The fundamental concepts of online commerce are starting to break down, said Paul Conroy, CTO at Square1, as he looked back at last week’s Stripe Sessions in San Francisco.
Stripe bills Sessions as its “internet economy conference.” For a few days in San Francisco, thousands of people from all over the world gathered last week to talk about the future of Internet commerce.
But for all the product launches and big-name keynotes, one important change keeps emerging: the fundamental assumptions behind online trading are starting to break down.
For over 20 years, payment systems have been built with the assumption that bots are a problem. The ideal customer browses, hesitates, clicks, and finally makes a purchase. The suspicious customer lands directly on the payment page, gives almost zero behavioral signal, and comes from a server farm instead of a smartphone.
Stripe Sessions 2026 made it abundantly clear: that assumption is dead. In the next stage of trading, it is possible that it is a bot is something the customer.
Agents need salespeople who understand them
One of the clearest examples of this change is the soon-to-be idea of asking an AI agent to buy something for you. Not just “find me this jacket”, but something like a concierge: “Find me the perfect outfit for hiking in France in July, within this budget.”
That request asks a lot more of the seller than a typical product search. One can glance at the product page, read the missing information, decide whether the two items might work together, and gauge if the return policy sounds right. The agent needs that information in an organized, reliable format. It requires understanding sizes, materials, compatibility, and, most importantly, whether the seller can be trusted.
For sellers, agency marketing raises a practical question: can your products, prices, and policies really be understood by machines?
This is why new trade agreements are suddenly so important. The Unified Commerce Protocol (backed by companies like Stripe, Shopify, and Google) is an attempt to standardize how this should work. If agents are going to buy, sellers need a common way to tell those agents what they are selling and how it can be bought. Businesses with dirty product data will soon find themselves effectively invisible to customer devices.
AI’s New Economic Unit
This trend is also evident when we look at agents who pay for digital work in small amounts.
One demo at Sessions involved a code review tool that charges based on tokens used. That sounds niche until you consider the economics of AI more broadly. As more companies rely on AI, the cost of guesswork becomes a major operational risk. We’ve all seen the funny screenshots where someone convinces a fast food chatbot to ignore the menu and write a React app instead. That unintended use is funny until a service is used that has real cost thinking behind it. If the use of spikes, it costs a spike.
In the demo, the price of the tool was one thousand cents for each token used. That’s too little to make sense with regular credit card processing, so late processing charges are common, albeit dangerous, for this type of merchant today. However, if an agent can call an API, use an authorized wallet, and make thousands of micropayments when they use the service — while keeping processing fees low — functional microtransactions suddenly look real.
How do you charge for native AI services when the unit economics are too small, too fast moving, or too risky for traditional payment models? This is where stablecoins graduate from crypto buzzword to functional infrastructure.
A view from Europe
Spending a few days in San Francisco makes a difference of pace that is hard to ignore. Coming from Dublin, where bus shelters are likely to sell phone plans or supermarket specials, it’s surprising to arrive somewhere where every billboard seems to advertise a novel AI startup, or a company with a new way to transfer money.
Some of that is definitely hype. But what is absolutely true is that the US is busy putting together the infrastructure to support these changes. Stablecoin adoption and agent wallets are moving quickly from theoretical concepts to live commercial deployments.
From a European perspective, that should make us a little uncomfortable. We tend to approach new financial infrastructure with regulation first. The release of the MiCA (Markets in Crypto-Assets) framework is a perfect example. While it provides Europe with much-needed regulatory clarity, our focus on compliance often means that commercial shipping lags behind.
Consumer protection and stability are important, of course. But there is a difference between moving carefully and moving so slowly that the next generation of infrastructure is built elsewhere, with someone else’s interest at heart. If native AI commerce, agent wallets, and real-time stablecoin microtransactions become the basis of how online commerce is done, Europe cannot watch from the sidelines. The challenge is good control without late control.
The arms race for fraud is intensifying
The angle of fraud is where this natural process becomes more complex.
Historically, using fraud tools has automatically treated bot-like behavior as suspicious. No normal browsing pattern, one quick request to make, and a data center IP were strong signs that something bad was going on. In the commercial world of the agency, a legitimate transaction will look exactly like that.
This creates a catch-22 for marketers. Block good agents, you lose income. Allow bad agents, you lose money. Old signs fail in both directions.
This came up several times during the Session. There’s a new arms race developing: fraudsters using AI to measure attacks and investigate vulnerabilities, and Stripe using its AI models in Radar to detect and respond. What I found most interesting was the frankness of most of the speeches. There was no victory, just a lot of, “we haven’t fully figured this out yet”. How do we verify intent? Who owns the activity when the user sends a decision to the agent?
These are the questions that exist for businesses that operate on low margins. The same automation that makes new experiences possible makes abuse less expensive to try.
Clean up APIs and the human element
Between lectures in the main hall, instead of piped background music, a live string quartet played pop covers. It sounds like a small thing, and it won’t show up in anyone’s ROI model, but it was a smart decision made by someone somewhere at Stripe, to make the room a great place to be.
That theme of the hidden uses of beauty came up during Patrick Collison’s interview with Sam Altman. Altman noted that Stripe has taken an almost absurd level of care about design and aesthetics in its APIs for years. That aesthetic consistency was designed to appeal to human developers, but ironically, it may be their biggest advantage in the world of agents. Agents, it turns out, benefit from the exact same things that human developers do: clear APIs, consistent abstractions, and predictable behavior. Stripe has spent years making it easy for developers to choose, putting them in a surprisingly strong position as software also begins to choose tools.
In that same conversation, there was an interruption when a protester carrying a guitar walked around singing saying that music and art should be made by people, not machines. It was a strange and funny moment. The acoustics of the Moscone Center are so good many thought he was part of the show at first, before he was quickly escorted out! There were plenty of callbacks to this during the discussions that followed – John Collison noted that the AI demo that was taking too long to run could have used a guy with a guitar to cheer people up – but it was another reminder that AI is changing more than commerce; it is in direct conflict with culture at large, both good and bad.
The future is distributed unequally
For visitors to San Francisco, Waymo’s self-driving cars roaming the hills still feel like a tourist attraction of the future. For locals, they are a lot of traffic.
Agent Commerce feels very similar to those self-driving cars. It reminds us of William Gibson’s famous line about the future already being there, just not distributed equally.
While the agent trade is evenly distributed, there is a lot more here. Businesses that prepare now by cleaning up their data, rethinking their microtransaction rates, and strengthening their fraud controls will be ready for a new type of customer.
Those waiting can find agents who have already learned to shop elsewhere.
Paul Conroy is the CTO at Square 1an award-winning digital transformation agency specializing in online payments and publishing. He was also among the first group of Stripe Partner Advocates – a group of technology leaders with deep experience in payments, chosen to work directly with Stripe’s product teams. Disclosure: Square1 is a longtime editor of Silicon Republic.
Don’t miss out on the information you need to succeed. Sign up for Daily BriefSilicon Republic’s digest of must-know sci-tech news.
