Agentic Commerce at 0.5: Why Infrastructure Beats AI

Adyen's Karan Katyal put agentic commerce at 0.5 on a five-point scale and at 1.5 a year from now. That second number triples where we are and still lands nowhere near the middle.

Consumers already knock at the door. AI is great at helping people find things to buy. It is terrible at letting them actually buy them.

The hard part was never the AI. It is the plumbing underneath: machine-readable catalogs, trust, fraud, liability, and a pile of competing protocols nobody has reconciled yet.

The Infrastructure Bottleneck: Catalogs, Trust, and Fragmentation

Katyal, global head of agentic commerce at Adyen, laid out the gap in an interview with PYMNTS. A traditional product listing needs maybe a dozen attributes to show up on a webpage. An AI agent trying to tell two similar products apart needs far more. Detailed specs, reviews, context -- the stuff a human reads between the lines. Most catalogs simply do not carry that. Making them legible to machines is a project, not a tweak.

“Infrastructure is a much bigger block than folks thought about, perhaps,” Katyal said.

Then there is trust and liability. “Trust and risk and fraud all need to be at the center of the foundation of whatever you end up building,” he said. Not bolted on at the end. Because the moment an agent acts on your behalf, the questions get more precise. Who is liable when the agent books the wrong flight? A botched checkout is annoying. A botched delegation is a dispute about responsibility.

Part of why agentic commerce sits at 0.5 is that it is not a single product anyone can ship. Katyal described four models: human-in-the-loop, autonomous buying with rules, machine-to-machine transacting, and business procurement. Four directions, four sets of standards, no referee.

“Fragmentation will continue for quite some time,” he said. “Agentic commerce is really almost like a direction of travel.”

Adyen Agentic: Rationale and Read-Through

Against that backdrop, Adyen rolled out Adyen Agentic, a modular bundle of product discovery, checkout orchestration, and payments. It helps merchants make product data machine-readable, keep existing checkout and fulfillment systems, and extend authentication and fraud controls into agentic environments without forcing a rebuild for every new platform.

In a world stuck at 0.5, where nobody knows which protocol wins, the smart play is not betting on one. It is being able to plug into all of them.

Katyal expects the low-drama purchases to lead: repetitive household restocks, low-consideration buys where handing off the task removes friction without making anyone nervous. PYMNTS CEO Karen Webster added a twist: AI may be just as valuable at the top of the consideration curve, on the big researched purchases, where the real payoff is gathering and comparing, not the click that ends it.

Both agreed there will not be one universal way to shop. It will bend to the person and the moment.

Katyal was blunt about the revenue timeline. “Agent commerce is almost like a series of experiments,” he said. “If the thing that you need to justify this is hard numbers, now is not the time for it yet.”

Low maturity is not permission to wait. Webster pointed to the lessons of the digital and mobile shifts: businesses that dragged their feet spent years clawing back ground. The consumers are knocking. The path from intent to purchase is broken. Merchants who fix their half of that path now, during the experimentation window, are the ones who will be ready when the loop finally closes.

The TCEHY (Tencent Holdings Limited) stock page carries an Alpha Score of 47, a mixed label. Tencent, which recently launched a new payment app for overseas visitors, sits at the intersection of cross-border commerce and AI infrastructure -- a helpful proxy for the sector's readiness gap.