Will AI agents become the new gatekeepers of ecommerce?

Wolf Ruzicka, Chief Commercial Officer at Unlimit, examines what happens to merchants and payment methods once AI agents mediate the buying journey, from what decides inclusion in an agent's shortlist to where liability sits when a mandate is breached.

When a shopping agent returns a shortlist of three or four options instead of a full search results page, the merchants that do not make the list effectively disappear. In your view, what will decide inclusion? And how much of that is genuinely within a merchant’s control?

Brands competing for consumer attention will increasingly need to build a digital reputation that AI agents can understand and evaluate. Inclusion is likely to come down to three things: whether an agent can understand what a merchant is offering, whether that proposition is competitive, and whether the merchant has demonstrated that it can reliably deliver what it promises.

User reviews, ratings on platforms such as Trustpilot, and customer feedback across social media will all contribute to the information agents use when deciding which products or merchants to recommend. While merchants cannot fully control these signals, a lot remains within their control: price, promotions, product quality, availability, customer experience, and fulfilment.

We have spent years talking about SEO for humans and search engines. The next challenge is optimisation for AI agents. Merchants’ product catalogues, offers, pricing, and availability need to be machine-readable, up to date, and contextually relevant. Structured product feeds, API readiness, and accurate information on delivery and returns will become increasingly important.

An agent’s decision will ultimately reflect a combination of merchant-controlled fundamentals and wider signals. Merchants can influence those fundamentals, but they cannot fully control how an agent weighs them or the commercial and ranking logic of the platform itself.

Earning visibility will increasingly mean building a reputation that is trusted by people and legible to machines.

Payment method selection has traditionally happened at checkout and been made by the shopper. If an agent picks the rail based on cost, speed, or reward logic, what are the effects for card schemes, wallets, and local payment methods? Unlimit operates across markets where alternative payment methods are extremely popular, so does agentic commerce strengthen or weaken those local champions?

Local payment methods could actually become more important in agentic commerce, even if consumers become less aware that they are using them. If an agent chooses the rail based on what is most advantageous, the user should ultimately benefit. A credit card may be preferable for someone wanting to spread a purchase into instalments, while Pix in Brazil or UPI in India may be more attractive when paying upfront offers an advantage. Digital wallets can compete through cashback, loyalty benefits, or partner discounts.

We are already seeing payment methods compete and evolve. In Brazil, Pix has expanded into recurring payments through Pix Automático, bringing instant payments into areas traditionally dominated by cards. Card schemes are adapting too: Elo, for instance, has explored biometric authentication for Pix payments within merchant journeys. The boundaries between payment rails are beginning to blur.

Agentic commerce will accelerate this competition. If an agent selects a rail based on cost, speed, acceptance, rewards, or consumer preferences, strong local methods can become even more relevant. Rather than requiring a consumer to recognise or actively select a domestic payment brand, the agent can choose it because it offers the best outcome for that transaction.

Payment methods will therefore need to demonstrate why they are the best fit for both consumers and the agents acting on their behalf.

Unlimit’s role is to make this diversity accessible rather than forcing merchants or agents towards a single global rail.

With agent-initiated transactions, an authentication problem emerges: proving that the agent had authority to buy and that the mandate behind the purchase was genuine. How do you see this working alongside Strong Customer Authentication (SCA) under PSD2 and the incoming PSR, and where does liability sit when an agent buys the wrong thing or is manipulated into a purchase?

The key question in agentic commerce is no longer simply: ‘Did the customer authenticate this payment?’ It becomes: ‘What did the customer authorise the agent to do, and did this transaction remain within that mandate?’

Under the current SCA framework in PSD2, authentication is largely designed around the customer actively authenticating a payment. Agentic commerce introduces a different model. Here, the customer could authenticate a mandate in advance and delegate specific purchasing decisions to an agent. This process could involve cryptographic credentials, spending limits, and merchant or category restrictions that define what the agent is authorised to do. As a safeguard, transactions outside those parameters could require step-up authentication or renewed human approval. PSD3 and the Payment Services Regulation will update the wider European payments framework, but agentic commerce raises questions around delegated authority that regulation does not yet fully answer. The industry will therefore also need robust standards for agent identity, auditability, and the scope of customer mandates.

Liability becomes particularly important if an agent is manipulated or compromised, for example through prompt injection, and makes a transaction outside the customer's mandate. There must be a verifiable record of what the customer authorised the agent to do. Determining responsibility will depend on whether the transaction was genuinely authorised, whether it remained within that mandate, and whether each participant met its security and fraud-prevention obligations. As agentic commerce grows, it will require a shift from authenticating individual payments to authenticating and enforcing the mandate under which an agent is allowed to transact. Clear, auditable boundaries will be essential for trust, security, and liability.

The standards layer is fragmenting, with ACP, UCP, MCP, AP2, x402, and scheme-level frameworks from Visa and Mastercard all in play. Do you expect consolidation around one or two of these by 2028, and what is Unlimit doing to keep merchants from guessing the protocol to avoid?

By 2028, I expect convergence, but not necessarily one winning protocol. These standards are not all trying to solve exactly the same problem: different approaches address areas ranging from agent-to-merchant interactions and delegated payment authority to machine-to-machine transactions and the use of existing card networks. We may therefore see consolidation around a smaller number of interoperable standards rather than a single dominant framework.

What will ultimately determine which standards succeed is their ability to establish clear authority and context behind each transaction, while making transactions auditable and providing simple mechanisms for failures, disputes, and remediation. Delegated authority needs to be portable, transactions traceable, and consumers must be able to clearly define and control the boundaries within which an AI agent can act.

Merchants should not translate this evolution into another layer of complexity. When a merchant works with the right partner to expand into a new market today, it does not need to determine individually which licences to obtain, which payment methods to integrate or which local infrastructure to build. For example, Unlimit abstracts that complexity through a single integration which connects merchants to the capabilities they need across markets.

That is the approach we believe payments providers should bring to agentic commerce. Merchants should not spend the next two years trying to predict whether ACP, UCP, MCP, AP2, x402, or a scheme-level framework will prevail. The payments infrastructure underneath them should absorb that complexity and adapt as the standards mature.

If agents own the interface, aren’t merchants risking losing the direct customer relationship, the data trail left behind, and the loyalty mechanics built on top? Is there a realistic path for merchants to keep that relationship in an agent-led journey, or does the value shift permanently to whoever owns the agent?

Losing the interface is not the same thing as losing the customer relationship. Although the way people shop may change, that does not necessarily mean that the relationship between consumers and brands will disappear. Consumers will still have preferences around products, quality, delivery, customer service, rewards, and the brands they trust. What will change is where and how consumers and brands manage that relationship.

As the purchasing journey becomes mediated by agents, merchants will need to find new ways to capture customer preferences and build loyalty. The relationship will therefore become less about owning the interface and more about creating reasons for customers to maintain a direct connection with the brand.

We are already seeing emerging agentic commerce models explore account and identity linking, which can allow a customer’s existing loyalty status, preferences, or benefits to follow them into an agent-led journey. We could also see merchants rewarding customers with exclusive benefits, like discounts and loyalty points, for voluntarily sharing proof of purchase or transaction information through their own channels. These incentives could be built into the agent-led journey, allowing merchants to continue collecting valuable first-party signals with the customer's consent.

The agent may own the interface, but the merchant can still own the brand relationship. The real risk is that merchants give customers no reason to maintain it once an agent can handle discovery, comparison, and purchasing on their behalf.

Brands that continue to offer differentiated products, service, and loyalty benefits will still have something an agent platform cannot easily replace: the customer’s preference for that brand.

About Wolf Ruzicka

Wolf Ruzicka is Chief Commercial Officer at Unlimit, driving global commercial strategy and growth through AI solutions. With 25+ years of experience, he has built and scaled high-performing businesses, including growing EastBanc Technologies, Solvd, APIphany, and MicroStrategy. An entrepreneur, investor, and author of AI Driven, Wolf has served on multiple boards, advised global enterprises, and mentored extensively within the technology and venture community.

 

 

About Unlimit

Unlimit is a global financial infrastructure for the emerging agentic economy. Its proprietary payment network spans more than 180 countries and supports more than 150 currencies, integrating directly with hyper-local payment systems and powering cross-border commerce for businesses, consumers and AI agents. Unlimit operates Stable.com, its full-service financial platform for businesses and individuals. Founded in 2009, Unlimit is headquartered in San Francisco and London, with 17 offices worldwide. Its clients include Xiaomi, Telegram, Convera, and inDrive. The company is led by its Founder and CEO, Kirill Eves. For more information, visit unlimit.com and stable.com.