Who's Actually Shopping? Specno on the Infrastructure Agentic Commerce Still Doesn't Have

Venture money is flooding into AI shopping agents. Payment networks, browser makers and AI labs are racing to define how autonomous agents will shop, compare and buy. Specno, the product consulting firm that helps companies de-risk complex software builds, says the infrastructure behind that race isn't keeping pace with the capital. Trust, verification and accountability are still unsolved, and the technology is moving faster than the answers.

"Hey Agent, find me the cheapest one"

The shift Specno expects: people stop browsing and start delegating. Instead of opening an app, a consumer tells an agent what they want. "Find me the cheapest kettle." "Match these blinds to this photo." "Furnish this room to this budget." The agent searches, compares and buys. The person never touches a storefront.

"The interface of the future isn't a search bar. It's a conversation," said Joshua Harvey, CEO of Specno. "People are going to stop browsing and start delegating. They'll tell their agent what they want, show it a photo, give it a budget, and expect it to come back with the right thing already found or bought. That's a different relationship between a brand and a customer, and most retailers haven't touched their systems to account for it."

SEO is being rewritten for machines

If agents do the shopping, retailers need to be visible to agents, not just to people. That means structured, machine-readable product data, pricing feeds that update in real time, and content built for an AI working through hundreds of options in seconds, not a person scrolling a page.

Ranking well in human-facing SEO today guarantees nothing with an AI agent evaluating on different criteria. Specno expects a new discipline, agentic SEO or answer-engine optimisation, to matter as much to commerce as search ranking has for the last 20 years.

Who's liable when the agent gets it wrong?

Then there's the harder problem nobody has answered. If a shopping agent gets fooled by a fake storefront and completes the purchase, who's responsible? The consumer? The agent developer? The payment provider? The platform that surfaced the fake listing?

"Right now, if someone gets scammed online, there's a clear chain. You know who clicked the button and who authorised the payment," said Harvey. "Once an agent makes that call for you, the chain breaks. Did the agent fail to verify the site? Did the marketplace fail to flag it? Did the buyer even see what was purchased before it happened? Nobody has answered that, and it will matter a lot more once agents are transacting at scale, not just browsing."

Specno believes the winners in this category will treat verification, fraud detection and liability as core product features from day one, not something patched on after agentic transactions go mainstream.

"This isn't a problem you solve with a bigger team," said Harvey. "It's a problem you solve with the right specialist who holds the technical, commercial and regulatory picture at once, moving fast enough to build the trust layer before the fraud does. That's the model we've built Specno around, and it's exactly the kind of build this category needs right now."

Specno is already helping retailers and product teams get ready by structuring product data for AI discovery, designing conversational and visual-search experiences, and building the verification and trust layers agents need to transact safely.