Next-generation payments are not within the payment layer

Author: IreneDu

This is the 2.5th article in the Stripe AI strategy breakdown series.

The origin of this series is due to the release of 288 products at Stripe Sessions 2026 on April 30, where I observed that Stripe is trying to become the economic infrastructure for the AI Agent era.

The first article, "Stripe is Not a Payment Company," attempts to answer "Why Stripe" — its genes determine its ability to do this.

The second article, "KYC is Dead, the Agent Economy is Rewriting the Underpinnings of Financial Regulation," aims to break down the future Stripe is betting on — what the Agent Economy looks like and why traditional payment infrastructure will completely fail before it.

However, during the second article, I received a comment from a peer:

I completely agree with the first half. Whether it's AB 316 or the laws of any sovereign country, they will not recognize "Agents as legal entities" in the short term — the ultimate defendant will always be a specific person. This is something that Know Your Agent cannot change, nor can it be changed.

But regarding the second half — "the only change is in payment and settlement efficiency" — I reserve my opinion. The problem with this statement lies not in the conclusion but in the framework it assumes: it views KYA as an upgrade to the existing payment system.

This is what I believe deserves further discussion.

Let’s return to the muscle memory of a former payment practitioner:

Payment forms are scenario-driven, not designed from within the payment system.

The real leap in payment each time — online banking, mobile wallets, QR code scanning — did not occur because someone created a better product at the payment layer, but because a new transaction scenario emerged that broke through the underlying assumptions of the original payment system.

New payment forms grow out of the infrastructure required by that scenario, not optimized out of it.

I once worked on payment innovation at Ant Group. At a platform that was an absolute leader in the industry, having created "Quick Payment," "Mobile Payment," and "QR Code Payment," the greatest joy and pain was pondering: what is the next generation of payment forms?

We developed watch payments (along with heartbeat verification as an alternative to facial recognition), NFC payments (the original technology for "tap to pay"), participated in and wrote several "next-generation" payment protocols, and even tried to get my boss to support my exploration of metaverse payments.

Most of these projects did not come to fruition.

Looking back, the reason is the same: we tried to define new payments at the payment layer, but the scenarios driving payment transformation had not yet arrived — without the scenarios, the infrastructure required by the scenarios cannot grow, and no matter how clever your design at the payment layer, it cannot hold up.

The Agent Economy is the new scenario that I have been waiting for.

KYA is the layer of infrastructure that is emerging.

KYA is not a product at the payment layer; it is the infrastructure layer of the Agent Economy.

The five layers of KYA I defined in the previous article — Agent Identity, Authorization Scope, Intent Signature, Responsibility Chain Audit, Credit Rating — only the Authorization Scope and Responsibility Chain Audit layers fall on the payment link; the other three layers (Identity, Intent, Credit) are not part of payments at all.

• The Identity layer serves all scenarios that require identifying Agents: cross-platform calls, regulatory filings, internal audits — payment is just one of them.
• The Intent layer addresses the larger issue of AI alignment — payment is just one of its many verification scenarios.
• The Credit layer serves any system that needs to allocate permissions and limits to Agents — payment is also just one of the use cases.

So the judgment of that peer, "the only change is in payment and settlement efficiency," translated into infrastructure language means: viewing KYA as a subsystem of payments.

My judgment is the opposite: payments are a subsystem of KYA.

This reversal is the core of this article.

Stripe's investment actions at the industry front line are precisely empirical evidence.

The term Patrick Collison used at Sessions 2026 was not "AI payments," but "economic infrastructure for AI." This is not marketing language; it is a positioning choice. It indicates that Stripe does not intend to lock itself into the identity of a "payment company"; it is betting on being the foundation for the Agent Economy.

Specifically regarding product layout:

The Agentic Commerce Protocol (ACP) co-built by Stripe and OpenAI is now used by Microsoft Copilot, Meta, and Google Gemini, which joined in April this year — it is essentially an identity and session protocol, not a payment protocol.

The Shared Payment Token separates Agents from real card numbers, functioning at the authorization layer, not the settlement layer.

Stripe's acquisition of Bridge for stablecoin infrastructure, acquisition of Privy for embedded wallet capabilities, and the establishment of its own Tempo blockchain for settlement pipelines — this entire layout does not fit within the framework of "payment efficiency optimization."

This investment portfolio only makes sense under the judgment that "KYA is the infrastructure layer." If the Agent Economy were merely a payment efficiency issue, Stripe would not need to create stablecoins, embedded wallets, or build its own L1. What it is doing is gradually occupying positions in those five layers of KYA.

Emily Glassberg Sands, Stripe's data head, provided several numbers in an interview with Every this April, which corroborate the same point from another angle: a large AI client is intercepted 250,000 times a week for fraudulent free trials; she has seen an AI company burn $25 in computing power for each free trial with a conversion rate of 4%, meaning each new paid user costs $625; the overall abuse of free trials has increased fourfold in the past six months.

These numbers collectively indicate one thing: in the AI economy, the real judgment of whether a transaction can succeed and is worth pursuing no longer occurs at the moment of checkout — it happens upstream in questions like "who is this, what do they want to do, and is it worth allocating resources." This is why Stripe wants to move risk control Radar from "the moment of transaction" to "the entire user lifecycle": it is not about making old risk control faster, but about shifting the focus of risk control from "is there a problem with this payment" to "is there a problem with the entire behavior of this user/Agent." The former is a payment layer issue; the latter belongs to the KYA layer.

Returning to that peer's question: who ultimately bears responsibility?

He is correct — the ultimate legal entity is still a specific person. This has already been legally established by AB 316.

But this is precisely the real problem KYA aims to solve: when the responsibility chain becomes distributed, finding "which specific link falls on which person" is something that did not need to be done in the KYC era but must be done in the KYA era.

In the KYC era, the responsibility chain is linear (user → payment/bank → merchant); when a transaction goes wrong, you intuitively know who to look for.

In the KYA era, the responsibility chain is networked (user → Agent platform → model provider → payment protocol → bank → merchant, possibly involving other Agents in between), and even if the law tells you to "find a person, not an Agent," you still do not know which person to look for — because responsibility is distributed among 5 to 7 entities.

KYA cannot change the ultimate legal attribution. But it can, within the networked chain, use cryptography to solidify the roles and actions of each entity — who authorized what, who executed what, who settled what, who fulfilled what. It transforms "unable to find evidence" into "able to find evidence"; it changes "which link went wrong is unverifiable" into "verifiable."

This is not an enhancement of payment efficiency.

This is the first time that accountability traceability can occur within the Agent network.

Therefore, the statement "the only change is in payment and settlement efficiency" misrepresents the infrastructure and functionality.

What is truly happening is:

• Because a new type of economic actor (Agent) has emerged, a new layer of infrastructure (KYA) has been forced to grow;
• This layer of infrastructure redefines "who is on the other side, what they can do, and who to turn to when things go wrong"; on top of this layer of infrastructure, payments will reorganize themselves in a form we cannot fully see today.

What will the next generation of payment forms be? It is still unclear, which is precisely the new species Stripe is trying to define.

But in a world of uncertainty, one thing I am very sure of — it will not be designed at the payment layer.

It will grow out of the scenario after the KYA infrastructure is laid out.