We don’t buy on price — we buy on passion
Why Agentic Commerce is getting the human behaviour wrong.
Mark Sage - 5 min read - 14/04/2026
I don’t like country-and-western
I don’t like rock music
I don’t like rockabilly or rock ’n’ roll particularly
Don’t like much really, do I?
But what I do like, I love passionately
The Pet Shop Boys were onto something. If I look at my record collection, it makes no sense on paper. There’s rock, classical, even a bit of jazz. From Creedence Clearwater Revival to Charli XCX. It’s not a genre I like or a rule I follow. It’s just… what I felt drawn to at the time.
I like what I like. Or more precisely, I like what I feel something for.
In the agentic future, as some would have it, machines will talk to machines. They’ll decide on our behalf. They’ll automate the mundane, optimise the basket. The weekly shop will simply… happen.
The big consumer win here is that through automation, we’ll get the best deals, and the cheapest price, almost instantly. As if price was ever the defining factor in a considered purchase.
Because sometimes I buy the album for the cover. Sometimes for the colour of the vinyl. Sometimes for a single track. And sometimes for no better reason than I want to try something new.
That moment of curiosity, of irrational pull — that’s not something an agent is particularly well equipped to replicate.
Groceries, of course, feel different. There’s a logic to automating them. Toilet roll, fruit, canned soup — these are the things we tell ourselves can be left to the machine. Remove the drudgery of the weekly shop and free ourselves up to do something more interesting.
Like shopping.
Walmart has been experimenting in this space. When they enabled checkout directly within ChatGPT, conversion dropped by around two-thirds. And this wasn’t even fully agentic. It was simply collapsing the journey — turning discovery into purchase.
What it revealed is something quite simple.
People don’t just buy items. They explore. They discover. And that process isn’t just functional — it’s where the decision is actually made.
Baskets don’t appear fully formed. They evolve. You add something because it goes with the salmon. You pick something up because it’s on offer. You try something new because it caught your eye. The final basket is rarely what you set out to buy.
The act of buying isn’t efficient. It isn’t logical.
But it is where value is created.
Because we don’t just buy what we need — we buy what we like.
My daughter loves raspberries, but they have to come from M&S. She loves fish and rice, but the rice has to be Japanese. Chocolate isn’t just chocolate either. For me, it’s Cadbury’s — every time. You can send me a Hershey’s bar, but the only place that will land is in the bin.
These aren’t decisions made on price or even on objective quality.
That’s not to say price doesn’t matter. It does.
PepsiCo’s 50% increase in the price of its crisps and snacks over the past few years is a reminder that consumer behaviour can be highly price elastic.
Until it isn’t.
Push too far, and people push back — as PepsiCo discovered when volumes collapsed and market value followed, with a $50bn hit.
Price sets the boundaries of choice. It tells us when something is no longer acceptable. It doesn’t though tell us what we want.
Instead, these decisions are anchored on something much less tangible — familiarity, identity, and experience.
In short, they’re anchored on memories.
We struggle to articulate why we choose something, but feel very strongly when that choice is violated.
The book The Mind Is Flat, by Nick Chater, makes a compelling argument that we don’t have a deep, stable inner self, guiding our decisions. Instead, we construct our preferences in the moment, drawing on memory and context, and then rationalise them as if they had always been there.
In one experiment, participants were asked to state their political views. Their answers were then quietly reversed before being shown back to them. Most didn’t notice. Instead, they went on to justify the opposite position as if it were their own.
This phenomenon — known as choice blindness — suggests that preference is often constructed after the fact, not before.
But there’s a catch.
The illusion only works when the person believes the choice was theirs.
The moment the substitution becomes visible, the spell breaks.
In grocery retail, this plays out every day. When an item goes out of stock after an eCommerce order has been placed, the retailer has to decide on a replacement. It’s a small decision on the surface, but a revealing one.
Research into these substitution decisions shows that even when retailers apply sophisticated logic — matching on brand or flavour, or selecting items the customer has bought before — dissatisfaction remains stubbornly high, often in the range of 30 to 45%.
There are patterns that help. In categories where taste dominates, matching flavour improves acceptance. In categories where quality and hierarchy matter more, matching brand performs better. And if the replacement is something the customer has bought before, acceptance increases significantly.
But even with all of this, the outcome still falls short of what would have happened had the customer made the choice themselves.
Interestingly, price barely moves the needle. A cheaper substitute doesn’t resolve the issue, nor does a more premium one. The resistance isn’t economic. It’s psychological.
Because what’s being disrupted isn’t just the product — it’s the sense of self.
We are remarkably willing to rationalise our decisions, even to the point of defending choices we never actually made. But only if we believe those choices were ours. When that sense of ownership is removed, even a perfectly reasonable substitute can feel wrong.
This is where the tension lies for agentic commerce. Machines can optimise for similarity. They can learn from history. They can apply logic at a level far beyond human capability. But unless the outcome still feels like something we would have chosen, acceptance will always be fragile.
The challenge, then, is not simply how to automate the act of buying, but how to preserve a sense of ownership within it — how to ensure that even when decisions are made on our behalf, they still feel like our decisions.
Without that, the system may be more efficient — but it will also be less human. And if we remove that sense of self from the process, those decisions won’t scale. They’ll simply fail.
Agentic commerce isn’t just a race for automation.
It’s ultimately a balance between automation and humanisation.