John Gibbard

What happens when the system finally learns to listen.

Last week in Part One, I described why Christmas shopping feels hostile, why even the most basic purchase turns into a strange performance of archaeology, jargon and filters masquerading as understanding. The real problem wasn’t the products but the machinery. The fiction that a PLP grid is somehow an acceptable translation layer between human intent and retail stock.

This week is the other half of the story: the thing that replaces it.

Because the truth is, we’ve spent twenty years designing for systems that never deserved that level of obedience. We pretended the homepage was the grand entrance, the digital lobby with its scented candles and seasonal banners. We treated it like the flagship store: polished, high-stakes, endlessly debated at internal stakeholder meetings. Meanwhile, almost no one arrived through it, or if they did, they were there for a split second. Most people dropped in sideways, via Google, a WhatsApp link, an email, or a moment of panic at 11 p.m. The homepage was the UX and UI theatre we performed for ourselves and our clients.

Agentic systems make that fiction impossible to sustain. They don’t care about your reception desk and your neatly prioritised way finding. They don’t even see it. They take what you mean, “something thoughtful, about forty quid, she hates clutter, nothing scented” and drop you straight into the one, tiny corner of the site where the decision will live or die. A place that, inconveniently, most retailers still treat as a functional afterthought: the product-detail page.

The PDP becomes the real front door because in an agentic journey the start isn’t a place, it’s a sentence.

This is where that old inventory-obsessed model buckles. Catalogue commerce was built on the premise that customers begin at the top and drill down. Agentic commerce begins at intent and works sideways. The sitemap is your fiction, not theirs. The system no longer needs your categories. It needs your clarity.

Be under no illusion though, this ain’t easy. This only works if the agent can explain itself. When a system gives you two options instead of two hundred, you need to know why. Not academically, emotionally. Why this jacket and not the other one? Why this feels like her. Why this fits your mental model of who she is. The explanation is the reassurance loop. Without it, the whole thing becomes another opaque machine; efficient, yes, but untrustworthy in all the ways that matter.

And then there’s the serendipity problem. Efficiency is addictive, but clinical. If we strip out every detour, we drain the pleasure along with the friction. The answer isn’t a return to the grid; it’s controlled looseness. A suggestion or two just off-axis. Something adjacent. Not twelve rows of “you may also like” tat, just enough to keep the experience human. Discovery without the search-and-filter trauma.

None of this is a theoretical exercise for me. I genuinely spent years trying to push natural-language intent into car retail at JLR, long before the technology was mature enough to meet the ambition. I saw how people really shopped: not by wheelbase or trim code, but by anxiety, context, and use-case. “Capable in the mud.” “Seven-seater that doesn’t look ridiculous.” “Can get all the family crap in it for Cornwall, without a roof box.” All perfectly rational human requests – treated as nonsense by the old machinery. The ideas weren’t wrong. They were simply early.

Now the technology has finally caught up. And with it, the entire structure of how we design retail subtly shifts. From catalogue to conversation. From homepage theatre to product truth. From filters to language. From the warehouse to the person.

None of this saves Christmas, of course. But it does save us from the annual pantomime of pretending that people enjoy buying gifts and products more generally through a system that refuses to understand how they think or consume any of the deeper context that matters. The future isn’t more choice. It isn’t more filters. It isn’t even more intelligence.

It’s fit.

Fit between intent and suggestion.
Fit between the context you’re in and the thing you’re shown.
Fit between the human messiness of December and the machinery that finally stops treating you like a clumsy clinical user story.

Christmas shopping isn’t a test of skill. It’s a test of whether the system knows how to listen. And for the first time in a long time, it might.

AI: This piece was assisted with Ai. I used it for the tags, the post excerpt, image generation and some sub-editing. Ideas, references, and anecdotes are all mine.