Somewhere in Britain right now, a customer is asking an AI assistant to find a pair of running shoes under £100, in their size, delivered by Friday. The assistant is comparing retailers, checking stock and, increasingly, completing the purchase.
The retailer on the other end may have no idea a machine just bought from them.
This is agentic commerce: AI agents researching, choosing and buying on behalf of people. It is no longer a forecast. A survey by The Payments Association of 100 senior finance, payments and risk leaders at UK online retailers found that 58% believe AI-initiated transactions have already reached their platforms, and 72% are preparing or planning their approach. Yet only 41% feel very confident in current liability frameworks (Retail Gazette).
The Association's chief executive, Emma Banymandhub, put it simply: agentic commerce is "arriving faster than many merchants expected."
Three gaps that matter
Liability. When an agent buys the wrong thing, buys too much or is manipulated into a purchase, who pays? The survey makes the confusion plain. Asked about a disputed £2,000 agent purchase, 24% said it depended on the circumstances, 21% backed shared liability and 18% said the AI vendor should be responsible. There is no settled answer yet. The Financial Conduct Authority is reviewing how existing rules apply to agent-initiated payments.
Fraud detection. Most fraud engines are trained on human behaviour: how fast people type, how they move through a site, what time of day they shop. Agents behave nothing like that. As Banymandhub warned, "fraud engines built around human behavioural signals are misfiring on legitimate agent traffic." That means two costs at once: genuine sales being blocked, and real fraud potentially hiding inside agent traffic.
Visibility. Many of the identifiers and protocols that would let a merchant recognise an agent-initiated transaction are not yet widely available in the UK. If you cannot see it, you cannot measure it, price it or protect it.
Agents are a new kind of customer
Here is the commercial twist. Agents do not browse. They do not see your homepage hero, your brand campaign or your carefully designed product photography. They read data. They compare prices, availability, delivery promises and returns terms, and they pick the offer that best matches their instructions.
That shifts competitive advantage towards retailers with clean, structured, machine-readable data. Logistics provider nShift argues that delivery options, cut-off times, collection points and returns eligibility all need to be expressed in a structured, consistent form, and that the accuracy of delivery promises becomes a selection signal in its own right: an agent comparing offers will learn to avoid retailers that miss them (nShift). Emerging open standards such as the Universal Commerce Protocol aim to make that kind of agent-to-retailer interaction consistent.
In other words, your product feed is becoming your shop window.
What to audit this quarter
The Payments Association's working group urges retailers to audit fraud systems, review internal controls and check payment provider readiness. In practice, that translates into a short, focused list:
- Measure agent traffic. Work with your fraud and analytics teams to estimate how much of your traffic and how many of your declines may involve agents.
- Write a policy. Decide how you will treat agent-initiated orders, disputes and returns. The survey found merchants with formal policies were nearly four times more likely to feel confident about liability.
- Fix your data. Audit product, pricing, stock, delivery and returns data for completeness and consistency across channels.
None of this needs a big programme. It does need someone to own it.
The opportunity hiding in the risk
It is easy to read agentic commerce as a threat. It is also a chance to win customers who never visit your site in the traditional sense. Retailers that make themselves easy for agents to understand, trust and transact with will pick up demand that their competitors never see.
The rules will catch up. The Payments Association is calling for common standards on authentication, accountability and consumer protection within 12 months. The retailers who prepare now will not need to scramble when they arrive.
Our Experience and Engineering teams help retailers get their product data, checkout and fraud tooling ready for a world where some customers are machines. If you want to know how agent-ready you are, talk to us. For the security side of agents, read our note on securing AI agents that can act.
Frequently asked questions
What is agentic commerce?
Agentic commerce is when AI agents research, compare and buy products on behalf of people, often completing the purchase without the person visiting the retailer's website.
Are AI agents already buying from UK retailers?
A Payments Association survey found 58% of senior payments and risk leaders at UK online retailers believe AI-initiated transactions have already reached their platforms.
Who is liable when an AI shopping agent makes a mistake?
There is no settled answer yet. Retailers are divided, and the Financial Conduct Authority is reviewing how existing rules apply to agent-initiated payments.