efficient to do it that way,” he said. “You’re having to put a lot of that trust up front.”

Given the potential volume and velocity of agent-driven transactions, trust must rest on a firm foundation. Achieving that will require broad industry alignment—a necessary, though potentially challenging, step.

“One of the interesting things here is that trust means something different for each participant in a transaction like this,” Miller said. “There is what a merchant needs to trust, there’s what an issuer needs to trust, there’s what a processor needs to trust, and there’s what consumers need to trust. There’s just a lot here to think about in terms of how we can get all the participants to agree to do the transaction.”

Driving the Next Generation of E-Commerce

This industry-wide agreement between merchants and financial services firms will be paramount because the roles and responsibilities within agentic transactions remain fluid.

“You’re setting conditions around more of an event-driven architecture,” Holland said. “When something happens on this system, then do something else for me without me having to initiate it. But who defines what the criteria for that is? Who designs the guardrails around that and who—I suppose legally and philosophically—holds the responsibility for saying, ‘I want this?’ And now the AI has translated that into a set of conditions that it’s going to use.”

“It’s the same concept in fraud prevention as in retail banking,” he said. “We don’t expect the end consumer to be the perfect guardian of their own financial health. We accept a certain level of responsibility across the injury to help them in that regard. I think the same is going to be true of agentic AI.”

Like modern payments infrastructure, agentic commerce will likely include baseline controls. However, banks will still need to implement their own safeguards, policies, and compliance frameworks to protect customers and their institutions.

Larger financial institutions may need to take the lead, gradually introducing customers to agentic commerce through limited, well-defined use cases that build familiarity and confidence over time.

“You’ll probably see something similar to the use of Zelle in the U.S. where you have banks coming together and putting those safeguards around it at a common level,” Ostrowski said. “It can drive the growth of agentic AI usage within various financial services, within payments, and within retail itself.”

“You’re also going to continue to see the growth of trust registries, where you go through verification processes to be placed on the registry to show that I have proven my ability to be trusted, and that information can follow along with the agents,” he said, “especially within the blockchain space of being able to cryptographically assign transactions and agents with certain rights. All of that can be facilitated at these larger institutions that are already learning it in other areas, to help drive this next generation of e-commerce.”

The Messaging Standard

A consortium-driven approach to agentic commerce will hinge on clear, standardized communication. Although the ISO 20022 messaging protocol was not developed specifically with agentic commerce in mind, its rich, structured data model is well suited to this paradigm.

“ISO 20022 has been designed deliberately so that much clearer information is available about what this transaction is and who’s involved,” Holland said. “Whether you need to identify the name and location of the ultimate debtor, the ultimate creditor intermediaries and so on, that new standard was designed from the ground up to do that.”

“It’s important because when you look at how AI within compliance is starting to take off, data is the foundation to that,” he said. “If you haven’t got good foundational, reliable data about who’s involved and who the counterparties are, making a good, accurate, and certainly more automated decision comes with significant risk.”

A common messaging standard becomes even more critical as transactions accelerate towards real time. For example, stablecoins and agentic commerce share significant synergy: both are real-time, highly efficient, and capable of leveraging ISO 20022’s enhanced data capabilities.

For stablecoins to integrate fully into mainstream financial systems, however, transactions must embed sufficient data to distinguish them from other cryptocurrency transfers. They must also incorporate compliance-related information, including support for travel rule requirements.

“That whole sphere comes back to the standard ISO 20022 fields and that consistency we’re starting to get to be able to go forward in these various ways,” Ostrowski said.

Making the Final Decision

More advanced communication standards, efficient infrastructure, and stronger safeguards are all critical to fostering trust in an agentic commerce ecosystem. Yet none of these solutions can replace distinctly human qualities—creativity, empathy, curiosity, and judgment.

“It’s a true saying that if you design a very fixed, very structured, automated system, us humans will always find a new scenario, a new circumstance that is all of a sudden going to break it,” Holland said. “Introducing humans into it is that creativity buffer where I can see that Chris has bought 10,000 rolls of toilet paper, I can see that it meets his preferences, but I as a human know that’s unlikely.”

“That curiosity whereby humans can still intervene and say 99.9% of the time this might be right, but with my insightfulness, with my creativity, I can introduce that human factor back into this overall very tightly structured process,” he said. “I become that level of flexibility that’s not going to break the system.”

The human element won’t disappear, because AI agents are ultimately designed to act on behalf of individuals. Preferences differ widely and evolve constantly.

An AI agent may learn a consumer’s favorite restaurants, events, or airlines. But human priorities shift. Tastes change. Context matters.

In the end, even in an agent-driven economy, trust will remain deeply human.

“Maybe that day you feel like a window seat instead of an aisle seat, and your agent would say, ‘No, that’s not your typical pattern, you normally do this,’” Ostrowski said. “There’s still that level of independence that the human wants and over time the agent will try to mimic that, but you’re still never going to completely replace that.”

“It’s similar to what we’re seeing within the regulatory environment, where regulators aren’t ready to hand off agentic decisions for risk evaluation or compliance approvals to agents entirely,” he said. “They still want to see a human reviewing the cases, making decisions on whether I should onboard or reject a type of transaction. I want to be the one approving it; I want to be making that final decision. It’s doing 90% of the work for me, but I want that last 10% to stay with me.”

The post The Fate of Agentic Commerce Hinges on an Elusive Resource: Trust appeared first on PaymentsJournal.

The capabilities of artificial intelligence have improved exponentially, and AI agents are being delegated increasingly complex tasks every day. However, data from Coinbase’s protocol suggests consumers may not be on board with agentic commerce yet.

The crypto giant launched the x402 protocol last year, designed to use the existing HTTP “402 Payment Required” status code to facilitate instant stablecoin payments. The objective was to give AI agents a mechanism to exchange digital assets as seamlessly as they exchange data.

However, as CoinDesk notes, reports on last month’s usage of the x402 platform found average daily transactions of approximately 131,000, with average daily payment volume of roughly $28,000. Notably, even on the highest-volume day, the activity appeared to consist entirely of testing and trial runs.

Not a Referendum

At first glance, the data runs counter to the prevailing narrative that agentic commerce will soon reshape the retail landscape. In reality, it likely reflects less a referendum on the technology and more a reminder that agentic commerce remains in its early stages.

That point can be easy to lose sight of as examples of agentic-driven transactions become more frequent. Mastercard, for instance, recently facilitated two transactions that expanded both the global reach of agentic commerce and the scope of ongoing pilots.

The Tech Will Have Its Day

As agentic commerce platforms began emerging last year, companies like Google and Visa developed protocols designed not only to power AI agents but also to establish guardrails ensuring those agents operate within defined parameters.

While many of these platforms reached the market last year, transaction volume has yet to materialize. This makes this year pivotal, as merchants, financial institutions, and consumers begin exploring the technology and ironing out any wrinkles.

Given the hype surrounding these platforms, their adoption will be closely scrutinized—as Coinbase has already experienced.

“The potential is still there, I just think the AI integration at scale is what’s holding back adoption.” said Joel Hugentobler, Cryptocurrency Analyst at Javelin Strategy & Research. “It’s going to take some time for most FIs to adopt agents and therefore use Coinbase’s agentic solution. This tech will have its day but might just take some time to adopt.”

The post Agentic Commerce Traffic on Coinbase’s Protocol Has Yet to Accelerate appeared first on PaymentsJournal.

Consumers can now pay with everything from digital wallets to crypto and real-time payment rails—and merchants increasingly expect to support them all. That shift is putting new pressure on merchant acquirers, the financial institutions responsible for processing payments on merchants’ behalf.

Acquirers now face fresh challenges as payment methods grow more complex. To help address these, Visa is launching its Intelligent Authorization platform for acquirers in the Asia-Pacific region.  

The platform is designed as a single-API pathway that allows acquirers to process transactions using the infrastructure of card networks. This can eliminate the need for banks to build dedicated infrastructure while also delivering operational efficiencies.

“This is an extension of what Visa has been building with their Visa Acceptance Solutions platform,” said Don Apgar, Director of Merchant Payments at Javelin Strategy & Research. “Historically, Visa has relied on acquirers to provide connectivity to the Visa network, either through their bank’s own tech stack or via a third-party processor like Fiserv or WorldPay.”

“Visa now offers merchant-level network connectivity that enables acquirers of all sizes to bring a competitive processing offering to market,” he said.

Finding the Right Rail

One reason payments orchestration platforms have proliferated is that many banks were never designed to process transactions via technologies such as mobile wallets or stablecoin platforms. However, payments orchestration platforms have the added benefit of providing intelligent payments routing.

This means these platforms can determine the optimal payment rail for a transaction based on factors like speed and cost—and dynamically switch to another option if needed. This flexibility can dramatically reduce declines while saving merchants and acquirers both time and expense.

The Absence of Parameters

When artificial intelligence is added to the mix, routing decisions can become even more dynamic. AI agents now have the capability to navigate the complexities of the modern payments landscape and make real-time routing decisions, even without predefined parameters.

As AI agents begin to play a greater role in payments orchestration, these platforms could also help accelerate the growth of agentic commerce. With many financial institutions evaluating infrastructure strategies for AI-driven transactions, streamlined payments orchestration should be a key consideration.

The post Visa Enhances Payments Orchestration Tools for Merchant Acquirers appeared first on PaymentsJournal.

Two of the most consequential trends in payments are converging as Affirm and Klarna begin working with Stripe to allow AI agents to initiate buy now, pay later purchases. The integration relies on Stripe’s Shared Payment Tokens, which enable AI-assisted shopping without exposing sensitive payment credentials.

In a blog post, Viraj Gupta, Product Manager for Cards Vault at Stripe, said the system supports both agentic network tokens and BNPL tokens within a single integration. He added that merchants offering BNPL through Stripe can see revenue gains of up to 14%, along with higher conversion rates and average order values.

A Defensive Position

Affirm and Klarna together serve nearly 150 million users worldwide. Still, the more compelling reason for partnering with Stripe may be a desire not to be sidelined as the payments ecosystem evolves.

In that sense, adoption is a defensive move as agentic commerce starts to take off. BNPL providers don’t want shoppers locked out of installment options simply because a bot, rather than a human, is making the purchase.

“You could assume that the same percentage of buyers who use BNPL today will want to use it with agentic commerce,” said Don Apgar, Director of Merchant Payments at Javelin Strategy & Research. “Secondarily, nobody can use BNPL if those payments options aren’t available in agentic commerce.”

Categories Ripe for Agentic AI

Agentic commerce remains a small slice of the market, but Apgar’s research highlights several retail categories where shopping bots could have an impact.

Commoditized goods such as paper towels or dish soap, for example, involve little brand differentiation and are often purchased primarily on price. For gift givers, an AI agent could narrow choices based on details about the recipient. And travel sites like Trivago and Expedia already aggregate and compare options. Any of these categories could be ripe for BNPL-enabled agentic purchases.

“A sample prompt could be: ‘Find and purchasethe highest-rated office chair with the best back support that costs less than $400, can be delivered within 10 days, and I can pay for in four installments with no interest,’” Apgar said. “BNPL and payment attributes can actually be a purchase requirement, along with product attributes.”

The post Stripe Brings Agentic AI to BNPL Loans appeared first on PaymentsJournal.

An AI agent recently bought a T-shirt—an unassuming purchase that nonetheless marks two milestones for agentic commerce.

First, the transaction took place in Spain, making it Europe’s first fully agentic payment. Second, it was processed by Banco Santander, the first time an agentic commerce transaction has been executed within a regulated banking environment.

The trial, facilitated by Mastercard Agent Pay platform, was designed to ensure that AI agents can complete transactions while banks and customers retain full visibility and control. By successfully operating within the stringent compliance standards of a regulated financial institution, the pilot offers an early blueprint for how agentic payments could be integrated into banking.

Increasing the Scope

Since launching Agent Pay last year, Mastercard has rapidly expanded the platform’s reach. In November, the company piloted it in UAE with Majid Al Futtaim, the retail and hospitality conglomerate, including a use case that enabled an AI agent to purchase movie tickets at a local cinema.

At a recent AI summit in India, Mastercard conducted another agentic commerce pilot with expanded scope. That trial incorporated Mastercard-issued cards from two banks, along with multiple payment processors and merchants—demonstrating how the model can function across a more complex ecosystem.

Scaling the Infrastructure

As AI agents take on a larger role in payments, issues such as security, accuracy, and privacy move to the forefront. Building a durable framework is important—one reason major technology companies are developing agentic commerce protocols that define how agents authenticate, transact, and operate within set guardrails.

The Mastercard and Banco Santander pilots show that agentic commerce can function within the current regulatory and financial infrastructure. But scaling it will still be challenging. Risks ranging from fraud and misuse to technical errors underscore the need for strong controls.

Banco Santander has indicated it will continue testing agentic commerce internally and explore additional use cases. Mastercard, meanwhile, is expected to keep expanding Agent Pay globally—potentially moving beyond consumer transactions and into commercial and enterprise payments.

The post Santander and Mastercard Pilot Agentic Commerce appeared first on PaymentsJournal.

Agentic artificial intelligence moved a step closer to real-world commerce last week when Mastercard demonstrated an AI agent in India that located a product and completed the purchase without the user opening an app or manually entering card details.

Presenting at the India AI Impact Summit 2026, Mastercard described the transaction as the country’s first fully authenticated agentic commerce payment. The demonstration offered a glimpse of a model in which software agents—using stored payment credentials—can execute purchases on a user’s behalf.

Security Is Paramount

The focus of the exercise was security. Built on Mastercard’s Agent Pay framework, the transaction took place within a controlled environment designed to verify both the user and the AI agent before payment authorization. The goal was to show that AI agent-driven commerce can meet the authentication and fraud prevention standards required for financial transactions.

The emphasis is likely to become even more important as agentic systems expand into B2B use cases, where transaction values are significantly higher and governance requirements more complex.

In comments to Business Today, Nitendra Rajput, Senior Vice President and Head of Mastercard AI Garage, said the company’s approach centers on secure automation and fraud prevention, with strict safeguards and user control built into the payment architecture.

Opportunities in the B2B Market

The broader commercial implications are even stronger. Data from Deloitte projects that  agentic AI could reach $17.5 trillion in gross merchandise value by 2030, with the majority—roughly $15 trillion—coming from B2B transactions and roughly purchases, and $2.5 trillion from B2C activity.

“Visa and Mastercard have both been very tech-forward by offering secure payment solutions to authenticate both the buyer and payment credentials in agentic commerce,” said Don Apgar, Director of Merchant Payments at Javelin Strategy & Research. “These are important first steps to building the infrastructure that will make agentic commerce viable at scale for both merchants and consumers. While it’s fun to think of shopping bots keeping our pantries stocked, the more interesting application is enterprise B2B purchasing tasks, one of the categories that our research shows will among the first to leverage the power of agentic commerce.”

The post Mastercard Advances Agentic AI Commerce in India appeared first on PaymentsJournal.

The next phase of agentic commerce may not be about browsing or checkout, it may be about control of the wallet. Coinbase is introducing agentic wallets designed to function as full-service, autonomous money management tools, enabling AI agents to make payments, trade tokens, and earn yield on investments without constant human oversight.

The objective is to add financial functionalities to any AI agent through a plug-and-play wallet solution. One of the main benefits for users is that AI agents never sleep, allowing them to constantly search for opportunities and respond in real time.

“This is a big deal because it lets AI agents hold and spend money on-chain autonomously, and it even offers spending limits and features like risk screening,” said Joel Hugentobler, Cryptocurrency Analyst at Javelin Strategy & Research. “What this really will do is increase stablecoin activity dramatically and has the ability to automate DeFi activity and other on-chain transactions.”

“This will draw a lot of volume to Coinbase’s native chain Base as well,” he said. “On the flip side, if agents get compromised, mistakes and losses can add up quickly. So adoption of this tech will depend on of they have strong enough controls for developers and institutions.”

Minimal Human Intervention

Security and accuracy have been top of mind since the emergence of the nascent agentic commerce paradigm, as AI agents are designed to make decisions with minimal human intervention. As a result, there are many ways agentic AI could disrupt the current retail landscape—especially if these systems are exploited or misused.

For its part, Coinbase has implemented guardrails within its agentic wallets to ensure agents stay on task, including programmable spending limits and session-level controls. Users will also have access to a streamlined interface to monitor their agent’s status, fund wallets, and issue new prompts.

Building the Shared Language

Developing stable infrastructure to support AI agents has become a top priority for many of the world’s leading financial players in recent months. Google, Visa, and others have launched agentic commerce protocols designed to serve as a shared language among merchants, financial services providers, consumers, and AI agents.

Coinbase has also launched its x402 protocol, which leverages the previously unused HTTP “402 Payment Required” status code to facilitate instant stablecoin payments. The company’s agentic wallet launch is an offshoot of x402, which has reportedly gained significant traction. According to Coinbase, the protocol has facilitated 50 million transactions since its launch last year.

The post Coinbase Unveils Agentic Wallets to Power Autonomous AI Spending and Investing appeared first on PaymentsJournal.

It has been less than a year since Visa and Mastercard unveiled platforms designed to give AI agents a larger role in retail—and actual purchasing power. In the months since, there has been a rush to build agentic commerce protocols, plan merchant integrations, and map out the fraud risks and potential liabilities.

Amid this push to prepare for the next big thing, many financial institutions are struggling to balance modernization efforts with compliance obligations and customer protections. As Matthew Gaughan, Payments Analyst at Javelin Strategy & Research, detailed in the Agentic Commerce Approaches: How Can Banks Prepare? report, there are concrete steps organizations can take to lay the groundwork for agentic integrations and chart a path forward.

Leveraging the Shared Language

While the payment settlement process itself will likely remain unchanged in agentic commerce model, new front-end infrastructure capable of interacting with AI agents will be required. Some initiatives are already moving in this direction, including Google’s recent rollout of its Agent Payments Protocol (AP2) platform.

AP2 is a neutral, open-source framework that enables merchants, consumers, and third-party companies to interact with agentic AI. The platform also includes built-in safeguards, known as mandates, which are designed to verify that an agent has accurately followed a user’s instructions.

While Google’s protocol has attracted a strong group of backers, several other organizations have launched competing solutions. Although none of these platforms has reached widescale adoption yet, financial institutions should begin evaluating which approaches best align with their strategic and operational needs.

“Those protocols essentially are establishing a shared language that allow payments to occur as they normally do,” Gaughan said. “The common thread through a lot of these developments is that the payment itself is handled normally on the merchant’s back end. It’s the protocol more so just makes it possible for this payment flow to occur.”

“The main takeaway is that agentic commerce goes hand-in-hand with modernization efforts in general,” he said. “Banks are going to need to be aware of these different protocols that are coming into play and that probably will require them to overhaul some of their internal systems to be more interoperable and accessible through APIs.”

Drawing Lines in the Sand

The growing number of nascent agentic commerce platforms can muddy the waters for financial leaders attempting to map out a clear strategy. Compounding this complexity are broader, unresolved questions about how agentic commerce will ultimately work.

For example, if a customer authorizes an AI agent to make a purchase and something goes awry, who’s ultimately responsible? This question becomes even more complex in scenarios involving first-party fraud, deliberate customer manipulation, or cases in which an AI agent is deceived into transacting with a fraudulent merchant.

“I’m sure banks are keenly watching this, just because in a lot of ways they’ll probably be the one—at least from a regulatory standpoint—that is ultimately on the hook,” Gaughan said. “In documentation available to developers, OpenAI made it clear that it thinks merchants own the payments associated with the transaction and that any settlement, refunds, chargebacks and compliance remain with the merchant and their payment service provider.”

“They’re all trying to draw lines in the sand, but I don’t think any of them truly know where it’s going to end,” he said. “You’re going from a standard where you might not have a card present at a transaction, but there still was always human involvement in the process.”

As generative AI adoption has expanded, the need for human oversight has become increasingly apparent. While models continue to improve, they still produce outcomes that are difficult to explain or plainly incorrect.

These uncertainties have contributed to a healthy degree of skepticism about whether agentic commerce will achieve widescale adoption.

“It’s an area that’s ripe for mistakes,” Gaughan said. “Also, there are bad actors out there optimizing fraudulent websites to look real and that are perfectly positioned and made for AI agents to interact with. Ultimately, the customer loses out on their money and they don’t get what they were buying.”

“It’s going to be an issue,” he said. “It’s going to continue to develop as the technology gets more popular—if it gets more popular. But it’s an area where the players involved are keenly aware of what’s at stake.”

A Nebulous Topic

The potential upside of the technology means organizations can’t afford to ignore it altogether. Instead, financial institutions should begin educating themselves on the emerging protocols within the broader agentic ecosystem and determine how these technologies may impact different areas of the bank.

Each protocol comes with its own nuances, and banks will likely need to support multiple platforms to meet diverse customer needs. While much of the current discussion focuses on consumer use cases, many banks also serve merchant clients with distinctly different requirements.

In parallel with infrastructure planning, banks will eventually need to address fraud risk and compliance considerations—though there is no immediate need to dive deeply into those issues.  

“It’s a very big and still kind of nebulous area, but there are some important big-picture considerations that they should be mindful of when approaching this new framework,” Gaughan said. “It’s going to be a topic of conversation for any bank or board of directors, because everybody hears it nonstop every single day.”

Getting Ahead of the Game

Agentic commerce may still be in its early stages, but its potential to reshape payments makes it more than a passing buzzword. Given the industry’s mixed track record in responding to transformative technologies, it’s imperative for FIs to begin developing strategies now.

“Banks are still going strong, but many executives would admit that if you go back 10 years ago, they were a little behind on modernizing their technology,” Gaughan said. “It’s important that they be aware of what’s going on and how they can get ahead of things and set themselves up for a potential future where agentic commerce becomes more of the norm.”

“It’s not a given that that will be the case, but it’s important that they’re doing what they can to not only facilitate these transactions, but also to preserve any of their products—be it card products or accounts—anything to stay top of wallet for consumers,” he said.

The post Demystifying the Agentic Commerce Enigma appeared first on PaymentsJournal.

The over two billion consumers within Tencent and Alibaba’s ecosystems use these platforms to send messages, shop, and pay in a unified super app. Now, many of these users will be able to leverage the platforms’ AI agents for full-fledged agentic commerce.

For example, Alibaba has just unlocked full access to its e-commerce ecosystem for its Qwen AI agent. This means that if a consumer wants to order food or buy tickets, they can prompt Qwen with their desired item, and the AI agent will handle everything—including payment through Alibaba’s affiliate Alipay.

Previously, the Qwen chatbot was limited to providing recommendations in response to user prompts. Consumers still had to visit other platforms to complete their orders.

Developing the Moat

One of Alibaba’s goals in integrating agentic commerce is to increase user engagement and build a stronger competitive moat. Currently, Tencent’s WeChat holds a commanding position in China’s thriving super app market.

However, Tencent has also indicated that it is pursuing agentic commerce initiatives, and its chatbot Yuanbao could soon see similar upgrades as Qwen. What’s more, TikTok owner ByteDance has already upgraded its AI chatbot to handle certain tasks autonomously within the e-commerce segment of its Douyin app.

The Crucible for Agentic Commerce

This agentic commerce zeitgeist was sparked by Visa and Mastercard, who rolled out iterations of the service last year. Their launches prompted many of the largest U.S. tech players to either introduce agentic commerce platforms or begin building the necessary infrastructure.

Despite these foundational efforts, little headway was made in implementing true agentic commerce last year. This suggests that 2026 could be a crucible for a technology with both dynamic upsides and compelling concerns.

One frequently cited issue with agentic commerce is that, in regions like the U.S., services are splintered across different apps and companies. Super apps, however, could present agentic capabilities to consumers at scale. This makes the super apps from Alibaba, Tencent, and ByteDance compelling proving grounds for agentic commerce.

The post Tencent and Alibaba’s Super Apps Evolve with Agentic Commerce appeared first on PaymentsJournal.

There has been considerable fanfare surrounding the emergence of agentic commerce, followed by a race to build the supporting infrastructure. While there has been less hoopla accompanying biometric authentication and digital identification cards, there are lessons from the evolution of these technologies that can be applied to agentic commerce initiatives.

As Christopher Miller, Lead Emerging Payments Analyst at Javelin Strategy & Research, notes in the 2026 Emerging Payments Trends report, one main commonality is that none of these technologies are close to achieving ubiquity. That said, this year will bring more frequent and tangible interactions with each of them for consumers, merchants, and financial institutions alike.

Face Pay for BBQ

The benefits of biometric authentication in payments are well established, including greater transaction security and reduced friction at checkout. As consumers have grown accustomed to fingerprint and facial recognition through everyday smartphone use, many have speculated that the widespread adoption of biometrics in retail payments is imminent.

While the underlying tech has existed for years and there have been notable implementations—most prominently Amazon’s pay-by-palm rollout in its physical stores—there is still a ways to go.

“In November, I published the first scorecard about biometric authentication providers at the point of sale, and the fundamental finding was that there was very little in this space that was actually in market,” Miller said. “But a number of these products planned to be available and suggested that they in fact had clients who would be launching in the U.S. in 2026.”

To date, adoption has largely been limited to pilots and trial runs within defined use cases. These range from iris-scanning programs tied to exclusive Visa cards to facial recognition systems used for venue entry and concessions at the New England Patriots’ Gillette Stadium.

This year, more of these pilots and trials are expected to move beyond experimentation and into broader, real-world deployment.

“Some people will get a chance to see it for real, and not just as a pilot somewhere in one place,” Miller said. “Last year I went to San Francisco, and I did a face pay in a pilot. But realistically, that’s one BBQ shop in one arena in one city in the United States. The point is that people will start to see this in the wild and not just in very controlled environments.”

“It’s not going to be commonplace, it’s not going to take over the world, but there will be people who do this, and that’s a step forward,” he said.

The Chicken and the Egg

As with biometrics, digital ID cards offer clear security advantages and can reduce friction in use cases such as airport queues or purchases that require age verification.

However, much like biometrics, the road to digital ID adoption has been rocky.

“The rollout of digital ID has been a classic case of uneven awareness and uneven availability,” Miller said. “Some states have had this for almost 10 years, and then other states still can’t get out of their own way. We’re reaching the point where it’s going to be more than half of states that offer it.”

“You had a chicken and the egg problem,” he said. “Why should merchants go to the trouble of building the infrastructure to accept digital IDs if nobody had digital IDs? Well, why should I get a digital ID if nobody’s going to accept it? It’s a classic problem, but the availability problem is mostly over. We can say with reasonable confidence that within a decade or so, every state is going to issue something.”

As adoption increases, merchants and financial institutions will gain greater confidence to invest further in acceptance technologies, both online and at physical points of sale.

As this infrastructure matures, innovators are likely to identify additional use cases for digital IDs. For example, financial institutions could integrate digital ID acceptance directly into customer onboarding flows. This approach could prove superior to the current paradigm, in which users take pictures of their identification documents and submit them for manual review.

“There will be mainstream visibility to this,” Miller said. “We’ll then start to see who is interested in adopting it. It’s still going to be a self-selected group of people. It’s not like we’re immediately going to go zero to 60 on digital IDs. This is where there’s more likely to be pushback—interestingly enough—than biometrics.”

“One of the reasons is that like almost all of the biometric implementations, it’s optional,” he said. “If you don’t want to do it, you don’t have to do it. The TSA does facial recognition, but if you don’t want to do it then that’s fine, we’ll do it this other way. It’s the same with Digital IDs, it’s going to grow; it’s going to be more visible.”

The First Encounter

Many of the challenges facing biometric authentication and digital ID adoption also apply to agentic commerce, where AI agents perform the lion’s share of a consumer’s shopping. While delegating purchases to AI agents offers clear benefits, many consumers may be reluctant to give them full autonomy over transactions.

What’s more, as with biometrics and digital IDs, users must first become aware of agentic commerce programs and then opt in. Early deployments will likely consist of pilots and trial runs in narrowly defined use cases, the results of which should be interpreted cautiously.

“Pilot participants are not always representative and in fact are probably anti-representative of consumers as a whole,” Miller said. “If you participated in a pilot, you’re already willing to try new things. That means that your feelings about it, your reactions to it, if you would use it again, all of those things—they’re just different than many other people’s.”

All of these factors—coupled with growing skepticism around AI’s accuracy—point to a more methodical rollout of agentic commerce than some reporting has suggested.

“Almost nothing was even remotely in production in 2025, which is an important call-out because people talked like it was happening and there were these huge growth waves,” Miller said. “No, false. That sets up 2026 to be the first encounter that many people across the entirety of agentic tech will have with the products, ranging from the consumers using them to the merchants accepting them to the payment processes interacting with them.”

“That is the trend of the year, which will be underwhelming for many people because this is where the kinks are being worked out and where the problems will in fact be surfaced,” he said. “That is a natural part of the development of emerging tools, but this is happening under a pretty bright glare.”

The post To Forecast Agentic Commerce Adoption, Look to Biometrics and Digital IDs appeared first on PaymentsJournal.

Artificial intelligence is helping shoppers find items and compare prices, but it also introduces new risks. According to Experian, the top fraud threat this year is AI agents and their potential to disrupt the evolving retail landscape.

In the past, merchants and financial institutions relied on blanket defenses to identify and neutralize any activity originating from bots. That approach is no longer sufficient as agentic commerce gains traction this year, leaving organizations struggling to distinguish malicious bot traffic from legitimate AI agents.

Experian notes that issues are heading into a crisis that will demand proactive responses, pushing organizations and regulators to examine liability obligations and the regulatory framework governing agentic commerce.

An Accelerating Threat

Even without widespread agentic activity, fraud is accelerating. Criminals can now craft highly targeted messages, and social media has become a breeding ground for scams.

These fraudulent messages are increasingly difficult to detect because cybercriminals use AI to generate realistic communications. Bad actors also leverage AI to create deepfakes, tricking employers to gain access to remote jobs or deceiving consumers into sending funds. This functionality has made deepfakes the second most impactful fraud trend of the year, per Experian.

“Consumers will always be the weakest link,” said Tracy Goldberg, Director of Cybersecurity at Javelin Strategy & Research. “Socially engineered schemes, whether driven by AI or not, will continue to fool consumers into clicking on malicious links, friending malicious actors they do not know, and giving out personally identifiable information about themselves. AI just makes the risk of socially engineered attacks more targeted and personal, which is a real worry for businesses’ customers and employees.”

“Enhanced email and firewall security while become increasingly critical to protect employees from themselves, and more businesses, financial services in particular, should consider providing ancillary security services, such as identity theft protection, to their customers that includes firewall provisions, virtual private networks, and spam filtering for text messaging and emails.”

Dynamic But Nascent

It is no coincidence that these threats leverage one of the most powerful technologies of our time: AI. Unfortunately, bad actors have been able to harness AI capabilities faster than many organizations, which are often constrained by regulatory, customer, and internal considerations.

Some organizations, however, have begun to take defensive action. Amazon has blocked third-party bots, including AI agents, from interacting with its platform. The e-commerce giant even went so far as to take legal action against AI platform Perplexity, seeking to block its AI agents from shopping autonomously on Amazon.

While this may provide a short-term solution, consumers are increasingly comfortable with using AI in retail settings—at least in certain contexts. As a result, agentic commerce is reaching a crossroads, where all stakeholders must define the roles and permissions AI agents should be granted.

These considerations could further delay the fully adoption of this dynamic yet still nascent technology.

The post Experian Raises Concerns Over Emerging Agentic Commerce Fraud appeared first on PaymentsJournal.

Despite macroeconomic concerns, the holidays are still a priority for consumers. Early data from Visa shows that U.S. holiday retail spending rose 4.2% year-over-year, excluding inflation.

The report is based on payments activity tracked from early November and excludes automotive, gas, and restaurant spending. While in-store purchases still account for nearly three-quarters of holiday spending, e-commerce emerged as the primary driver of growth.

Artificial intelligence also played a role in the online shopping boom. Visa found that roughly half of surveyed consumers said they planned to use AI tools for comparison shopping and to help narrow down gift choices.

“It’s good news that holiday sales are up over 4% and e-commerce sales largely drove that by growing almost 8%,” said Don Apgar, Director of Merchant Payments at Javelin Strategy & Research. “Certainly, AI is playing a role in this growth by making it easier to search for products with less-defined input.”

“So-called agentic search uses AI to discern meaning and intent from user search requests, rather than just matching on key words,” he said. “Because agentic search can deliver more accurate context-aware results, users are more likely to follow through with purchasing one of the suggested items.”

Picking It Up Today

It’s important to distinguish between agentic search—where AI agents help users discover products or services—and full-fledged agentic commerce.

For example, a recent agentic search partnership between PayPal and AI platform Perplexity allows users to consult an agent and complete purchases within a chat interference, but the final decision remains with the shopper.

This trend is prompting many merchants to rethink their strategies, though it doesn’t diminish the importance of the brick-and-mortar experience.

“While there’s no question that agentic search is giving e-commerce a boost, what’s notable in all of this is that physical retail still accounted for almost 75% of total retail spending this holiday season, according to Visa’s research,” Apgar said. “While agentic search can only discover results on the web, this underscores the importance of omnichannel alignment for retailers.”

“In-store inventory and buying options need to be available for agentic search engines so the consumer still gets meaningful results when they add ‘that I can pick up today’ to the end of their search query,” he said.

The Agentic Holidays

Despite these dynamic shifts, the retail sector is still far from true agentic commerce, in which AI agents independently handle most or all aspects of a transaction with minimal user involvement.

While agentic commerce has yet to gain the same level of adoption as agentic search, Visa expects that to change. The company said it already completed hundreds of AI-driven transactions through trials of its Intelligent Commerce program and anticipates that millions of consumers will rely on AI agents as personal shoppers during next year’s holiday season.

The post Holiday Shopping Soars, Driven by AI and E-Commerce appeared first on PaymentsJournal.

After the initial buzz around agentic commerce, skeptics questioned whether artificial intelligence agents would ever gain real traction in retail. Yet developments suggest that hesitation may be premature.

Most recently, Visa completed hundreds of AI-driven transactions in pilots of its Intelligent Commerce program and expects rapid consumer adoption of the technology.

The goal of agentic commerce is to take the heavy lifting out of shopping for consumers, enabling agents to perform complex purchases with minimal prompting.

Visa highlighted research showing that roughly 47% of U.S. shoppers currently use AI tools for at least one shopping task, such as product recommendations or price comparisons. From there, the payments giant extrapolated that the AI-powered e-commerce environment could drive millions of consumers to task AI agents with completing purchases by next year’s holiday season.

It remains to be seen whether agentic commerce will gain that much traction so quickly, but momentum is clearly building around the concept. Data from Javelin Strategy & Research indicates that among consumers who have yet used agentic AI tools, 40% may be willing to trust them.

“That’s evidence that they are willing,” Christopher Miller, Lead Emerging Payments Analyst at Javelin Strategy & Research told PaymentsJournal. “Will they use it over time? Will they fully change their behavior? We don’t know. But our data suggests that 40% of those people would be willing to expand their usage.”

“Sometimes when you run surveys like these, you see a substantial portion with a categorical unwillingness to use the technology for some reason, whether it’s lack of comfort or distrust of the companies providing it or whatever,” he said. “If we have at this point a high degree of willingness even among those who have not yet used, I think that suggests that there’s room to grow.”

Keeping on Task

While AI has undoubtedly permeated many aspects of the retail experience, there are still challenges to the widespread adoption of agentic commerce. One of the main concerns is ensuring that AI agents carry out instructions accurately and to the user’s satisfaction.

“Merchants, payment processors, and card issuers are all going to think about this in terms of liability and consumers are going to think about it in terms of experience,” Miller told PaymentsJournal. “If they have an experience that doesn’t meet their expectations, that has implications for the growth of this ecosystem.”

“If a consumer doesn’t believe that they’re going to get what they want by delegating authority to choose or to purchase some piece of software that we’re calling an agent right now, they might not use the agent,” he said. “That’s a fundamental limiter on growth here.”

Misinterpretation by the agent or unclear instructions from the user can increase the likelihood of transaction disputes in agentic commerce, especially during its early stages.

Safeguarding the Agents

There are also concerns about the security of agentic transactions, both from error and potential fraud. Organizations face the dual challenge of detecting fraudulent activity while minimizing false positives in this emerging ecosystem.

“We should be expecting situations where criminals are creating fake websites and apps that offer a similar service,” Suzanne Sando, Lead Fraud Management Analyst told PaymentsJournal. “They’re going to try and convince consumers to sign up for what they think is a legitimate agent service and then in turn, they will be giving up a whole host of PII and payment information and data for this particular scam.”

“On top of that, we should be expecting a surge in text and email scams from fraudsters that are impersonating legitimate agent services,” she said. “Not only do we have to worry about fake services, but now we’re worried about the use of generative AI that has already made impersonation scams easy for criminals to commit. I don’t think it’s at all far-fetched to assume that agentic commerce will be affected as well.”

To safeguard these transactions, financial services companies like Visa, Google, and Klarna have launched protocols designed to keep AI agents on task and protected from harm.

Ready to Serve

Questions have also arisen regarding whether demand for agentic commerce is substantial enough to justify the extensive investment the technology has received.

Both Visa and Mastercard have launched agentic commerce platforms that quickly expanded to additional use cases and markets. For example, Visa has worked with more than 100 partners in pilots of its Intelligent Commerce platform and plans to launch pilot programs for Intelligent Commerce in Asia and Europe next year, alongside other global initiatives.

This suggests that more AI agents will be ready to serve consumers next year, whether consumers are fully prepared or not.

“Skepticism is warranted, but this is happening,” James Wester, Co-Head of Payments at Javelin Strategy & Research, told PaymentsJournal. “If we are saying, ‘I can’t imagine why somebody would do something,’ that shows the limits of our imagination, not the limits of where this is going to go.”

“Approaching this with an open mind and understanding that there is going to be an entire industry of developers, systems integrators, and folks that are going to be aimed at this (is important),” he said. “It’s understanding that this is bigger and important, and we need to understand that in the context of our entire industry, as opposed to just saying this seems like a lot of hype.”

The post Will Agentic Commerce Break Through Next Year? appeared first on PaymentsJournal.

The launch of agentic commerce platforms like Mastercard’s Agent Pay represents a new phase in how artificial intelligence is integrated into the shopping experience. Still, agentic commerce is far from ubiquitous, and Mastercard is now moving forward with plans to pilot Agent Pay in the UAE.

These platforms aim to give users an AI agent that acts as a personal shopper. With minimal input, the agents are designed to handle the entire shopping process—from product selection to completing the transaction.

Agent Pay’s UAE trial will be conducted in partnership with Majid Al Futtaim, a corporation that owns shopping malls, hotels, and various retail stores across the region. One of the initial use cases being explored is using Agent Pay to shop for and purchase movie tickets at VOX Cinemas.

Choosing the Right Ticket

Although these platforms unlock powerful use cases, questions remain about how effectively AI agents can perform these tasks.

For example, if a user wants to purchase movie tickets, how specific must their instructions be for the AI agent to fulfill the request to their satisfaction? If customers need to provide strict guidance on the film and showtime, the AI agent’s value may be limited.

Conversely, many users may be uncomfortable giving AI full autonomy to select and purchase their evening’s entertainment. This could lead to a surge in disputes if customers are unhappy with the AI’s choices.

What’s more, there are still concerns about the security of agentic commerce transactions and the protocols needed to prevent fraud and misuse.

Not a Novelty

All of these factors contribute to the obstacles agentic commerce faces in achieving broader acceptance. While many users are open to AI-assisted shopping, they often want the final say before a payment is made.

Despite these lingering questions, many of the largest financial services players have invested heavily in the infrastructure to facilitate this new paradigm. For example, both Visa and Google have launched protocols designed to establish guardrails around AI agents.

This investment, combined with the promise of agentic AI, indicates that organizations can’t discount agentic commerce as a novelty. While there may be no need to rush adoption, companies should consider how this disruptive technology could shape their operations.

The post Mastercard Expands Agentic Commerce Efforts to UAE appeared first on PaymentsJournal.

The next iteration in the rapid evolution of artificial intelligence has arrived, and organizations are racing to harness the potential of AI agents to create a dynamic new shopping experience. However, as powerful as agentic commerce can be, the road to adoption won’t be without hiccups—many of which will lead to a surge in disputes.

In a recent PaymentsJournal podcast, Joseph McLean, CEO and Co-Founder of Quavo, and Christopher Miller, Emerging Payments Analyst at Javelin Strategy & Research, discussed the challenges that can arise in the agentic commerce dispute process, the steps financial institutions can take to prepare, and how disputes can serve as an opportunity to engage and retain customers in the age of agentic commerce.

Navigating Uncharted Waters

Traditionally, as the volume of payments has grown, the number of disputed transactions has remained relatively stable. However, as agentic commerce gains traction, this pattern is unlikely to hold.

This shift raises many questions for organizations attempting to navigate these uncharted waters.

“There is going to be fraud on these transactions; there are going to be mistakes that are made by consumers or by AI,” McLean said. “The regulations aren’t super clear on who is liable in these scenarios when consumers are making purchases. Is it the consumer? Is it the merchant? Is it the issuer? This also opens up new attack vectors for fraudsters, where they can get into the agentic commerce area themselves posing as other people and making purchases.”

In particular, there may be a rise in first-party, or consumer-engaged, fraud. For example, an AI agent might follow its instructions perfectly, yet if the customer is dissatisfied with the outcome, they may still dispute the transaction. Alternatively, a consumer could intentionally make a purchase with the plan to dispute it later—claiming fraud or an AI error.

These situations create grey areas, as liability becomes unclear when a consumer authorizes an agent but doesn’t directly complete the purchase themselves. It’s therefore critical that these issues are resolved before agentic commerce scales further, since confusion and ambiguity could be detrimental to adoption.

“Merchants, payment processors, and card issuers are all going to think about this in terms of liability and consumers are going to think about it in terms of experience,” Miller said. “If they have an experience that doesn’t meet their expectations, that has implications for the growth of this ecosystem.”

“If a consumer doesn’t believe that they’re going to get what they want by delegating authority to choose or to purchase some piece of software that we’re calling an agent right now, they might not use the agent,” he said. “That’s a fundamental limiter on growth here.”

Trusting the Process

To develop a stronger framework around the dispute process, several factors should be considered by financial institutions.

First, FIs will need a mechanism to gauge the consumer’s intent when they instructed and authorized the AI agent.

Given that AI systems can hallucinate or misinterpret instructions, it will be important to verify whether the agent accurately carried out the customer’s request. Understanding consumer intent is also critical because bad actors may attempt to manipulate AI agents—for example, by creating fraudulent websites or impersonating legitimate services to trick AI into making unauthorized transactions.

These challenges also raise broader questions about how to proactively address fraud in an agentic commerce environment.

“When it was a fake website that consumers visited, we could take that head on and teach people what are the ways to recognize a fraudulent website,” Miller said. “If it is your agent that is deceived—if one platform impersonates another within an agentic integration flow—those are entirely outside the sphere of consumer, they can’t do anything about it. It’s interesting to think about not just who is liable, but who will be perceived as having responsibility for solving that problem.”

Issuers, merchants, and agentic AI developers may all need to take on new roles in educating both consumers and AI systems. Considering the potential scope of agentic commerce, an industry consortium approach might also be required to set up comprehensive safeguards.

Regardless of the specific path forward, developing a framework for agentic commerce will likely be necessary sooner rather than later.

“A lot of consumers are using this, and we’re going to see it happen a lot more in 2026 and going forward, but consumers will need to trust what’s happening through the agent,” McLean said. “They will need to trust their merchants, and they will need to trust that their banks can handle it appropriately when something does go wrong.”

Fighting Fire with Fire

To develop this trust, financial institutions can take proactive steps to prepare for the increased volume and complexity of agentic commerce disputes. Historically, many FIs have responded to spikes in fraud or dispute cases by simply adding more personnel to the process. However, this approach is unlikely to be effective in the new paradigm.

“The best way to solve this is going to be pulling in more technology, better solutions that solve the problem end-to-end so that the users at the issuing institutions can spend more time focusing on the complex pieces of the work,” McLean said. “These disputes, they will look very similar, but it’s not going to be just more of the same. It’s going to be much higher volumes that are coming through the door and the complexity of these disputes are certainly going to be different than how they’re used to working through disputes today.”

As financial institutions take stock of the dispute process lifecycle, several important questions will arise. For instance, how will the bank handle communications with the cardholder? How will it manage accounting or reconciliation? And how will institutions handle issuing a new card if one is compromised?

These complex challenges can’t be effectively solved by adding more staff or connecting disparate systems. Doing so often creates siloes, which can lead to delays, errors, and poor experiences for both consumers and merchants.

To address these issues, a comprehensive technology solution that manages the end-to-end dispute lifecycle will be paramount.

“One of the things that we need to look at is fighting fire with fire,” McLean said. “How can we bring in AI and those sorts of technologies into the issuing space to help solve these problems, make faster decisions, augment investigations with better data and better materials to help those solutions work through faster.”

“Making sure resolution times aren’t increasing for consumers, making sure that consumers are made whole, and following all the regulations. There are so many moving parts here that the technology is going to have to solve, especially when we start talking about the first party fraud piece,” he said. “It’s another layer of complexity that we’re going to have to deal with, and an effective dispute technology solution is going to be needed by every issuer to handle this problem.”

A Moment that Matters

As financial institutions search for technology solutions, they should consider platforms that handle the full dispute lifecycle—starting from intake. Platforms like Quavo’s offer a unified data solution to receive and track information, allowing institutions to create audit trails and leverage this data within their fraud systems to fight fraud more proactively.

As disputes surge with the rise of agentic commerce, issuers will no longer need to rely on a patchwork of vendors, technologies, and in-house solutions—unlocking significant efficiency gains and potential revenue improvements.

However, one of the most powerful benefits of a streamlined dispute process is its ability to strengthen customer relationships.

“When a consumer has an issue with their accounts—and largely it’s going to be transaction-related—it can go one of two ways,” McLean said. “It can go very poorly and be a bad experience, where your customer may look to leave your institution—and all the research that we’ve conducted says that absolutely can happen.”

“On the flip side, you can take this into what we’ve always called a moment that matters,” he said. “It’s one of those pieces of banking where you can build real trust and build a much deeper relationship with your account holder.”

The post How FIs Can Prepare for the Surge in Agentic Commerce-Driven Disputes appeared first on PaymentsJournal.

Much like with generative artificial intelligence, the emergence of agentic AI has been accompanied by substantial hoopla. However, as more organizations race to incorporate the next big thing into their operations, many are struggling to plot the road map to agentic commerce.

As Christopher Miller, Lead Emerging Payments Analyst at Javelin Strategy & Research, detailed in the report Making Sense of Agentic Commerce: How Do We Get Started?, businesses and financial institutions can take tangible steps to prepare for agentic commerce. However, depending on the organization, a speedy implementation may not be the best approach.

The Vision Is Real

Agentic commerce has been defined as leveraging an AI agent to automate all aspects of a transaction, including the purchase, with minimal user interaction. However, if this model were implemented, it would create a dramatic shift in the shopping paradigm.

“Depending upon the form that automation takes, it could have substantial impacts on the ways that companies interact with individuals,” Miller said. “This means if you’re interacting with other people’s software instead of them, you have to do different things. You need to build a different type of front door, you need to build a different ecosystem to respond to the needs of agents, and you potentially have to change your advertising models.”

These are important questions, and many organizations are feeling the pressure to provide answers. This pressure has only increased as leading financial services companies like Visa and Mastercard have launched agentic commerce platforms. Additionally, Visa and Google have created protocols designed to be the framework for the future agentic environment.

These launches have prompted many organizations to wonder if, and how, they should proceed with their own agentic commerce initiatives.

“It’s important to say, ‘No, we’re not there,’” Miller said. “There is a vision; the vision is real; the potential impact is not fake, but what is going on this instant is almost nothing—and I can’t overstate that. There are no agents, that is not a thing that exists yet—in terms of a fully capable agent that can do all of this stuff for you. Zero, and not going to happen anytime soon, but there are emerging capabilities that are parts of that vision.”

An Iterative Move

The first step in getting started with agentic commerce is to identify where the technology is in its evolution and map the ways that AI agents could interact with a business’ products.

Currently, AI primarily factors into the shopping experience is by streamlining the search process and providing curated selections. In some instances, the user can even pay for the subsequent purchase within the AI platform.

“You could search in ChatGPT and say, ‘I want to see some shoes,’ and it’ll show you some shoes and then you could click a ‘buy now’ button—but that’s not an agent in anyone’s vision,” Miller said. “That is, architecturally and infrastructurally, an iterative move from a subscribe and save.”

As far as this reality is from the vision of agentic commerce, the current model has many limitations.

For example, Perplexity users can pay through PayPal directly in the AI chat, and ChatGPT users can make purchases at Etsy and Shopify in the app. These partnerships are steps in the right direction, but they also exemplify the barriers to agentic commerce.

“You could buy one thing from one place using a certain card—so there’s a long way to go,” Miller said. “But it gives companies a good way to understand things: What are the capabilities in the market? How are they emerging? What can they actually do? What are they integrated with? What are the limitations?”

Boring, Straightforward, and Infrastructural

Once organizations understand AI agents’ functionalities, they can begin to identify where the technology is headed.

“You can say, ‘I can see where the next stage of that is—we’re going to see an announcement in three months or we’re going to see an announcement in six months,’” Miller said. “Identify where the foreseeable agent capability maturity curve overlaps with those existing capabilities, and those are the places that you want to try to build to.”

This may prove difficult, as the definition of agentic commerce has already been stretched beyond the boundaries in many instances. Organizations will likely have to continue to scrutinize this technology as they map the ways the capabilities intersect with their products.

Additionally, many businesses may find that they have limited use cases for AI agents, or none at all. For those organizations moving forward with agentic commerce, one of the key factors will be to develop a shared language across any agentic capabilities within the business.

In larger organizations, many leaders and teams are likely to be tasked with implementing agentic AI. This means that it will be critical to coordinate agentic commerce projects across the organization to ensure this game-changing technology is deployed in a routine fashion.

“This sounds fancy and new and futuristic, and the reality is it’s mostly crushingly dull,” Miller said. “Automation is not interesting, in and of itself. Agents are just accelerated work, and if it was boring work to start with, it doesn’t become interesting because you automated it. This thing gets handed to people as if it is new and crazy and futuristic, but it must be delivered as something that is boring, straightforward, and infrastructural.”

The Word Agentic

For all the hype around agentic AI, in many ways it is simply a bolt-on, customer-facing layer within the shopping experience.

“The word agentic is doing a lot of work right now,” Miller said. “Agentic is being used to describe this notion of things being done for you, and it is being presented from the perspective of the end of an evolutionary stage where somehow all this stuff magically happens.

“But the software has to tap into other existing infrastructure—there’s no new payment infrastructure. There is payment infrastructure necessary to identify the agent—that bit is new and interesting and challenging—but at the end of the day, there’s a box in a warehouse that has to get on a truck and end up on my front porch, and none of that is new.”

The post How Organizations Can Chart the Course to Agentic Commerce appeared first on PaymentsJournal.

As artificial intelligence plays a growing role in purchasing decisions, Visa is launching its Trusted Agent protocol to give merchants more visibility into the process.

In the emerging agentic commerce environment, merchants will need the ability to screen AI agents and filter out bots and bad actors. Visa’s platform was developed to enable exactly that—using agent-specific cryptographic signatures and other unique identifiers.

These identifiers can convey information about an agent’s intent, such as details about the products being sought or evidence of prior interactions between the consumer and merchant. Visa’s system can also determine whether an agent has payment functionality compatible with a merchant’s preferred checkout methods.

The Continued AI Emergence

Visa underscored that this solution was necessitated by the continued rise of AI in retail transactions. The company cited data from Adobe showing that generative AI traffic increased by 4,700% year over year as of July, with most consumers who have used AI reporting that it has improved their shopping experience.

These trends have driven more companies to deploy AI across an array of use cases. For example, Klarna teamed up with Google to leverage its AI models to create personalized visuals and customized marketing campaigns within Klarna’s app.

Shepherding the Agents

Agentic commerce takes this a step further, giving AI agents the power to initiate and complete payments. This has naturally raised concerns about the safety and security of agentic transactions.

In Google’s Agent Payments Protocol (AP2)—another framework designed to shepherd AI agents—safeguards are implemented through the use of mandates. These digital contracts securely verify that an AI agent has followed a user’s instructions, including detailed data about the parameters and timing of a purchase.

Regardless of the specific protocol, security and fraud mitigation controls are necessary for agentic commerce to advance. This presents a challenging task: beyond detecting bots and fraudulent activity, these systems must also minimize false positives—while maintaining full transparency for both consumers and merchants throughout the process.

The post Visa Aims to Safeguard Agentic Commerce Transactions appeared first on PaymentsJournal.

As the race to build the infrastructure for artificial intelligence agents heats up, Klarna has announced support for Google’s open-source Agent Payments Protocol (AP2), a neutral framework designed to connect merchants, consumers, and third-party platforms to enable agent-driven commerce.

AP2 supports multiple payment types, including debit and credit cards, stablecoin transfers, and real-time payments. However, it’s not clear how Klarna’s signature buy now, pay later services will integrate with the framework.

Individualized Shopping

The announcement follows last week’s news of a broader partnership between Klarna and Google Cloud aimed at enhancing personalization across Klarna’s platform. Under the collaboration, Google’s AI models will help create tailored visuals for Klarna’s e-commerce offerings and power customized marketing campaigns.

Personalization and hyper-personalization have been effective use cases for AI, and the technology has seen widespread implementation. With the continued growth of e-commerce and digital payments, companies have more access to consumer shopping and transaction data.

While once the domain of the largest merchants such as Walmart and Amazon, AI-driven personalization tools are now being adopted by smaller merchants to better understand customer preferences and deliver individualized recommendations.

In AI Agents’ Domain

Agentic commerce goes beyond personalized recommendations, placing the entire transaction process within the domain of AI agents. It has become a high priority for many of the world’s largest financial services firms, with both Visa and Mastercard introducing their own versions of the technology.

Google’s AP2 aims to serve as the foundational infrastructure—or rails—of this emerging ecosystem, and the tech giant has made significant progress toward that goal. In addition to Klarna, Google has attracted support from companies including Mastercard, American Express, PayPal, Alipay, Coinbase, Etsy, and Intuit.

Yet despite this momentum, questions remain around the security and liability implications of agentic commerce—particularly when transactions go wrong. There’s also debate over whether consumer demand for AI-driven shopping experiences will be broad enough to justify large-scale investment in the supporting infrastructure.

Still, few doubt that AI will become an increasingly integral part of how people discover, evaluate, and complete purchases.

The post Klarna Leans Further Into Agentic Commerce appeared first on PaymentsJournal.

Artificial intelligence has become embedded in the shopping experience, and new collaborations between OpenAI and Stripe are expanding the technology’s role in payments.

First, ChatGPT’s U.S. users can now make Etsy and Shopify purchases directly on the platform. For example, once a shopper engages with the AI interface to explore a product and refine their search, they can confirm shipping details and pay using Apple Pay, Google Pay, Stripe, or a credit card—without ever leaving ChatGPT.

Second, Stripe and OpenAI are working together on an open-source Agentic Commerce Protocol (ACP). The goal is to build the infrastructure that allows merchants, consumers, and developers to integrate AI agents into the shopping journey.

Continued Advancement

These initiatives highlight the ongoing momentum behind the agentic commerce movement, which has picked up steam in recent months. Perplexity announced its Pro subscribers can now pay directly within its chat using PayPal or Venmo, a model similar to the one ChatGPT and Stripe are rolling out.

Meanwhile, Google introduced Agent Payments Protocol (AP2), an open-source protocol designed as a framework to support agentic AI. The framework accomodates multiple payment types, including debit and credit cards, stablecoin transfers, and real-time payments.

Continued Questions

As agentic commerce platforms continue to emerge, questions have risen about the role of AI and AI agents in the e-commerce landscape. The most pressing question is how these platforms are safeguarding against errors and fraud.

In Google’s AP2, for example, the tech giant uses mandates—digital contracts that verify the AI agent has followed its directives—to help protect payment data. In the ChatGPT and Stripe model, Open AI explained that all orders, payments, and fulfillment are managed directly by merchants through their existing systems, with ChatGPT acting only as an intermediary to relay data between users and merchants. 

Beyond security, there are also questions about how these platforms will leverage the data they collect and whether AI models will remain unbiased and natural in their interactions. For example, if a user consults ChatGPT about a particular item, some may wonder whether the chatbot would steer the shopper to Etsy instead of a competing marketplace.

OpenAI addressed this concern in a blog post, noting that product results are organic, not sponsored, and ranked according to relevance for the user. The company also noted it will charge merchants a nominal fee for completed purchases. 

The post OpenAI and Stripe Launch Agentic Commerce Initiatives appeared first on PaymentsJournal.

Artificial intelligence agents’ ability to shop and purchase products on behalf of consumers is set to advance with the launch of Google’s Agent Payments Protocol (AP2).

The protocol is designed as a neutral, open-source framework that enables merchants, consumers, and third-party platforms to leverage the benefits of agentic AI. It supports multiple payment types, including debit and credit cards, stablecoin transfers, and real-time payments.

While giving AI agents full autonomy to shop on users’ behalf may raise concerns, Google is introducing safeguards through the use of mandates—digital contracts that securely verify an AI agent has followed the user’s instructions.

For example, if a user asks an AI agent to buy tickets for the upcoming baseball playoffs, they would sign a mandate detailing the desired price, purchase timing, and other key conditions. The initiator would then sign a separate mandate granting the AI agent authority to complete the transaction once conditions are met.

A Seal of Approval

Although use cases for AI agents continue to emerge, the promise of Google’s protocol would mean little without industry adoption.

On that front, Google has earned a strong seal of approval. It has secured support from credit card giants like Mastercard and American Express, fintechs such as PayPal and Alipay, and crypto companies including Coinbase and MetaMask. Google has also attracted backing from Etsy, Intuit, and Salesforce.

In total, over 60 companies have backed AP2, marking a significant industry-wide collaboration. As with its recent blockchain launch, Google’s goal with AP2 is to provide an open, agnostic framework for the industry.

Building Consumer Confidence

While this backing is noteworthy, questions remain about whether customers will find value in agentic commerce. Mandates can help build consumer confidence in the process, yet AI agents have already been exploited in many cases by bad actors.

In AP2, Google has incorporated safeguards that create an auditable trail, allowing fraudulent transactions to be reviewed. Still, these guardrails may not go far enough to entice consumers to fully hand over control to AI agents.

“Interesting topic and aligns with our recent research on agentic commerce,” said Don Apgar, Director of Merchant Payments at Javelin Strategy & Research. “Several folks are quoted in this article questioning where liability falls when the agent operates outside its authorized scope: the consumer, the merchant or the card issuer? We took this a level deeper and asked, ‘How does the consumer know who the agent is truly working for? Is the agent delivering the best deal for the consumer or steering the consumer toward purchases where the agent receives a commission from the merchant?’”

“Look no further than the Google search engine where companies pay for placement to appear at the top of the search results, even though they may not be the best answer to what the user was searching for,” he said. “Kudos to Google for taking a leadership role and establishing a framework within which agents can be validated and operate securely, but there are significant business and financial questions that remain for consumers.”

The post Google’s Agentic Commerce Protocol Gets an Array of Backers appeared first on PaymentsJournal.

The movement to inject agentic artificial intelligence into the shopping experience has taken another step forward.

Mastercard has unveiled updates to its Agent Pay platform while expanding agentic commerce partnerships to drive adoption, including recent collaborations with Stripe, Google, and Ant International’s Antom.

The platform’s reach is also broadening. Mastercard noted that Citi and U.S. Bank Mastercard cardholders will get early access to Agent Pay, with availability extending to all U.S. cardholders by the holiday season. An international rollout will follow soon after.

Driving Organizational Adoption

In addition to the expansion, Mastercard announced new features aimed at driving organizational adoption of Agent Pay. These include a developer toolkit that enables integration of AI agents with Mastercard’s APIs, as well as a consulting service to help issuers, acquirers, and merchants get up to speed.

The company is also introducing insight tokens, designed to protect consumer data while delivering a more personalized experience in the agentic commerce environment. In parallel, Mastercard is working with the FIDO Alliance’s Payments Working Group to develop industry standards for this emerging technology.

The Magic of Agentic Commerce

Implementing standards and protections is key to the broader adoption of agentic commerce. Consumers may feel comfortable consulting AI platforms like ChatGPT or Perplexity during the shopping experience, but entrusting the entire process—including payment—to an AI agent will likely cause some reticence.

Despite these concerns, Craig Vosburg, Chief Services Officer at Mastercard, noted the transformative promise of agentic commerce: “Payments must be native to the agentic experience.”

While stronger infrastructure will no doubt go far towards agentic commerce adoption, another obstacle remains: agentic commerce also requires both consumer awareness and active participation.

“In the product discussion that Visa had when they did their product launch talking about agentic, one of the things that resonated with me was it was one of the products where people said, ‘We are going to have to pull consumers along,’” James Wester, Co-Head of Payments told PaymentsJournal. “‘We are going to have to show them and educate them on the magic of agentic commerce.’”

The post Mastercard Upgrades Agentic Commerce Platform Ahead of the Holidays appeared first on PaymentsJournal.

As agentic AI promises to take on the role of a personal shopper, Antom, a division of the Chinese financial giant Ant Group, has launched a new agentic payment solution for the Asia Pacific region. In a partnership with Mastercard and Visa, Antom is introducing tokenized, card-based agentic payments that would enable AI agents to complete payments with cards, bank transfers, and a wide range of digital wallets.

Antom says the technology is designed to provide a secure checkout process, an AI-ready payment mandate model, and enhanced risk management. Users can link their digital wallets directly to the checkout page without being redirected to external apps.

The platform lets payments flow naturally through conversations with AI agents, so purchases can be confirmed directly in chat. Shoppers can also arrange pre-authorized transactions such as spending limits or scheduled flash sales. 

Financial Heavyweights

The new offering involves some of the biggest players in payments. Antom offers unified, vertical-specific digital payment solutions in support of more than 300 merchant payment methods. It operates in more than 200 markets and accepts payments in more than 100 currencies. 

Mastercard’s Agent Pay and Visa’s Intelligent Commerce platforms were unveiled in April of this year. They have spearheaded the technology that enables AI agents to effectively act as personal shoppers.

Concerns About Confidence

The appeal of agentic AI is easy to see. The technology has the potential to not only find items to purchase but also merchants from which to purchase and with the optimal form of payment. But industry analysts question whether AI is advanced enough to handle these transactions and to warrant consumer confidence in its abilities.

“If you’ve used AI, the answers that it returns are just as often wrong as they are right,” Don Apgar, Director of the Merchant Payments Practice at Javelin Strategy & Research, said in a recent PaymentsJournal podcast. “When you give a shopping agent a test to go find something and buy it for me, it’s not the buying part that worries me, it’s the finding the right thing. In today’s environment, you’ve also got AI hallucinations. It’s not everybody’s worst fear, but if you get a message that says, ‘Hey, I’ve bought this for you,’ it’s like, ‘Wait, who asked for that?’”

Beyond concerns about security or accuracy, one of the main barriers to agentic commerce is the need for active customer participation. The challenge for Antom will be whether it can persuade consumers to entrust their payment data to an AI agent and let it work on their behalf.

The post Antom Hopes Asian Shoppers Are Ready for Agentic AI appeared first on PaymentsJournal.

Perplexity’s newly launched artificial intelligence web browser, Comet, was previously available only with a $200-per-month subscription or a special invitation, but now PayPal and Venmo users will gain early access.

Much like ChatGPT, Perplexity created a chatbot platform that users can query to receive detailed answers along with source links. Comet incorporates this functionality into a web browser that also includes an AI assistant. According to Ryan Foutty, VP of Business at Perplexity, the AI agent is akin to a “personal shopper and personal assistant all in one.”

The two firms are leveraging an existing partnership, as PayPal recently gave Perplexity users the capability to purchase products and services directly within the platform. This means a user could chat with a Perplexity AI agent about a theme for an upcoming birthday party, then make purchases without leaving the platform.

The Agent Is Everywhere

The PayPal release stopped short of specifying whether this same functionality would exist within Comet, but the agentic commerce model is gaining significant traction. Visa and Mastercard have launched platforms through which AI agents are designed to be virtual personal shoppers.

However, Mastercard’s Agent Pay and Visa’s Intelligent Commerce platforms have the autonomy to make payments.

“In this vision, the agent is everywhere,” Christopher Miller, Emerging Payments Analyst at Javelin Strategy & Research, told PaymentsJournal. “It’s making your life easier; it’s saving you time; it’s relieving you of the burdens of your side of any transaction. It can find items for you to purchase, it can choose merchants for you to purchase from, and it can select which form of payment you wish to use at any given point in time. There’s a lot behind that vision, and many technical aspects will have to be addressed for a system like that to operate.”

Market and Regulatory Risks

These technical aspects are likely a reason Perplexity and PayPal haven’t unlocked full-blown agentic commerce yet.

However, the partnership could still provide substantial benefits for both companies. In addition to Comet access, PayPal and Venmo users will also receive a free yearlong subscription to Perplexity’s premium tier, Perplexity Pro. The fintech aims to leverage these offers to drive users to its new subscription hub, where they can view, manage, and pay their subscriptions in one place.

These offers will likely draw more users to PayPal’s platform, a feat Perplexity also hopes to achieve, as the AI firm trails ChatGPT and Google’s Gemini significantly. Similarly, the Comet launch pits Perplexity against Google Chrome in what is likely another uphill battle. So far, those odds haven’t deterred Perplexity. The company recently made an ambitious, unsolicited offer to buy Chrome for $34.5 billion, an offer far lower than the leading browser’s valuation.

The post PayPal and Venmo Users Can Access Perplexity’s Anticipated AI Browser appeared first on PaymentsJournal.

In the latest convergence of artificial intelligence and payments, OpenAI will integrate a payments checkout system into ChatGPT.

Earlier this year, ChatGPT and Shopify partnered to upgrade the shopping feature within the AI platform. The collaboration enabled users who search for a product on ChatGPT to see the top results with prices, reviews, and links to relevant sites. However, to buy the product, users were still directed to the merchant’s platform.

According to Reuters, consumers will soon be able to complete their purchases directly on ChatGPT. Open AI is working with Shopify and other brands to develop the system and negotiate rates, as merchants fulfilling orders through ChatGPT would pay a commission to OpenAI.

Growing Payments Integrations

This integration reflects a growing trend of AI chatbots and agents being empowered to perform transactions. Perplexity recently announced that its Perplexity Pro subscribers would be able to make payments directly within its AI interface.

This functionality is enabled by PayPal, with both PayPal and Venmo payment methods supported. The goal is to give Perplexity users the ability to make one-click payments once they’ve selected their preferred product through the AI platform.

The Rush Toward Agentic Commerce

Taking this model a step further, both Visa and Mastercard have rolled out agentic commerce platforms designed to make AI agents into full-scale shopping assistants. Mastercard’s Agent Pay and Visa’s Intelligent Commerce platforms are built to handle every aspect of a transaction—including payment—with little customer interaction.

However, giving AI this level of control has raised concerns, particularly around the technology’s potential for inaccuracies and hallucinations. These risks are somewhat mitigated in the Perplexity and ChatGPT models, where the final payment decision still rests with the user.

Nonetheless, privacy and security concerns remain across all these scenarios, as bad actors could exploit these still-nascent AI models in various ways. Still, for all the valid concerns, the rush toward agentic commerce doesn’t appear likely to lose momentum.

“Skepticism is warranted, but this is happening,” James Wester, Co-Head of Payments at Javelin Strategy & Research told PaymentsJournal. “If we are saying, ‘I can’t imagine why somebody would do something,’ that shows the limits of our imagination, not the limits of where this is going to go.”

“Approaching this with an open mind and understanding that there is going to be an entire industry of developers, systems integrators, and folks that are going to be aimed at this (is important),” he said. “It’s understanding that this is bigger and important, and we need to understand that in the context of our entire industry, as opposed to just saying this seems like a lot of hype.”

The post OpenAI to Add Payments Checkout in ChatGPT appeared first on PaymentsJournal.

Walmart has Sparky; Amazon has Rufus. These AI-powered shopping assistants have begun to take a more prominent place in the e-commerce apps of the world’s largest retailers. Although chatbots have been an early use case for artificial intelligence, they are just the beginning of how merchants can leverage this powerful technology.

As Don Apgar, Director of Merchant Payments at Javelin Strategy & Research, found in the report AI in the Payments Ecosystem merchant use cases for artificial intelligence  cover such areas as transaction routing and regulatory compliance. However, merchants also must consider risks as they race to implement AI.

Fraud, Compliance, and AML

Along with customer service, one of the most frequent implementations of AI has been in fraud detection. Artificial intelligence can dig through vast amounts of data and identify patterns and red flags. This capability is especially applicable in card-not-present environments like e-commerce.

Although AI excels at parsing data, fine-tuning can be done in how models analyze their findings and present conclusions. If AI has too much autonomy in fraud response, unintended consequences can occur.

“Sometimes a decision is very obvious, but in cases where it’s not, if you’re not restrictive enough, you’re going to take a fraudulent transaction,” Apgar said. “If you’re overly restrictive, you’re going to alienate a good customer who was trying to make a legitimate purchase.”

Despite these issues, artificial intelligence has the potential to supercharge the fraud defenses of not only merchants but also the payment processors that serve them.

Another area where AI can make an impact at the processor level is in compliance. Payment processors have been increasingly held responsible for anti-money-laundering (AML) monitoring.

In this use case, AI can ensure that processors are compliant by verifying that a merchant account is legitimate. Artificial intelligence can scour the internet and provide troves of data that help processors vet their customers.

“AML is a little trickier because of the amount of data,” Apgar said. “A lot of banks and processors are having trouble with this because just simply the volumes of data that have to be analyzed to be able to detect these patterns. In today’s compliance environment, whether or not those rules continue to be enforced as vigorously as they were in the previous administration is unclear, but that doesn’t mean that AI won’t have a role in that going forward.”

Routing the Transaction

Another operational area where AI will play a larger role is transaction routing. As more payment types have become available, organizations have increasingly explored payment orchestration efforts. Selecting the most efficient payment method can dramatically cut costs and improve the customer experience.

However, determining the right path for sending a payment can be complex, especially when cross-border elements come into play.

Today, many of these platforms are rules-based, whereby the user will program rules to define the process. Some degree of adaptive learning and machine learning still comes into play, but adaptive learning is limited because it can handle only cases that it has seen before. The model understands that when a certain event occurs, a certain result was obtained.

As more variables are introduced, adaptive learning is likely to struggle.

“Machine learning is based on experience with transactions that share similar attributes, but the first time that transaction comes in the door and a transaction with those attributes has never been seen before, how do you make that decision?” Apgar said. “That’s where AI comes in. AI is able to handle broader amounts of data beyond the task at hand, which is how do I route this transaction?”

Pushing the Envelope

Though artificial intelligence can provide efficiency gains throughout an organization, the promise of AI means that it will continue to be implemented in customer-facing situations.

“If you think of AI like a search engine on steroids, it’s extremely useful,” Apgar said. “It creates a lot of efficiencies—especially for merchants—where customers come to the site and say, ‘Hey, I need help finding this; I have a question about that.’ It can bring them right to the point in the FAQ, and some small percentage of inquiries still go to a live operator.”

Although AI has been successful in many chat use cases, some organizations will want to push the envelope.

In fact, some of the world’s most dominant financial companies have already given the technology a larger role. Visa and Mastercard have rolled out platforms built to harness agentic AI. In this model, AI agents can shop and make purchases with little customer interaction.

To some consumers, it would be a substantial boon to simply give AI a general direction—find a 25^th anniversary gift for my wife, for example—and have an agent do all the legwork and make the purchase. However, many customers would be hesitant to give AI the reins due to the tech’s potential to make a mistake, spend too much, or disclose private data to the wrong party.

For these reasons, merchants still must maintain a buffer around any public-facing AI initiatives.

“You never want AI right now to be in the critical path of anything, because AI is found to make mistakes,” Apgar said. “It hallucinates, as they say—it makes up stuff that’s not there. You want to be able to leverage the efficiencies of AI, but you never want it to create a point of failure in a workflow.

“It’s easy to fall into that trap where, as in the chatbot example, ‘AI is handling 80% of the inquiries—what if we just didn’t have staff?’ True, but you’re never going to get to 100%, at least not in today’s technology. At some point in the future, you will, but not now. So you always want to have that backstop, and it’s the same thing if you look at the operational side.”

The post Merchants Find More Use Cases for AI Amid Risks appeared first on PaymentsJournal.

Although generative artificial intelligence has powerful proficiencies, its limitations have become more apparent as the technology is deployed at scale. Now, however, Visa and Mastercard have unveiled platforms aimed at giving AI an even more prominent role—AI agents will soon be able to shop and make purchases with minimal consumer input.

In a recent PaymentsJournal podcast, Christopher Miller, Lead Emerging Payments Analyst, Suzanne Sando, Lead Fraud Management Analyst, Don Apgar, Director of Merchant Payments, Jordan Hirschfield, Director of Prepaid, and James Wester, Co-Head of Payments at Javelin Strategy & Research, discussed the barriers to agentic commerce, potential use cases for the emerging technology, and the broader future of AI.

A Vision of the Future

In both Mastercard’s Agent Pay and Visa’s Intelligent Commerce platforms, AI agents are effectively taking on the role of personal shopper.

“In this vision, the agent is everywhere,” Miller said. “It’s making your life easier; it’s saving you time; it’s relieving you of the burdens of your side of any transaction. It can find items for you to purchase, it can choose merchants for you to purchase from, and it can select which form of payment you wish to use at any given point in time. There’s a lot behind that vision, and many technical aspects will have to be addressed for a system like that to operate.”

Beyond concerns about security or accuracy, one of the main barriers to agentic commerce is the need for active customer participation. Until now, organizations have largely implemented AI initiatives behind the scenes.  

Persuading consumers to entrust their payment data to an AI agent will likely require a deeper understanding of AI than many users currently have.

“In the product discussion that Visa had when they did their product launch talking about agentic, one of the things that resonated with me was it was one of the products where people said, ‘We are going to have to pull consumers along,’” Wester said. “’We are going to have to show them and educate them on the magic of agentic commerce.’”

In addition to the awareness obstacle, there is growing skepticism about the accuracy of AI-produced information. While many concerns about artificial intelligence are unfounded, there are documented limitations that could cause consumers to be wary of agentic commerce.

“If you’ve used AI, the answers that it returns are just as often wrong as they are right,” Apgar said. “When you give a shopping agent a test to go find something and buy it for me, it’s not the buying part that worries me, it’s the finding the right thing. In today’s environment, you’ve also got AI hallucinations. It’s not everybody’s worst fear, but if you get a message that says, ‘Hey, I’ve bought this for you’ it’s like, ‘Wait, who asked for that?’”

The Potential for Exploitation

Along with concerns that an AI agent could make a costly mistake, it is inevitable that bad actors will attempt to exploit or manipulate these systems.

On one hand, fraud prevention has been a standout use case for AI, as it can detect suspicious activity across vast amounts of data. On the other hand, the emergence of a new technology connecting consumers, merchants, and financial institutions introduces risks that must be addressed.

“There’s obviously a major need for continuous authentication throughout the entire life cycle of the agent interaction,” Sando said. “At this point, we’re thinking data points, what behavioral analytics can I use to continuously authenticate the consumer to the agent? This is just like any other identity verification and continuous auth scenario.”

“It doesn’t matter what the channel is, whether it’s your digital wallet or you’re going straight to a merchant or you’re using the actual agent, even down to when you’re planning for situations where social engineering is suspected,” she said. “The best scam detection solutions are looking for those real-time clues where social engineering is taking place where a criminal is convincing a consumer to make a certain purchase.”

Many organizations use behavioral analytics to identify such clues—for example, when a user makes a purchase that is out of character or from a significantly different location.

However, if it’s unclear to a retailer whether a purchase is made by an AI agent or a human, it can be difficult for the business to distinguish fraudulent activity. In this scenario, merchants must rely on the agent service to have conducted proper due diligence in detecting illegitimate or unauthorized behavior before initiating the purchase.

Additionally, the services themselves present a potential avenue for cybercriminals exploitation.

“We should be expecting situations where criminals are creating fake websites and apps that offer a similar service,” Sando said. “They’re going to try and convince consumers to sign up for what they think is a legitimate agent service and then in turn, they will be giving up a whole host of PII and payment information and data for this particular scam.”

“On top of that, we should be expecting a surge in text and email scams from fraudsters that are impersonating legitimate agent services,” she said. “Not only do we have to worry about fake services, but now we’re worried about the use of generative AI that has already made impersonation scams easy for criminals to commit. I don’t think it’s at all far-fetched to assume that agentic commerce will be affected as well.”

The Training Wheels

While significant concerns surround agentic commerce, there are also powerful benefits. Many consumers struggle to fully leverage loyalty points, gift cards, and other rewards. In these split-tender scenarios, an AI agent could dramatically impact the customer experience.

“If you can split-tender, I can get the most value for my money by saying it’s going to understand how it can redeem reward points in a stored value account,” Hirschfield said. “How can I benefit if it’s earning more reward points through a particular card that it’s going to use? How does it pick and choose how to split up a payment to buy something that most benefits the consumer, so they get the most value for their money?”

Loyalty and rewards have become integral to prepaid accounts, which may also serve as the bridge connecting consumers to agentic commerce.

“Prepaid can be the training wheels,” Hirschfield said. “Obviously, you need to train the models and you need to train the consumer. With prepaid, consumers are already used to turning money over to prepaid accounts and stored value accounts.”

“Part of the issue is how do I trust a third party—be it the retailer themselves with a stored value account or an agent—with my money,” he said. “If I limit how much I give them and I’m giving them a pot of money, this is a more direct way to program that money.”

Agentic AI could also offer potential benefits for merchants. If a retailer can recognize that an AI agent is making a purchase, it may open the door to a new dynamic in commerce.

“Can the merchant now offer agent incentives to buy at their store versus somebody else’s store?” Apgar said. “If Target and Walmart have the same item and Target is now paying a sales performance incentive fund to the agent to make that purchase at Target versus making it at Walmart, now the merchant has the ability to apply leverage to the automated shopping process.”

The Limits of Imagination

While the impacts on consumers and merchants are yet to be determined, the promise of agentic AI suggests that an entire industry may emerge to power this new paradigm.

“At the end of the day, to the extent that this vision comes to pass, it is the providers of the agents themselves who are most likely to insert themselves as a new layer of the commerce stack and extract some form of value,” Miller said. “We’ve seen this play out repeatedly, so I’m going to suspect that’s who will benefit the most. Whether other participants benefit or not, I think is a TBD.”

Although many questions remain about the trajectory of agentic commerce, there is little doubt that it is gaining momentum. The launches of Visa and Mastercard’s platforms are imminent, and they are likely just the first of many agentic AI-powered initiatives.

“Skepticism is warranted, but this is happening,” Wester said. “If we are saying, ‘I can’t imagine why somebody would do something,’ that shows the limits of our imagination, not the limits of where this is going to go.”

“Approaching this with an open mind and understanding that there is going to be an entire industry of developers, systems integrators, and folks that are going to be aimed at this (is important),” he said. “It’s understanding that this is bigger and important, and we need to understand that in the context of our entire industry, as opposed to just saying this seems like a lot of hype.”

The post The Agentic Advent: How the Next Iteration of AI is Shaping Commerce appeared first on PaymentsJournal.