Key Takeaways
- At Visa Payments Forum 2026 on June 10, Visa made a wave of announcements spanning AI, tokens, and stablecoins, and tied them together under a single idea it calls Intelligent, Programmable Commerce.
- The individual announcements fall into three layers — the front end where AI agents start shopping, the trust that rides along with a transaction, and the back end where money actually settles — and Visa is positioning itself as the trust layer across all three.
- For e-commerce and booking businesses, whether your site is readable by AI agents, and whether your payment partners adopt the new standards, is becoming a competitive condition on par with search engine optimization.
The Single Thread Running Through Scattered Announcements
Visa unveiled a series of new AI, stablecoin, and tokenization capabilities at Visa Payments Forum 2026, unifying them under Intelligent, Programmable Commerce.
usa.visa.comOn June 10, 2026, at Visa Payments Forum 2026 in San Francisco, Visa announced an almost overwhelming number of new capabilities at once: a partnership with OpenAI, expanded stablecoin settlement, and a cluster of tools with unfamiliar names. The press covered each one separately, but the more you follow them piece by piece, the harder the whole picture becomes to see.
The thread running through all of it is the keyword Visa chose: Intelligent, Programmable Commerce. Chief Product and Strategy Officer Jack Forestell captured the intent in a single sentence.
AI is transforming the front end of commerce. Stablecoins are reshaping the back end. Visa's role is to enable it to work securely, reliably and at global scale, for every participant in the ecosystem.
Follow that framing, and the announcements make sense once you split them into three layers: the front end where buying begins, the trust that decides whether a transaction can be believed, and the back end where money finally moves. Visa is not claiming any one of these. It is positioning itself across all three as the layer that provides security, trust, and control.
| Layer | What it handles | Key announcements |
|---|---|---|
| Layer 1 — Front end | Where AI agents begin shopping | Visa Intelligent Commerce, OpenAI partnership, Agent Score, Agentic Directory, Trusted Agent Protocol |
| Layer 2 — Trust | Evidence for whether a transaction can be trusted | Enhanced Token Data, Token Assurance Signal, Large Transaction Model |
| Layer 3 — Back end | Where money finally settles | Stablecoin settlement (~$7B annualized), tokenized deposits, 160+ linked-card programs |
What was announced in each layer, and what it means for our businesses? Let's go through them in order.
Layer 1: The Front End That Welcomes a New Kind of Customer — the AI Agent
In the commerce to come, the party that visits a site and clicks the button shifts from a human to an AI agent. What Visa lined up on the front end were the parts needed to welcome that new customer safely.
At the center sits Visa Intelligent Commerce, a vision underway since April 2025. It is a platform that bundles the trust, control, and connectivity an AI agent needs to discover products and to initiate and complete transactions — and the headline strategic partnership with OpenAI is one element of it. We covered that partnership, including spending limits and the division of fraud liability, in a separate article.
Two merchant-facing pieces matter most in practice here. The first is Agent Score, a metric — built with the AI startup New Generation — that diagnoses whether AI agents can navigate your site, understand it, and complete a purchase. A site that looks beautiful to a human is not necessarily easy for an agent to operate, and Agent Score puts a number on that gap. The second is the Agentic Directory, a registry of agents and merchants that Visa has verified, so each side can confirm the other is legitimate.
What technically underpins this front-end trust is the Trusted Agent Protocol. Introduced in October 2025 with more than ten partners, it is a shared standard that lets merchants tell a legitimate agent apart from the malicious bots that plague their sites. The heart of it is cryptographically signed HTTP messages: an agent transmits its intent, a verified user identity, and payment details under a cryptographic signature, and the merchant validates that signature using Visa's public keys. The signature carries a timestamp and a session-specific identifier, preventing reuse and replay. The backdrop is a 4,700% surge in AI-driven traffic to U.S. retail sites — a real reason to build a checkpoint that lets only genuine agents through.
As the entry point moves from humans to agents, machines have to judge "can I trust the party on the other side?" in an instant, and with one another. Layer 1 is the foundation that makes that judgment possible.
Layer 2: Giving Transactions Trust, Through Tokens and AI
A transaction that clears the entry point next faces the question: can this transaction be trusted? Visa answers by strengthening the tokens it has relied on for years.
Visa's tokens have long replaced card numbers with secure substitute values. Two enhancements are now added. Enhanced Token Data gives the token richer context — the type of transaction, where it is used, who is paying. Token Assurance Signal evaluates how a token was provisioned and used across its full lifecycle, and turns that into a single indicator of the trust behind a transaction.
Paired with this is an AI trained on billions of transactions, the Large Transaction Model. Its aim is to ease a dilemma the payments industry has lived with for years. Screen too hard for fraud, and legitimate transactions get wrongly rejected — the false declines. Loosen approvals, and fraud slips through. A model trained on enormous transaction volume raises the precision of that tug-of-war, pursuing both fraud detection and higher authorization rates.
In agent transactions, where a human cannot check each one, the trust judgment has to be left to machines. Embedding identity and behavioral history inside the token, so that trust travels with the transaction, is what Layer 2 makes concrete. A fuller explanation of these token enhancements also appears in our article on stablecoins and tokenization.
Layer 3: The Quiet Redesign of the Back End Where Money Moves
Of the three layers, the change on the back end is the least flashy and the most structural. On screen a payment feels instant, but behind it banks run a step called settlement to finally reconcile funds — traditionally bound to banking days and taking several of them.
Visa has been running pilots that replace this with stablecoins, and that settlement reached an annualized run rate of about $7 billion as of March 2026. That is live transaction volume, up roughly 50% in a short span from about $4.5 billion in January. Alongside it, Visa announced tokenized deposits — letting banks gain stablecoin-like immediacy while keeping funds on their own books — and more than 160 stablecoin-linked card programs globally.
Because this back-end redesign sits slightly outside this article's focus, we leave the breakdown of the numbers and the mechanics of VTAP (Visa Tokenized Asset Platform) to a dedicated explainer. What to hold onto here is the shape of it: both ends — the front (AI) and the back (stablecoins) — are moving at once, and Visa wants to stand in the position that connects them.
Visa's Pragmatic Answer: "Connection, Not Full Replacement"
Hearing all of this, a business might brace for the idea that it must rip out and replace its entire stack or be left behind. Visa's answer is clear: full replacement is not required.
What the company kept stressing was modular, phased adoption that connects to existing infrastructure. For issuers, the cloud-native core banking platform Pismo enables a gradual move toward real-time processing. For the merchant side, Unified Checkout brings card and non-card payments into one orchestration layer, and Visa Intelligent Authorization helps optimize approval rates. None of these demand a rebuild from scratch; each is designed to be added onto what already exists.
This pragmatism ties to the timeline for returns. Visa CFO Chris Suh said of the investments in agentic commerce and stablecoins that they "won't pay off in the next six months, but could pay off over six years." Rather than leaning too heavily on a flashy narrative, Visa keeps its existing business running while laying long-term plumbing — a consistently cautious posture.
What E-Commerce and Booking Businesses Should Prepare
So, faced with this three-layer shift, where should a business start? It will not change tomorrow's revenue — yet the entry point to preparation is surprisingly close at hand.
The first question to ask is whether your own site or app is "readable by AI agents." That is precisely what Agent Score sets out to diagnose: whether the flows you polished for humans have become obstacles for agents. This is becoming a new competitive condition, the equivalent of search engine optimization. Indeed, S&P Global's 2026 merchant survey (cited by Forbes) found that 65% of merchants are considering adding a new payment partner to handle agentic commerce, and more than two-thirds expect at least 10% of e-commerce transactions to be agent-initiated within three years.
Booking businesses in travel and transport are no exception. An experience that hands an agent a "money-touching, complex transaction" — a booking with changes and cancellations — only works once all three of today's layers are in place: spending limits, approval requirements, and trust-bearing tokens. How the payment providers and issuers you depend on will adopt this new standard is, for now, the most practical thing to understand.
Conclusion
Seen in fragments, the Visa Payments Forum 2026 announcements look like a pile of product news. But rearrange them into the three layers — front end, trust, back end — and a consistent strategy emerges: as AI changes how buying begins and stablecoins change how money ends up moving, Visa intends to sit at both ends and in the middle as the trust layer. You do not need to memorize the product names. Holding two questions — is my site visible to agents, and where is the back end of payments heading — is the first step toward not being left behind by this larger shift.
