By Xavier Sheikrojan, Director, Risk Intelligence at Signifyd

Retailers have spent decades refining the online customer journey, optimising landing pages, streamlining checkout and layering in just enough friction to manage risk without losing conversions. But what happens when the customer no longer takes the journey at all?

That’s the shift agentic commerce represents. AI agents are beginning to act not as assistants but as autonomous actors, capable of initiating and completing purchases independently of the shopper.

With major platforms now enabling agent-led purchases, more transactions are taking place outside the traditional storefront and often without the retailer having clear visibility into how the purchase was made.

It tends to be a quicker, more convenient experience for the shopper, but it also raises some tricky questions.

When an AI agent handles the purchase, who’s responsible for verifying the payment? Who gets the credit for the sale? And who’s responsible if something goes wrong? That means less oversight for retailers, as well as a growing risk of fraud and disputes.

Agentic commerce isn’t a thought experiment. It’s an active disruption. And as the line between shopper and system continues to blur, retailers can adapt now before they’re left out of the transaction altogether.

When the buyer skips the store

In traditional ecommerce, the retailer owns the customer journey. From first click to final checkout, merchants can observe behaviour and influence decisions in real time. Agentic commerce shows to be removing that visibility; the AI agent becomes the shopper’s proxy, operating independently to search, compare and transact, often without the merchant ever seeing a session.

This creates a fundamental disconnect. Orders still arrive but stripped of the signals retailers rely on to assess legitimacy or optimise performance.

In some cases, especially when the checkout happens entirely offsite, there’s no device data, no behavioural footprint and no marketing attribution. The order arrives with little context, making it harder to assess intent or risk. The product may have been selected via an interface the merchant doesn’t control. Even the checkout might be completed offsite, using saved credentials unknown to the retailer.

The technology here is nascent but the implications promise to be broad. Fraud detection becomes harder when there’s no discernible shopper behaviour.

When agents strip out key details – sizing, materials or delivery terms, for instance – customers are more likely to be disappointed with the final product they receive, leading to more returns and a higher chance of disputes.

At the same time, merchants may lose out on the ability to shape the buying experience or explain how a purchase came about. Without important context clues like these, it’s harder to spot suspicious activity or apply the right checks before an order goes out.

Autonomy without accountability

Agentic commerce promises to streamline the path to purchase but for retailers, it’s a path paved with blind spots.

When an AI agent takes over the transaction, the retailer doesn’t just lose visibility, they also can lose control over how their brand is represented, how their policies are applied and how the buyer’s identity is verified.

This matters because much of today’s fraud prevention is rooted in context. Understanding intent, spotting inconsistencies or flagging buyer anomalies all depends on access to behavioural signals that disappear in agent-led flows. Retailers are instead left approving orders they can’t authenticate and defending transactions they didn’t directly facilitate.

The risks extend past fraud. When agents misinterpret product details or apply out-of-date discounts, the result is a spike in returns, policy disputes and margin loss. If customers don’t recognise the transaction or the merchant’s name, chargebacks are likely to increase. And with no clear record of user consent, liability becomes harder to assign.

What’s still missing is a verification layer designed for non-human buyers – a way to confirm that the agent is trusted, the shopper has authorised the purchase and the information presented was accurate.

Until those safeguards are in place, agentic commerce will continue to expose retailers to a growing set of vulnerabilities with fewer tools to manage them.