Imagine planning an international trip. Rather than searching for flights, comparing accommodations, and completing multiple checkouts, an AI agent receives the consumer’s preferences (including dates, budget, loyalty programs, and preferred payment methods) and manages the process end-to-end.

Artificial intelligence has already reshaped how people interact with the world, and that, of course, includes how consumers discover products, compare prices, and receive recommendations. A recent McKinsey study found that 10% of consumers already use AI to start their online shopping journey. The next shift is set to move beyond this initial assistance, with 20% of respondents saying they would be comfortable asking AI to make a purchase on their behalf, according to McKinsey.

A new generation of so-called “agentic AI” systems is beginning to act on behalf of users, not only informing decisions, but initiating and completing transactions independently.

“The irony is profound. After spending decades building systems to shield away robots, we now have to develop tools that help us trust the self-acting entities we spent so long trying to keep out,” highlights Eduardo de Abreu, Chief Product Officer at EBANX.

Industry leaders compare the emergence of agentic AI to the early days of e-commerce, when online storefronts first began to replace physical ones, on a trajectory that is gradually becoming difficult to ignore.

“Agentic commerce is a paradigm shift to how e-commerce as we have known it during this century will be for the next few years,” states Rene Salazar, Head of Partnerships and Expansion LatAm at Stripe. 

Projections from Deloitte indicate that by 2030, up to 30% of e-commerce transaction value could be influenced by agentic AI, with automated systems potentially driving as much as USD 17.5 trillion in global commerce.

Traditionally, Salazar notes, businesses have focused on optimizing their appeal to consumers through search engine optimization, targeted keywords, and marketing. “Now, agentic commerce really means the creation of a third party, an agent, that mediates transactions between a merchant and a consumer.” 

“Ultimately, the promise here is that you remove the emotion and the impulse, and consumers would potentially get better pricing”, believes Daniel Kornitzer, EBANX's VP of Global Partnerships. 

“Agent-driven commerce has massive potential for value creation and for disruption. It would force merchants to up their game in terms of pricing, resulting in better value for consumers,” he adds.

Amidst this rise of agentic commerce, another shift is simultaneously beginning to take shape in how consumers search for products. Clients will gradually move away from search-led purchasing on platforms like Google and marketplaces such as Amazon and toward conversational buying through AI tools such as ChatGPT and Gemini, predicts Javier Kaniewicz, Country Growth Senior Manager for South LatAm at EBANX.

“If e-commerce platforms don’t allow agents to operate within their ecosystems, they risk losing the sale", he says. “There may be regulations around how agents are allowed to enter and interact, but this shift is inevitable. All marketplaces will eventually have to open their doors to agents." 

Salazar, from Stripe, adds that, through their agents, consumers will be able to access businesses that either they didn’t know existed, or that were “buried in pages and pages of a search engine”.  

“What will drive the consumer’s preference won’t be a fancy website, with impressive UX, but the quality of the prompt to the agent and the detail to which it knows the consumer”, he says, adding that, with agents, “the purchase is completed without visiting the merchant’s website”. 

For businesses, it provides an opportunity to compete head-to-head with every single retailer. “Agents open a possibility of bringing millions of consumers to a storefront that is no longer their website”, completes Salazar.

The rise of agentic commerce poses a particularly consequential question for emerging markets. These economies combine rapid digital adoption, fast-evolving payment infrastructures, and high consumer openness to new technologies, while also operating under some of the world’s most complex fraud, trust, and identity challenges. 

As a result, rising economies are poised to become both early adopters of agentic commerce and critical stress testers for its scalability.

At first glance, this may seem counterintuitive, since one might expect the most established financial systems to lead the transition toward AI-driven shopping and payments. 

But according to Eduardo de Abreu, Chief Product Officer at EBANX, the opposite is true: “Emerging markets are becoming the laboratories where the future of AI-driven payments is being invented”.

The absence of deeply entrenched legacy systems has historically created room for experimentation in many emerging economies. 

“These markets are much more open to new technologies because there are fewer incumbents and fewer dominant rails,” Eduardo de Abreu explains. “In the US or Europe, however, payments are largely built around cards and long-established bank transfer systems. In more open environments, new technologies can gain traction much faster.”

That openness has already translated into mobile-first payment infrastructures and real-time settlement systems operating at a national scale. Platforms such as Pix in Brazil, UPI in India, and mobile money networks across Africa have enabled account-to-account payments that rival — and in some cases surpass — traditional card rails in adoption.

These systems, being digital-native and API-first, can be easier for AI agents to integrate with than legacy card and bank rails.

Real-time APMs and mobile money already support instant settlement and programmable flows, potentially allowing agents to initiate and complete payments autonomously. By contrast, legacy payment stacks were built for human-led processes and batch approvals, requiring heavier retrofitting to support AI-driven execution.

In that sense, Bruno Rossetto, Software Engineering Director at EBANX, notes that emerging-market fintechs and APM providers may be able to move more quickly than traditional card-based systems in mature economies.

“Card networks may be more cautious because of fraud exposure. But APMs and fintechs in emerging markets tend to be more flexible and more willing to experiment", says Rossetto.

Having long operated in environments with elevated fraud attacks and cybercrime risks,  emerging markets were forced to design payment systems that assume risk by default rather than treating it as an exception. This has direct implications for agentic commerce, where trust in automated decision-making is the primary barrier to adoption.

“Instead of layering controls on top of outdated frameworks, these systems were often built from scratch,” adds Eduardo de Abreu. “That ‘baptism by fire’ created a level of technical resilience and operational sophistication that is highly relevant for the AI era.” In practice, this means that identity verification, transaction limits, and real-time monitoring are embedded into the payment flow.

One example of this is that most alternative payment methods born in emerging markets require two-step authentication, whether with an OTP (one-time password) or biometric authentication. In this sense, Rossetto from EBANX believes that early agentic commerce models in emerging markets could initially rely on alternative payment methods that require explicit user confirmation. 

“APMs with in-app authentication, like Pix, may offer a safer starting point,” he says. In this setup, an agent can initiate a purchase, but the final confirmation still happens on the user’s device, reducing the need to share sensitive credentials.

Still, those same characteristics that make these markets fertile testing grounds also expose their limits. Elevated fraud levels, persistent trust gaps, and uneven financial and digital literacy complicate the delegation of purchasing power to autonomous agents. 

According to Eduardo de Abreu, Chief Product Officer at EBANX, these regions already operate under elevated fraud pressure, a reality that agentic commerce could amplify.

“Emerging markets tend to have higher fraud rates and an extraordinary level of creativity when it comes to new scams,” he explains. “That’s already a challenge for card and alternative payment methods transactions. With agentic commerce, this becomes a much bigger risk multiplier.”

Fraudsters, meanwhile, are not waiting for agentic commerce to mature. As noted by André Peixoto, Operations Director at EBANX, generative AI is already being weaponized at scale: from fully AI-generated fake storefronts and phishing campaigns to voice scams and synthetic identities engineered to bypass onboarding and risk assessments.

“The real challenge is matching the speed and sophistication of attackers as AI is already being used offensively. Defensive systems need to evolve at the same pace, or even faster,” argues Peixoto.

Inclusion and education also remain critical variables. In Brazil, for example, smartphone penetration is near universal, but consumer confidence in online payments still lags. “There is a gap between access and trust,” says Leandro Carmo, Country Growth Director for Brazil at EBANX. “People are connected, but not always confident in digital transactions.”

And, “although financial inclusion has advanced significantly, financial education still has several steps to go,” adds Sebastian Fantini, Product Director at EBANX.

Together, these dynamics help explain why emerging markets play a dual role in the evolution of agentic commerce.

Their mobile-first infrastructures, real-time payment rails, and high adoption of AI make them ideal environments for experimentation. At the same time, their exposure to fraud, trust deficits, and infrastructural constraints ensures that any agentic model that succeeds at scale will have been rigorously tested, and not just in theory, but in practice.

Another unresolved challenge lies in the structural bias of early agentic commerce protocols. As Eduardo de Abreu notes, much of the current discussion and emerging standards are still deeply rooted in the card ecosystem.

“Most protocols are being designed with cards in mind,” he says. “That’s not the reality of emerging markets. A big part of our role is to bring this perspective to the table early, so these systems don’t have to be retrofitted later.”

This mismatch matters, since in markets where real-time payments, A2A transfers, wallets, and other local APMs dominate, agentic commerce must be payment-agnostic to scale. Otherwise, it risks reinforcing exclusion rather than innovation.

While there is still no clear timeline for when agentic commerce will reach scale, a growing set of announcements, pilots, and protocols signal where the market is heading, and how agentic payments are taking shape.

Tech players and payment networks, such as OpenAI, Google, Microsoft, Stripe, and Mastercard, are racing to stake early positions, laying the infrastructure to make them viable before launching fully autonomous shopping experiences.

Collectively, these early initiatives suggest that agentic commerce will not arrive as a single breakthrough moment. Instead, it is being assembled piece by piece through protocols, standards, and tightly scoped pilots, with trust, accountability, and interoperability as prerequisites rather than afterthoughts. 

Whether the industry can align quickly enough around these foundations will determine how fast agents move from experimental assistants to active participants in everyday commerce.

“2026 will likely be a year of experimentation,” predicts Rossetto, from EBANX, estimating that broader adoption may only materialize closer to 2027.

Stripe has taken one of the most concrete steps to date in enabling transactions within AI interfaces. In September 2025, the company announced an instant checkout natively embedded in OpenAI’s ChatGPT. The solution is built on the Agentic Commerce Protocol, an open standard designed to support end-to-end AI-mediated transactions. In subsequent months, Stripe confirmed that the same protocol would also power checkout functionality within Microsoft Copilot integrations.

“We look forward to thousands of consumers, agents, and businesses adopting it in the future,” says Rene Salazar, Head of Partnerships and Expansion Latam at Stripe. 

ChatGPT's agent-led purchasing is initially available only to US consumers. Its launch marked the first time a large-scale AI interface had moved beyond discovery into native checkout, enabling purchases to be completed without leaving the conversational environment. 

The Stripe initiative is also notable not only for its reach, given that ChatGPT and Copilot count hundreds of millions of weekly users, but also for its architectural choices. 

The protocol is open-sourced, payment-agnostic, and designed to let merchants remain the merchant of record, using their existing systems for payments, fulfillment, and customer support. AI platforms act as a delegated shopping agent, securely passing intent, cart details, and payment instructions between the user and the merchant.

For Stripe merchants, enabling agentic payments can require minimal technical changes, lowering barriers to experimentation.

Security and trust, however, remain the industry’s central preoccupation. According to Eduardo de Abreu, Chief Product Officer at EBANX, the most meaningful recent advances have focused precisely on this layer, with particular attention on initiatives led by Google.

“Google’s approach, in particular, is very security-oriented. That’s where everyone is looking,” he says.

That assessment was reinforced in September, when Google announced the Agent Payments Protocol (AP2). Developed in collaboration with more than 60 partners across payments, fintech, and technology (including EBANX), AP2 establishes a shared framework to authenticate AI agents, validate user mandates, and define accountability when transactions go wrong.

At its core are cryptographically signed “mandates” that formally document user intent, whether for real-time purchases requiring human approval or fully delegated tasks executed autonomously.

In practice, AP2 creates a verifiable chain from intent to cart to payment, solving one of agentic commerce’s most challenging problems: proving that an agent is acting legitimately with explicit authorization.

Stripe and Google’s initiatives represent two complementary approaches to agentic commerce. Stripe, as a payments infrastructure provider, prioritizes open, merchant-driven adoption and rapid transactional enablement. Google, as an interface, identity, and OS actor, centers its approach on security and user authorization, embedding mandate verification and accountability at the protocol level.

Beyond individual companies, other clear signs that agentic commerce is edging toward the mainstream are emerging from the card industry itself.

Recently, EMVCo — a global technical body for card-based payments, owned by Visa, Mastercard, American Express, Discover, JCB, and UnionPay — quietly announced that it is working on how global EMV specifications can support agentic payments. By exploring extensions to 3DS, payment tokenization, and Secure Remote Commerce, EMVCo is addressing liability, authentication, and interoperability head-on. 

If standardized successfully, this could make agent-led payments compatible with existing card rails worldwide, dramatically lowering friction for adoption.

PayPal and Mastercard have also announced initiatives to enable AI agents to transact within PayPal wallets, while fintechs and compliance providers are proposing new frameworks, such as “Know Your Agent” (KYA). 

These initiatives combine agent identity verification, dynamic user consent, fraud data sharing, and step-up authentication, along with dashboards that enable consumers to monitor, limit, or revoke agent permissions at any time. The goal is clear: pair automation with guardrails that preserve human control.

As agentic commerce unfolds, shaped by local trust dynamics, consumer behavior, and regulatory maturity, the same fundamental questions continue to resurface across regions.

According to Daniel Kornitzer, EBANX's VP of Global Partnerships, the essential factor for driving the adoption of agentic commerce will be "solving the trust and technical challenges to ensure the human intent that the agent is attempting to fulfill."

Security emerges as an immediate and universal concern. Agents themselves may need distinct payment credentials so that e-commerce merchants can recognize them as legitimate rather than fraudulent.

“If the same card suddenly appears across multiple users or contexts, merchants might block it, assuming fraud,” explains Wiza Jalakasi, Sub-Saharan Africa Regional Director at EBANX. “Acceptance systems need to evolve to recognize and accept non-human tokens.”

One likely solution is the rise of trusted verifiers within the payments ecosystem. Card networks or banks could certify agents and act as guarantors, allowing merchants to query, via their PSPs, whether an agent is authorized and insured.

On this same page, several experts converge on the idea that agentic commerce requires a conceptual shift in compliance and risk frameworks. Sebastian Fantini, Product Director for EBANX, describes this transition succinctly:

“The industry will need to move from ‘know your customer’ to ‘know your agent.’ It has to be clear when a transaction is coming from a human versus when it’s being initiated or decided by an agent.”

This distinction has far-reaching implications, since merchants today rely heavily on behavioral signals that assume a human is behind the screen. As Leandro Carmo, Country Growth Director for Brazil at EBANX, points out, that assumption may no longer hold.

“Today, merchants can tell when they’re interacting with a human. That may disappear,” he says. “How do you evaluate fraud when a machine is making decisions? How do you know if a purchase is legitimate, authorized, and aligned with the consumer’s intent?”

Agentic commerce also challenges long-standing assumptions around disputes, refunds, and chargebacks. If an agent makes a mistake or even acts ambiguously, should liability be treated the same way as when a human clicks “buy”?

“Agents are more prone to errors,” Wiza Jalakasi notes. “So, do you have different thresholds for chargebacks? Can agents accept refunds on behalf of a user? Can they redirect refunds? These are not edge cases; they fundamentally change how commerce works.”

A final question concerns the future of the checkout process. Will checkouts endure in the age of agentic commerce? Fantini, from EBANX, asserts that they will.

“Agentic commerce won’t eliminate merchant checkouts,” he believes. “Just like e-commerce didn’t eliminate physical stores. Checkout still plays a role in branding, loyalty, promotions, and experience.”