Back
Key takeaways:
- Europe isn't behind on agentic AI, it's solving trust, consent, and liability before allowing machine-led purchases.
- In agent-led shopping, structured product data and system accessibility matter more than website design or branding.
- New performance metrics, time to purchase, product accessibility, and trust, are replacing clicks as the KPIs that count.
The early narrative around agentic commerce is simple: the US is moving fast, Europe is lagging. It is also wrong. Europe is not behind on agentic AI. It is attempting to solve a harder problem first: how to let machines transact on behalf of consumers without breaking trust, liability or control.
That distinction matters because agentic commerce is not a user experience upgrade. It reshapes how purchasing decisions are made, who makes them, and what infrastructure retailers need to compete.
The real divide is not capability. It’s permission
In the US, consumers can already complete purchases directly through large language model interfaces. The agent can search, decide and transact in one flow, even if access is still limited.
In Europe, that flow breaks at the point of payment. Transactions still have to move through merchant-controlled environments with explicit authorisation and regulated payment methods.
The difference is not technical. It is regulatory. Europe requires explicit consumer consent for agent-led transactions and is still working through questions such as liability: who is responsible when an agent makes the wrong purchase or a transaction fails.
The US model is to deploy first and resolve disputes afterwards. Europe is doing the reverse. This slows adoption, but it forces something the US has deferred: a workable trust model for machine-led purchasing.
Agentic commerce quietly removes the website from the centre
Most retailers are approaching agentic AI as an interface problem. That is the wrong place to focus.
In an agent-led purchase, the decision is no longer driven by browsing or merchandising. It is driven by structured criteria. An agent evaluates thousands of products against specific requirements and selects the closest match.
This changes the role of the website. It no longer needs to guide a human to a decision. It needs to expose data that a machine can query.
The practical consequence is uncomfortable. The competitive advantage shifts away from design, content and persuasion, and towards data structure and accessibility. Retailers with well-organised product data, clear attributes and real-time availability will be preferred by agents. Those without will simply not be considered.
This is why the most important system in agentic commerce is not the website. It is the combination of CRM and Order Management System that holds and organises the product and fulfilment data.
Payments become infrastructure, not a moment
The same shift applies to payments. In a human-led journey, payment is a visible step. In an agent-led journey, it has to disappear into the background.
In Europe, that only works if consent and security are preserved. Tokenization becomes central because it allows pre-authorised, repeatable transactions without reintroducing friction every time an agent acts.
Without it, the flow breaks back into manual authentication, which defeats the purpose of delegation.
This is not a marginal adjustment. It turns payments from a discrete interaction into persistent infrastructure that agents can use repeatedly and safely.
The first benefit is not automation. It is conversion
Much of the discussion around agentic AI focuses on efficiency. The more immediate impact is much simpler: fewer abandoned purchases.
A significant proportion of shopping journeys fail because the customer cannot find the right product. Agents address that directly by refining intent and matching it to inventory in real time.
That is where the commercial case starts. Not in replacing people, but in converting demand that currently falls away.
Over time, agents also introduce continuity. They remember context, anticipate needs and return with relevant options, turning one-off purchases into repeat behaviour.
New KPIs expose what really matters
If agents take over discovery and decision-making, traditional metrics such as clicks become less meaningful.
Three measures start to define performance instead.
- Time to purchase shows how quickly an agent can resolve a need.
- Product accessibility reflects how visible and usable a retailer’s inventory is to machine decision-makers.
- Trust becomes decisive because agents will deprioritise retailers that fail to meet expectations on fulfilment or product accuracy
Trust, in this context, is not branding. It is operational reliability encoded into data.
The uncomfortable conclusion
Retailers looking at agentic AI are asking how to redesign their websites or add conversational interfaces.
The more relevant question is different: can your systems expose the right data, in the right structure, with enough reliability that a machine will choose you without ever seeing your brand?
Europe’s slower path makes this clearer. Once agent-led transactions are fully permitted, the winners will not be those who moved first on front-end experiences. They will be those who quietly rebuilt their data, payments and fulfilment systems to work in a world where the customer is no longer the one doing the shopping.
Planet newsletter
Built for your industry. Focused on your challenges. Delivered to move you forward
We respect your privacy. Unsubscribe at any time.
