From autonomous agents executing purchases to network-level credential transformations, the payments ecosystem is entering a moment of rapid realignment in agentic commerce. Networks, issuers, and merchants all recognize that agent-driven transactions represent a new revenue surface, a new fraud surface, and a new control surface. And they're each lobbying, loudly, for the right to shape how these transactions are accepted, routed, and evaluated.
Below, let's focus on the merchant lens, while unpacking how networks, issuers, and AI platforms are positioning themselves and why merchants must engage now rather than react later.
Networks Are Racing to Define the Future: "Agentic Tokens"
The networks are in a full sprint. They're aggressively pushing new token formats, the emerging “agentic tokens,” that mirror the industry's earlier shift toward network tokens (DPANs in Apple Pay, network tokenization for card-on-file, etc.). These new formats are designed to be leveraged by autonomous agents rather than human cardholders.
The motivation is clear:
- Control the credential rails. If agentic commerce becomes a high-volume channel, networks want their token frameworks to be the standard.
- Expand visibility. New token types provide networks with the leverage to request more insight into the underlying funding source.
- Prevent disintermediation. If agents drive more consumer spend, networks want to ensure those flows remain on-network and on-token.
As a result, the rollout pace is unusually rapid. Standards are forming in real time, not in years-long cycles. While this is happening, merchants need to act fast, or they won't have time to provide their perspective.
Issuers: Interested but Wary (Especially Around Fraud Liability)
Issuers face a difficult calculus.
On the one hand, cardholders will want to engage in agentic transactions, such as automated price shopping, subscription management agents, and personal concierge models, to name a few. Issuers don't want to block spending that their customers explicitly approve.
But issuers are equally aware that:
- There is currently no liability shift. Issuers do not want to absorb the loss of fraud from an entirely new credential modality.
- There is no data yet. We lack evidence that agentic tokens perform better (or worse) on authorization than any other credential form.
- Behavioral signals are obscured. Human-driven transaction patterns are well-modeled. Agent-driven patterns are not.
The result? Issuer behavior is unpredictable. Some may authorize conservatively. Others may experiment. Some may treat agentic credentials like card-not-present (CNP); others may penalize them until fraud patterns stabilize.
For merchants, this creates volatility in conversion, and volatility is expensive.
AI Platforms: A Unique Role
The dynamic here is a bit unusual.
AI platforms are effectively playing both sides of the field. On one hand, they want to enable cardholders to transact seamlessly within their platforms. On the other hand, they need merchants to be able to successfully accept those transactions; otherwise, any friction or failure is likely to be attributed to the AI platform rather than the merchant.
These platforms also need transactions to work across as broad a set of merchants as possible. In practice, the platform that enables purchases at more merchants is better positioned relative to competitors. At the same time, these platforms are highly motivated to be first to power this new wave of transactions.
The Structural Risk: Abstracted Cards + Abstracted Identities
A broader industry truth sits beneath all of this:
Network tokens abstract the card.
Agents abstract the human identity.
Apple Pay DPANs, card-on-file network tokens, and now agentic tokens all distance the merchant and the issuer from the true underlying PAN. Meanwhile, agentic commerce distances the merchant from the true underlying human.
Combine the two, and the identity surface becomes dangerously thin.
- The credential no longer reliably represents the cardholder.
- The agent no longer reliably represents the cardholder's intent.
- Fraudsters now have two layers they can exploit instead of one.
This is why issuers hesitate. And it's why merchants cannot passively accept whatever standards emerge; they will be the ones paying when fraud occurs.
Merchants: The Most Impacted Stakeholder
As we've said, networks want to control the agentic commerce rails as quickly as possible; issuers want to provide a strong cardholder experience but must remain cautious about fraud; and AI platforms want the best possible checkout experience across all merchants, while still needing merchants to succeed in charging the credentials they pass through.
Despite the impact on all of these parties, merchants have the most to worry about.
Loyalty programs could disappear; their checkout and conversion experience may no longer be visible; they would know less about their customers; they wouldn't receive raw credentials (only tokens); they wouldn't be able to prevent or stop some transactions originating from front-end channels; and they would still be liable for fraud.
The most vocal merchants already understand what's at stake. Agentic commerce represents a new revenue channel, a frontier in customer experience, and a new opportunity to differentiate. But merchants also shoulder the bulk of operational and financial risk.
Specifically:
If an agent uses a stolen credential, the merchant absorbs the chargeback, not the network, not the issuer, not the agent platform.
- Was the agent actually authorized by the customer?
- Is the credential legitimate or compromised?
- Is the funding source one that the merchant is comfortable accepting?
Without visibility, merchants lose the ability to answer those questions.
A luxury retailer may require strong underlying identity signals.
A digital goods merchant may prefer frictionless authorization.
A subscription provider may prioritize credential stability over transparency in risk scores.
One size will not fit all. The ecosystem must account for merchant-level preference and policy—even when transactions originate from autonomous agents.
Why Merchant Involvement Is Not Optional
Merchants must insert themselves into the design of agentic commerce now, not later, because:
- They are the ultimate acceptors. If merchant acceptance falters, agentic commerce stalls.
- They own the fraud risk. Liability drives cost; cost drives margin; margin drives willingness to participate.
- They hold critical user experience power. If merchants reject agentic credentials, the consumer experience breaks.
In other words, merchants can be king-makers or gatekeepers. The ecosystem needs them engaged, or the agentic commerce vision collapses into fragmented adoption and inconsistent authorization.
Conclusion: Agentic Commerce Will Only Work If Merchants Are Involved in Shaping It
Networks want credential control. Issuers want fraud containment. Agents want autonomy. But merchants want, and need, something different: clarity, trust, and choice.
The success of agentic commerce depends on designing a framework where merchants can:
- Understand the agent making the purchase
- Understand the credential being used
- Manage the fraud risk that ultimately lands on their business
