When shopping stops being a decision

Shopping has always involved a series of small decisions. What to buy, when to buy it, whether you need it at all. Even when those choices are influenced by recommendations or marketing, they still feel like yours. That’s starting to change.

A new wave of AI systems is being built to do more than suggest products. They’re designed to anticipate needs, make decisions, and in some cases complete purchases on your behalf. Not “you might like this”, more “this has already been handled”. At first glance, it sounds convenient, but it also raises a different set of questions.

When suggestion becomes action

Traditional recommendation systems sit at a distance. They surface options, but the final decision still sits with the person. Agentic systems move closer to the point of action. They don’t just recommend a product, they decide when it’s needed, select an option, and follow through. That might mean reordering household items, booking services, or making purchases without a clear moment where you decide. The shift is subtle, but important, because the question is no longer just what should you buy, it’s whether you’re still the one deciding.

Where shopping now happens

At the same time, the idea of a “shop” is becoming less fixed. Retailers are no longer just competing through their own websites or physical stores. Increasingly, discovery is happening through AI tools, social platforms, and recommendation systems. John Lewis, for example, is investing in ways to meet customers in these environments, rather than relying on its own storefront as the main entry point.

That shifts the balance of control. If AI systems decide what gets surfaced, compared, or recommended, the “shop window” isn’t owned by the retailer anymore. It’s shaped by the system in between.

When it doesn’t quite work

These systems don’t always behave as expected. AI stylists can generate outfit ideas that look plausible at a glance, but fall apart on closer inspection, sleeves that don’t attach properly, buttons that lead nowhere, fabrics that don’t quite behave like real materials. Recommendation tools can surface irrelevant products, or oddly specific items that don’t match what was asked for. The same pattern shows up elsewhere too, suggestions that sound confident but don’t quite fit the situation. It’s a reminder that while these systems can act, they don’t always understand.

The quiet trade-off

Part of the appeal is clear. Less effort, faster decisions, fewer things to think about. But convenience isn’t neutral. To act on your behalf, a system has to interpret your behaviour, predict your needs, and decide what counts as the right outcome. That introduces a layer of judgement that sits somewhere between you and the result. In some cases that might be helpful. In others, it raises a more subtle question.

What happens to choice

If a system starts making decisions before you’re fully aware of them, what happens to the act of choosing? You can still override, cancel, or adjust, but over time the default shifts from actively deciding to passively accepting. That doesn’t mean people stop making choices entirely, but it does change where those choices sit and how visible they are.

Because the question isn’t just whether AI can shop for you. It’s how much of that process you’re comfortable handing over.