According to PYMNTS.com, Visa’s senior vice president Rubail Birwadker outlined the company’s strategy for the coming wave of “agentic commerce,” where AI assistants handle shopping. The core challenge is moving from an assistant that can search and compare to one that can actually complete a purchase, which requires continuous validation of the agent’s behavior. Visa’s solution involves creating a “trust layer” built on a concept called “Know Your Agent” (KYA) to verify an agent’s cryptographic identity and permissions. They’ve launched a Trusted Agent Protocol to let verified agents pass a digital signature to merchants with parameters like spend limits. The goal is to use Visa’s existing token network, which already handles tens of billions of tokens, to securely bind credentials to agents. Ultimately, Visa wants checkout to recede into the background, becoming an invisible step in an AI-driven transaction.
The Invisible Checkout Problem
Here’s the thing: Visa isn’t trying to invent a new “buy” button for AI. They’re trying to make the button disappear entirely. Birwadker frames this as the natural evolution of one-click checkout—maybe even “without a button.” The interface could be your voice, or a text chat with a bot. But the feeling at the end should be the same: you wanted something, and you got it, with minimal friction.
But this isn’t just about consumer convenience. It flips the script for merchants, too. Imagine your website traffic is now a mix of humans and hyper-informed AI agents, all with perfect price comparison data. That changes everything about demand, branding, and leverage. And yet, behind the scenes, merchants are staring down a “huge nightmare” of integration. They can’t build a custom pipe for every new AI shopping model that comes along.
Building Trust in a Bot World
So how do you make this work without chaos? Visa’s answer is a two-part foundation: identity and control. The “Know Your Agent” (KYA) concept is crucial. It’s not enough to know the human user; you have to verify the *agent* acting on their behalf and establish a “root of trust.” This links the authenticated human to the bot and defines what the bot is actually allowed to do.
Their Trusted Agent Protocol is the start. A verified agent can pass a signature to a merchant that says, essentially, “I’m authorized by Jane Doe to spend up to $200 on home goods today.” This is where tokenization becomes the superstar. Visa’s tokens are no longer just about hiding your card number. They’re becoming programmable containers that can hold identity and permissions, bound to a specific agent. And if something goes wrong? The token can be revoked instantly, killing the agent’s spending power without exposing any underlying personal data. That’s powerful.
The Fragmentation Trap
The biggest hurdle, as with any new tech wave, is fragmentation. Birwadker admits we’ll see a messy early phase where every AI platform and big tech company tries to build its own little walled garden of trust. But he thinks standardization is inevitable. Why? Because merchants operate globally. They can’t deal with 50 different agent verification schemes.
Visa’s bet is that core trust standards will have to be established through industry bodies, creating a common language that works everywhere. Custom “pipes” will still exist for specific needs, but they’ll plug into a standard foundation. Without this, agentic commerce just can’t scale. It would be like every online store inventing its own version of SSL in the 90s—a total non-starter.
Why This Matters Now
Look, AI shopping agents are coming. The question isn’t *if*, but *how* they’ll fit into the financial plumbing we all rely on. Visa’s play is classic Visa: become the trust and security layer in the middle of a chaotic new ecosystem. They’re leveraging what they already have (a massive token network) to solve a new problem (agent identity).
The real test will be in the execution. Can they get enough players to agree on those standards? Can the risk models that catch human fraud also catch high-velocity, automated agent fraud? Birwadker’s vision is that if they do their job right, you won’t even notice the difference between buying something yourself and having an AI do it for you. That’s the goal: making a fundamentally new and weird process feel boringly normal and safe. And in payments, boring trust is the most valuable kind.
