Agentic commerce, AI, and where payments really makes its money: Geoffrey Barraclough

Three of the loudest stories in payments right now are agentic commerce, artificial intelligence, and stablecoins. Geoffrey Barraclough writes about all of them. He will also tell you, cheerfully, that most of the money in payments is still being made somewhere a lot less exciting: in small businesses, in bundled software, and in the unglamorous work of keeping a merchant happy enough that it does not leave.

That gap, between what the industry talks about and where it actually earns, is more or less his beat. Barraclough spent a decade inside Europe’s merchant acquiring industry before starting The Business of Payments, the newsletter and blog he writes under a tagline that doubles as a worldview: how paytech makes money. An economist by training, he reads the sector the way an economist would, by following the cash rather than the press releases. Ahead of his appearance at Payments Unleashed EMEA, he walked through where the money is moving, and where the story is moving faster than the revenue.

The part he is most certain about is the structure of the acquiring market. A small group of modern processors is consolidating the market, and he does not find that surprising in the least. Players like Stripe and Adyen built unified payment platforms from scratch, he says, while the established names are still spending heavily to knit together the acquisitions they made in the boom years.

The model feeds itself. A single platform, one contract, and global customer service pull in large enterprise merchants. Being developer-friendly out of the box pulls in the software vendors who, in turn, distribute payments into thousands of small and mid-sized businesses. And from the very start, the modern processors treated customer experience and brand as things worth investing in, where, in Barraclough’s assessment, the established players have more often treated brand and marketing as a secondary concern.

Consolidation is not only a modern-processor story, though. The established players are also merging and refocusing after years of acquisitions, so this is less old versus new than it can look, and the real question, in Barraclough’s framing, is whether all that scale makes the platform simpler for the merchant or merely bigger. Can the established players catch up on the experience? It will be hard, he says, but not impossible, and he is clear about where he would put the effort. Ruthless platform consolidation first, to drive down the cost to serve and make sure new product development reaches every customer rather than only the newest ones. Then customer service, because there is a real gap in the market. The global champions, he notes, are built around the needs of the very largest merchants, which leaves mid-sized and large domestic merchants under-served and worth winning. And then brand, built on an actual statement of what the company stands for.

Barraclough’s verdict on the idea getting the most airtime is blunt.

The enthusiasm is real, he allows, but it is enthusiasm for the wrong part. People happily use chatbots to search and to get product recommendations. Handing a bot their card details and telling it to go and book a holiday is a different matter, and very few genuinely autonomous payments are being made without a human somewhere in the loop. He does not expect that to shift much over the next year. Some large merchants have blocked or restricted bots from their sites, the competing agentic-commerce protocols are too confusing for most online retailers to adopt, and shoppers remain nervous about trusting software with their cards. That reading has aged well. OpenAI’s Instant Checkout brought purchases into ChatGPT through partners like Etsy and Shopify, then ran into the limits of the model, uneven merchant adoption and fee friction, and was retired in March. Visa’s newer move to bring its network into ChatGPT shows how fast the infrastructure is forming, but it makes Barraclough’s larger point for him: the hard part is trust, approval, liability, and fraud control, not putting a checkout button inside a chatbot.

He thinks agentic commerce may run through the hype cycle faster than anything the industry has seen, and tip into disillusionment soon. That, in his view, is no bad thing, because it buys the industry time to solve the problems that actually matter. Three of them stand out. There has to be a way to prove what the shopper intended: an auditable record of what they told the bot to buy, on what terms, and confirmation that the bot bought exactly that. There has to be an agreed answer on liability when a bot buys the wrong thing. And merchant anti-fraud has to be rethought, because in the wrong hands this technology could enable social engineering and friendly fraud at industrial scale. Only validated bots can be allowed to transact, which means, as he puts it, the industry will urgently need a Know Your Agent framework to sit alongside the Know Your Customer rules it already runs.

His less dramatic point is that none of this requires tearing up the system. The existing card payment infrastructure can carry agentic commerce, he says, provided everyone involved is willing to cooperate.

Ask what is delivering a return on AI and what is not, and Barraclough is candid that it is an easy question to ask and a hard one to answer honestly. The industry has used AI and machine learning for years, notably in fraud screening. The large language models everyone now means when they say AI are only a few years old, and have been running inside real businesses for less time than that.

He is skeptical of the loudest claims. For all the headlines, he sees little firm evidence of an AI-driven jobs collapse. Companies announce layoffs constantly, and pinning them on AI productivity rather than the dozen other things that move corporate performance is mostly guesswork. He is just as skeptical of the idea that buying everyone a Copilot license will lift output. It is a general-purpose tool, and handing it to staff untrained tends to produce an endless supply of plausible, low-value prose that other people then have to read. The value comes from training, or better still from packaging the technology into vertical or task-specific applications that solve a real problem.

Two things, though, do interest him. The first is that AI can write code, which he expects to be a genuine productivity boost across the industry, with one warning: people still have to know what they are doing, because non-technical people producing bad code in one corner of a company can quietly drag down productivity in another. The second is that AI is making it dramatically cheaper to start a business. A founder can now stand up a technology company with a single developer, and lean on AI for the legal, accounting, and even early marketing work that used to mean hiring. AI, in his reading, will reward the proactive and the curious, whether they work for themselves or for someone else.

He applies the same filter to stablecoins, describing them plainly: a form of private money, usually dollar-denominated and backed by reserves such as cash and short-term government debt, most useful to people in countries that lack good access to dollar banking.

That is where the real case sits. Moving money in and out of somewhere like Argentina or Nigeria through correspondent banking is slow and expensive, and stablecoins promise near-instant settlement at low cost. Along the well-trodden corridors, though, the argument thins out. Moving sterling to dollars is already fast and cheap, and a stablecoin does not change the part of the job that actually costs money. Most of the expense in international payments is the compliance burden, the checks that make sure a provider is not serving criminals or terrorists, and that burden does not vanish because the rail is USDC. Stablecoins cut settlement friction. They do not cut the obligation to know who you are dealing with.

So he sees the demand showing up exactly where the theory predicts. Merchant acquirers are hearing requests for stablecoin settlement from customers in countries with high inflation, exchange controls, or other reasons that hard currency is hard to move. Shoppers in Europe or North America paying for their shopping in stablecoins, on the other hand, remain a long way off.

For all the talk of enterprise deals and new technology, Barraclough keeps returning to a fact he finds slightly counterintuitive about his own industry: most of the money in payments is made from small businesses. The encouraging news for providers is that small businesses will pay for payments when they arrive inside a bundle that runs the shop, good hardware paired with genuinely useful vertical software, an ePOS system or an eCommerce platform that makes the day easier. The implication is straightforward. Acquirers need to own that software themselves, or partner with the software vendors that have a winning product, and ideally do both.

There is a second, newer revenue line on top of it. Small businesses are increasingly willing to borrow from whoever provides that bundle rather than from a bank, which hands payment providers an embedded-finance income stream that deepens the relationship and makes customers far less likely to leave.

Enterprise is a harder game. Competition has squeezed processing fees, but Barraclough does not think value has disappeared from the top of the market, only moved. Large merchants will still pay for higher authorization rates, lower fraud, and the ability to operate across borders. And he makes a point that sounds almost old-fashioned next to the talk of agents and tokens: great customer service still matters when product gaps open and the feature race tightens, and it keeps clients who might otherwise look elsewhere. In payments, as in most things, people stay where they feel looked after.

Geoffrey Barraclough joins Mainstage Panel 3, AI in payments: optimization or transformation, at Payments Unleashed EMEA. The event opens with an evening reception on 29 June at 12th Knot, Sea Containers, and continues with a full day of content on 30 June at the Hilton London Bankside, bringing senior payments leaders from across EMEA together for sessions on real-time payments, fraud and scam liability, sovereignty in European payments, and the economics of where merchant payments go next.

If you spend your days being told what payments will look like in ten years, an afternoon with someone who tracks where the money is moving today is a useful corrective.