If AI makes the decision before the visit, what happens to checkout?

We’ve spent the better part of two decades refining checkout.

If you’ve worked in digital, product, or ecommerce, you’ve likely been part of that effort. Reducing friction, simplifying forms, improving payment speed, increasing conversion rates. It’s been a consistent, shared focus across industries. And for good reason, checkout has traditionally been the moment where intent turns into revenue.

But I’m starting to think we’ve been solving the wrong problem.

Not because checkout optimisation hasn’t been valuable, it absolutely has. But because the conditions that made checkout such an important moment are starting to change.

AI agents are beginning to alter how decisions are made online. Not in a distant, hypothetical way, but in a very practical sense. These systems don’t browse in the way people do. They don’t open multiple tabs, get distracted, compare endlessly, or abandon carts. They operate from a different starting point: defined intent.

When someone uses an AI agent to “find the best option and complete the purchase within these constraints,” the decision-making process happens before any website is visited. The agent arrives with context already established, preferences, budget, timing, and a clear outcome in mind.

At that point, the role of a website shifts. It’s no longer the place where a customer is persuaded or guided toward a decision. It becomes the place where a decision is carried out. This marks an important change.

It means that many of the things we’ve optimised for; navigation, layout, calls to action, may matter less in certain scenarios because they’re no longer part of the critical path to purchase.

Instead, what matters is whether your product or service can be understood, evaluated, and acted on by an agent. This is where the real shift sits.

We’re moving from designing experiences for people moving through a journey, to enabling systems to execute outcomes on their behalf. This will have implications well beyond checkout.

For example, in an agent-driven interaction, the payment step doesn’t sit at the end of a funnel. It’s embedded within the task itself. Payment authority travels with the instruction, rather than being triggered by a button click. The agent doesn’t need to be “converted” in the traditional sense, it needs to be able to complete what it was asked to do, within the rules set by the user.

That changes how we think about conversion entirely.

It also raises some bigger questions that I don’t think we’re fully grappling with yet.

If intent is formed before a customer reaches you, where does influence sit?

If an agent is making the comparison, how do you differentiate?

If the interaction is machine-to-machine, what replaces the experience we’ve spent years refining?

I don’t think this means checkout disappears overnight, or that human-led journeys become irrelevant. There will be a long period where both models exist side by side. People will still browse, explore, and make decisions themselves. But alongside that, we’ll see a growing number of interactions where the “journey” is compressed into a single instruction, and everything else happens in the background.

Because if it plays out the way it’s starting to, then improving checkout isn’t where the next wave of value will come from. It will come from understanding how your product, your data, and your capabilities show up when an AI is the one doing the work.

I’ve been exploring this more deeply, including what it means for payments and the role things like cards play behind the scenes in enabling this model.

If you’re working in digital or thinking about how AI will actually change customer behaviour, it’s worth paying attention to now rather than later.