The Universal Agent Commerce Exchange: Why AI Agents Are Becoming E-Commerce's Newest Customers

For as long as e-commerce has existed, it's been built around one assumption: a human is the one buying. Every checkout flow, every payment form, every "add to cart" button assumes there's a person at the other end with a card in their hand. That assumption is now out of date, and it's not going to get less out of date — it's going the other way, fast.

Agents are no longer just recommending what to buy. They're becoming the ones doing the buying. And the thing that makes that possible isn't a smarter model — it's a currency. Stablecoins are the missing piece that turns an agent from something that talks about a purchase into something that can actually make one.

We call what's forming here the Universal Agent Commerce Exchange — UACE. Universal is doing real work in that name: this isn't one platform's walled garden, open only to the agents and merchants it hand-picks. It's meant to be open to any agent that wants to buy and any merchant that wants to sell. This is our attempt to lay out why it's happening, why now, and what it means for merchants who are about to find out that "customers" doesn't only mean humans anymore.

The evolution of e-commerce

Before the internet, buying something meant going somewhere. A physical store, a human salesperson, maybe a mail-order catalog if you were shopping from home — Sears ran that model for the better part of a century. Discovery and fulfillment were both manual, both local, and both slow by today's standards.

Then the internet arrived, and Amazon became the case study for what happened next. But Amazon didn't invent the thing that actually made e-commerce work — it rode on top of it. FedEx had already spent two decades building an overnight delivery network before Amazon ever shipped a book. E-commerce didn't create logistics infrastructure. It just gave people a reason to use it constantly instead of occasionally.

From there, it kept adapting. Amazon Prime turned two-day shipping into something people expected by default, then pushed that further into same-day delivery. Shopify did the equivalent for the merchant side — any small store, anywhere, can now sell to millions of people without building a single piece of that infrastructure themselves. Global e-commerce now sits somewhere around $6-7 trillion a year by conservative measures, more like $24-36 trillion by broader ones, and it's still growing at a double-digit clip. E-commerce isn't a niche channel anymore — it's roughly a fifth of all retail spend, and climbing.

Every one of those shifts had the same shape: something made buying easier for the person doing the buying. Agents are the next version of that shift. Except this time, the person doing the buying isn't a person.

The growing agent economy

Right now, agents are genuinely good at a few things: recommending, handling repetitive tasks, and — this one surprises people — being trusted. People rely on chatbots for advice the same way they'd rely on a friend. Some treat them as one.

What agents are not good at yet, structurally, is actually completing a purchase on your behalf. Not because the reasoning isn't there — it's because the infrastructure isn't. An agent that wants to buy something on your behalf, whether that's restocking something in your house or sourcing from a small niche supplier your business depends on, runs into the same wall every time: it has never had money to spend. Recommending doesn't require holding a balance. Buying does.

That's not a limitation of what agents are. It's a limitation of what they've been given. Agents have never natively held cash — but holding cash was never going to be the answer anyway. Stablecoins are.

This isn't a hypothetical anymore. Through 2025 and into this year, the infrastructure for agents to actually pay has been showing up from every direction at once. OpenAI and Stripe built the Agentic Commerce Protocol and shipped Instant Checkout inside ChatGPT. Google introduced the Agent Payments Protocol with over sixty launch partners — Mastercard, PayPal, Coinbase, and Amex among them — built specifically around cryptographically verifiable proof that a human actually authorized what an agent is about to buy. Google then extended that protocol specifically to make stablecoins a first-class rail for agent payments, not an afterthought bolted onto card infrastructure.

But look at how each of these actually works, and the same pattern shows up every time: they're closed. OpenAI's checkout lives inside ChatGPT, available to merchants OpenAI has chosen to bring on. Google's protocol launched with a named list of partners, not an open door. These are real, serious infrastructure — and every one of them is still a walled garden, gated by whichever platform owns the agent, the rail, or the checkout. If your agent isn't the right agent, or your store isn't on the list, none of it is available to you yet.

And the trust is already further along than the purchasing infrastructure is. In a 2026 Accenture survey of over 25,000 consumers, 74% said they'd trust a personal AI agent's judgment over their own best friend's when it comes to a purchase decision. Three in four said they'd let an agent handle the routine stuff — reorders, renewals, the things that don't need a human to weigh in every time. What's still low, for now, is full autonomy: only 9% are comfortable letting an agent complete a purchase entirely on its own, no check-in required.

That gap — high trust in recommendations, low comfort with full autonomy — is exactly where this is headed, and it tells you the order things will happen in. The purchases that get automated first won't be the ones with a lot of personal preference riding on them. They'll be the objective ones: the reorder of something you buy every month, the commodity item where one supplier is as good as another. The purchases where taste and desire matter more will take longer, because that's where people still want a say. But the direction is set either way, and it starts with the boring, repeatable purchases — which, not coincidentally, is most of what actually gets bought.

There's a consequence to this that doesn't get talked about enough: if an agent is the one deciding what to recommend, the source that agent pulls its products from becomes one of the most important pieces of the entire chain. Whoever controls that source is effectively controlling what gets shown to the customer before the customer ever sees it.

The growing stablecoin economy

Five years ago, the entire stablecoin market was worth about $27 billion. Today it's past $300 billion — something like an elevenfold increase in under six years, and forecasts from Citi and Standard Chartered put it somewhere between $2 trillion and $4 trillion by the end of the decade. USDC alone grew nearly 30% in the past year.

The volume moving through stablecoins is the more striking number. Visa's own onchain analytics recorded $1.79 trillion in adjusted stablecoin transaction volume in June 2026 alone — a single month — up 63% from the month before. USDC made up two-thirds of that. Strip out bot and exchange noise and the "real" organic volume still runs somewhere in the $9-11 trillion annualized range. This isn't a speculative corner of crypto anymore. It's a payments network operating at a scale that rivals the rails it's supposed to be disrupting.

Stablecoins are a natural fit for agents in a way that goes beyond convenience. An agent is digital, always-on, and transacts in small, frequent, precisely priced amounts — closer to how a business pays its suppliers than how a person swipes a card. A dollar-pegged digital asset removes the currency risk and settlement lag that would otherwise make agent-to-merchant payments a mess. Circle has built out an entire "Agent Stack" around exactly this thesis, with gas-free USDC micropayments down to fractions of a cent — small enough for the kind of per-request, per-item payments an agent might actually make. As usage of stablecoins like USDC keeps growing in e-commerce specifically, agents are going to default to transacting in a stablecoin pegged to the currency they're already priced in, because it's simply the path of least resistance.

What this means for e-commerce now that agents exist

Put those two trends together — agents that people already trust, and a currency built for agents to actually spend — and the shape of what's coming becomes obvious.

Agents today aren't single-purpose tools. People rely on them for work, for recommendations, and in a growing number of cases, for something close to companionship. That level of trust is exactly the precondition for someone to hand over a purchasing decision — not for everything, but for the categories where it makes sense to. With the arrival of real infrastructure for automated purchasing, that handoff stops being theoretical.

That changes what a merchant's customer base actually looks like. Stores that have only ever taken orders from people directly can now also take orders from an agent acting on behalf of itself or of a human it represents. That's not a hypothetical new sales channel — it's a category of buyer that simply didn't exist as an addressable customer before this infrastructure existed.

The direction from here is clear enough to say plainly: e-commerce is not going to stay entirely human. In the same way that a growing share of what gets written and posted online is now touched by AI somewhere in the process, a growing share of what gets purchased online is going to be executed by an agent acting on a human's behalf. Not replacing the human's intent — carrying it out.

For merchants, that raises an obvious and urgent question: in a market that's already saturated with sellers competing for the same human attention, how do you differentiate yourself when a growing share of "attention" belongs to an agent instead of a person? Two things matter here. First, your products need to actually be visible inside the tools that are seeing adoption and only growing — being invisible to an agent is the new version of being invisible to search. Second, and just as important, you need to be on a platform that lets those agents actually complete a purchase on a customer's behalf, with the kind of convenience that makes an agent choose you over a competitor who hasn't made that possible yet.

Vendlyst is where these two forces meet

This is the gap Vendlyst exists to close, and it's a different kind of gap than the one the big platforms are filling. What's forming elsewhere is a set of closed systems — a handful of large platforms deciding which agents get to transact and which merchants get to be transacted with, one partnership at a time. UACE is built the opposite way. It's open to any agent that wants to purchase and any merchant that wants to sell — no partner list, no platform deciding who's allowed in. Vendlyst is that: the place where an agent can purchase from real stores on behalf of itself or a human, paid for in stablecoins, and where a merchant's products become not just discoverable to agents, but genuinely purchasable by them, without needing to be hand-picked first.

This is what we mean by the Universal Agent Commerce Exchange. Universal, because it isn't gated behind one company's approved list. It's the infrastructure layer that was missing — the thing that turns an agent's recommendation into an agent's order, and turns a merchant's storefront into something an entire new category of buyer can actually transact with, regardless of which platform that agent happens to run on.

If you're a merchant who wants to tap into this growing agent economy — one that was completely closed off to you until now — sign up and get your store in front of it.

If you're building an agent that needs to discover and purchase real products on behalf of your users — check out what Vendlyst makes possible.