Ask someone how advanced the dealership trade is and you’ll usually get an adjective. “early.” “Fledgling.” “Emerging.” Karen Webster wanted a number.
So she gave Karan KatyalGlobal Head of Agent Trading at Aden,scale. Zero means we are nowhere. The number five means that brokerage is just an integral part of how a trade happens, like clicking a card. Where are we now?
“We are at version 0.5,” Katyal said. “It’s early days. Still.”
Not one. half. In the most interesting twist in commerce since the advent of the smartphone, we put the person who runs Adyen’s dealership business a tenth of the way to the halfway point. This is a reality check. Now this is the part that should make every trader stand up.
Webster fired him for a year. Where are we next June? Katyal’s answer was 1.5.
0.5 to 1.5 is one point on the scale, which seems like a rounding error. But it’s also a trilogy. We will, in twelve months, accomplish three times more than what has been built so far. After tripling, we would still be 70% below the halfway mark. Both things are true at the same time. The slope is steep and the summit is far away. That’s the full form of this moment in one sentence.
The AI wasn’t the hard part at all
Here’s Katyal’s point of view. Intelligence works. People already rely on large linguistic models to search for products, compare options, and narrow down a list of twenty to three. Discovery seems easy. Everything falls apart after the discovery.
“Infrastructure is probably a much bigger cluster than people imagined,” Katyal said. And he is polite. A traditional product listing might need dozens of attributes to appear well on a web page. An AI agent trying to differentiate between two similar products would need much more, perhaps three times more. Detailed specifications, reviews, context, and the things a human reads between the lines. Most catalogs simply don’t carry it. Making it machine readable is not a modification. It’s a project.
Then there is the confidence that Katyal kept coming back to.
“Trust, risk and fraud should ultimately be at the heart of the foundation you are building,” he said. It was not installed in the end. In the center. Because the moment the agent acts on your behalf, the questions become more precise. Who is responsible when an agent books the wrong flight? When he misreads what you ordered and buys it anyway? A failed exit is annoying. Failed delegation is a dispute over responsibility.
This is not one thing, that’s why it’s slow
Part of the reason we stay at 0.5 is because “proxy commerce” is not a single product that anyone can ship. Katyal described several different models that arose at the same time, each with its own technical requirements.
In some cases, the person stays in the loop all the way. In other cases, the agent purchases independently once you set the rules. Then there is machine to machine, where software interfaces with software. And business procurement, which is its own animal altogether. Four directions, four sets of standards, and no judgement.
“Fragmentation will continue for some time,” Katyal said. “The proxy trade is really a direction of travel.” It is not the destination you arrive at. The address you point to as competing protocols, setup flows, checkout requirements, and payment methods sort themselves out beneath you.
That’s why he’s upfront about money. “The agent business is very much like a series of experiments,” he said. “If the thing you need to justify is hard numbers, now is not the time for that yet.”
But don’t read “0.5” as “come back later”
This is where Webster has fallen behind, and where traders should pay attention. Low maturity is not permission to wait.
I’ve seen this movie before. Companies that were slow to shift from stores to digital, and then again from digital to mobile, spent years regaining ground and didn’t have to lose. Waiting for certainty has never been a winning move. Consumers, as she puts it, are already “knocking on the door.”
The problem is that they can’t get past it. “Consumers can’t close the loop,” Webster said. The benefit is real. The intent is real. The path of purchase intent is broken. And the merchants who fix half of this path now, during the trial period, are the ones who will be ready when the loop finally closes.
So what gets fixed first?
Katyal expects to drive low-drama purchases: households restock frequently, and low-concern purchases where handing over the task removes friction without making anyone nervous about losing control. Webster added a twist. AI may turn out to be just as valuable at the top of the consideration curve, in the larger purchases considered, where the real payoff is the aggregation and comparison, not the click that closes them.
They both agreed that there would not be one universal way to shop. He will bow to the person and the moment. Sometimes you might want an agent to handle it. Sometimes you will want to drive. The infrastructure must support both.
Adyen’s Bet: Build links, not another platform
Against this backdrop, Adyen has introduced Adyen Agentic, which brings together product discovery, checkout coordination and payments into modular pieces aimed at reducing integration headaches across different AI commerce protocols.
Logic follows everything above. Help merchants make their product data machine-readable. Allow them to keep the payment and fulfillment systems they already run. Extend the authentication, encryption, and fraud controls they trust in proxy environments, without forcing a rebuild for every new platform that comes along.
In a world stuck at 0.5, where no one knows which protocol will win, Katyal says the smart play is not betting on one protocol. She can deliver them all.
a witness Complete conversation To learn more about
- Why should the math of “0.5 to 1.5” change the way you think about timing?
- Why AI shopping needs better product data before it needs better AI.
- How to experience proxy trading without marrying one platform.
- Why trust, not transaction volume, is the currency that matters now.
