TheCustomerJourneyIsDead.TheCustomerIntelligenceSystemHasReplacedIt.

Last month a friend replaced her mattress in four minutes.

No twenty tabs. No group chat poll. No nine-minute video of a man pressing his palm into foam. She typed one sentence into a chatbot, best mattress for a side sleeper with a bad back, under a grand, skip the marketing, and bought what it told her to.

It’s easy to say she was reckless, but honestly if you’d ask her she’s she was relieved.

Relief. Stay with that word for a second. Because the most important thing about the new customer journey isn’t technological. It’s psychological. She didn’t just save time. She offloaded the labour of deciding. And her brain quietly thanked her for it

For thirty years we drew the journey as a funnel. Awareness, consideration, purchase, loyalty. We printed it on tote bags. And for thirty years it described one thing: a human doing the labour of choice.

That human has found someone to do the labour for them.

Here is the claim. The customer journey is no longer a path people walk. It is an intelligence system they consult. And it did not come out of nowhere. Half a century of decision science predicted it.

The mind was always trying to get out of this.

Psychologists have a blunt name for how we think. They call us cognitive misers. The phrase belongs to Susan Fiske and Shelley Taylor, in 1984, and the idea is older still: the brain treats thinking as expensive and spends as little of it as it can get away with.

Daniel Kahneman gave the miser two gears. System 1 is fast, automatic, intuitive. System 2 is slow, effortful, deliberate. The whole funnel was a machine for dragging people from one to the other. Catch them on instinct, march them through deliberation, close before the effort wore them out.

We mistook the friction of that march for the market. All those tabs and reviews and comparisons looked like chances to interrupt. But friction was never the ultimate prize. Friction was the tax people paid because they had no alternative. The moment something offered to pay it for them, they were always going to say yes. We have always outsourced our thinking. Now we outsource the deciding.

This is where people get it wrong. They treat handing your judgement to a machine like it’s new. Like it’s unnatural. In reality neither.

In 1985 the psychologist Daniel Wegner described transactive memory: the way we store knowledge inside other people. You do not remember your sister's plans, your colleague's spreadsheet, the name of that restaurant. You remember who remembers. As Wegner put it, you don't have to know everything. You just have to know who knows it.

Then the internet became that someone. In 2011, Betsy Sparrow, Jenny Liu and Wegner published a study in Science whose finding has only aged into relevance. When people expect to be able to look something up later, they remember the information itself less well, and instead remember where to find it. They called it the Google effect. We had quietly promoted the search bar to a member of the team. Later work by Evan Risko and Sam Gilbert gave the habit its formal name: cognitive offloading. We push mental work onto the world whenever the world will take it.

Search engines stored the knowing. What is new is that the machine now does the deciding. It does not just hold the twenty options. It reads them, ranks them, drops eighteen, and hands you two. Kahneman gave us thinking fast and thinking slow. This is something like a third mode: not thinking fast, not thinking slow, but not thinking at all, and trusting that something else did it properly.

Close to 900 million people use ChatGPT every week, more than double a year ago, and it fields roughly two billion questions a day. About a third of people now begin a search with an AI rather than a search engine. In business-to-business buying, where the sums are large and the caution larger, more than nine in ten buyers say they used a generative AI tool somewhere in their last purchase. This is the cue for deliberation to leave the building.

Why we are so happy to hand it over?

Because choosing hurts, and subconsciously we’ve know this for a long time.

In 2000, Sheena Iyengar and Mark Lepper ran what became the most quoted study in marketing. A supermarket tasting table, some days stocked with six jams, some days with twenty-four. The larger display pulled the larger crowd. Then the twist. Of the people who stopped at twenty-four, three percent bought a jar. Of those who stopped at six, thirty percent did. More choice, almost ten times less buying.

Now, honesty, because this is meant to be rigorous. The jam study did not always replicate. A 2010 meta-analysis by Benjamin Scheibehenne and colleagues, pooling around fifty experiments, found the average effect close to zero. So choice overload is not a law. But the later work, including a large meta-analysis out of Kellogg, found something more useful than a law. It found the conditions. Choice overwhelms us most precisely when the decision is complex, and when we want to make it quickly with little effort.

Which is a near-perfect description of almost everything we now buy. So the machine's real gift is not the answer it gives you. It is the twenty-two jars it quietly takes off the table before you ever see them.

And now let me slide this over the table to you. We trust the machine more than we trust each other…

We assume people distrust algorithms. The opposite is closer to the truth. In 2019, Jennifer Logg, Julia Minson and Don Moore ran a series of studies and found what they named algorithm appreciation. Given the same advice from a person and from an algorithm, people leaned on the algorithm more. Not less. More.

But this trust is brittle, and the brittleness matters even more than the trust. Four years earlier, Dietvorst, Simmons and Massey had shown the flip side, which they called algorithm aversion. Let an algorithm visibly err just once and people abandon it fast, punishing it far more harshly than they would punish a human who made the identical mistake. We forgive people their errors. We rarely forgive the machine its own.

Put those two findings together and you get the strange physics of the new journey.

People will trust the machine's recommendation over a friend's, right up until it burns them once, after which they turn on it. That makes the system ruthless about credibility. It cannot afford to be wrong, so it hunts for sources that are unlikely to make it wrong.

And the warning lights are already on. In one early-2026 survey, fewer than half of people said they trusted chatbots to give them the best answer, and nearly eight in ten already assumed those recommendations were being shaped by advertisers, which, for now, they mostly are not. The suspicion has arrived before the corruption. That is the environment your brand is about to be judged in.

So now Trust becomes the asset.

The good news for everyone reading is that credibility is not a mystery. We have been studying it since the 1950s.

Carl Hovland's Yale research reduced source credibility to two things: expertise and trustworthiness. Does the source actually know, and will it level with you. Robert Cialdini added the social signals: authority and consensus, the expert and the crowd. And the availability heuristic, from Kahneman and Amos Tversky, tells us that whatever comes to mind easily feels truer and more likely, whether or not it is.

Stack those up and you can predict exactly what the machine reaches for in the hall of mirrors. The brand that genuine experts reference. The product that real communities defend without being paid to. The company whose story has not changed in a decade. It is no accident that two of the most-cited sources inside these systems are Wikipedia and Reddit. One is edited to exhaustion. The other is hard to fully buy. The machine is grabbing for whatever still smells like the truth.

So whats does media have to measure now?

For decades the scoreboard read impressions, reach, clicks, conversions. Human metrics, all of them. Did a person see a thing and do a thing. None of them tell you whether a machine understands you, believes you, retrieves you, or recommends you.

Byron Sharp's Ehrenberg-Bass work taught a generation that brands grow through availability. Mental availability, being easy to think of. Physical availability, being easy to buy. There is now a third. Call it machine availability: being easy for an intelligence system to find, parse, trust and cite. A brand can own all the mental availability in the world and still be invisible to the layer that now builds the shortlist.

This is where the quiet power of defaults arrives. Richard Thaler and Cass Sunstein built a discipline on a simple fact: people overwhelmingly take the option they are handed. Switch the default on organ donation and you switch a nation's behaviour, as Eric Johnson and Daniel Goldstein demonstrated in Science. The AI's first recommendation is the new default. And the default wins. The machine is not just answering the question. It has become the choice architect.

More than half of searches now end with no click at all, a figure that climbs past ninety percent inside pure AI answers. Publishers are turning from destinations you visit into sources the machine quotes. And recommendation and purchase are collapsing into a single motion. You can now check out inside the chat, and Amazon already credits roughly a third of its revenue to recommendation. The shelf is gone. The answer is the shelf.

Every brand is braced for competition. You can outspend it, out-create it, out-charm it. But the new threat is quieter than any of that.

Gerd Gigerenzer and Daniel Goldstein described a shortcut they named the recognition heuristic. When you recognise one option and not the other, you assume the recognised one is better. Often you are right. Recognition is a cheap, fast proxy for quality, and the mind adores a cheap, fast proxy.

Now run it in reverse. If the intelligence system does not surface you, you are not the rejected option. You are the unrecognised one. You are absent from a shortlist the customer never saw being built. The opposite of being chosen is not being rejected. It is never being named.

So what to do while the future is still arriving?

Make your claims legible: clear, consistent, evidenced, identical in every place you appear, because the machine is cross-checking. Earn the validation of the people and institutions it already trusts, because borrowed credibility is the fastest credibility there is. Build communities that defend you when you are not in the room, because that room is precisely where the decision now happens. And treat your reputation as collateral, because that is exactly what the machine is lending against.

The brands that win the next decade will not be the loudest. The volume era is closing. They will be the most trusted, the most retrievable, the most recommended and the most useful inside the systems where modern decisions are now made.

My friend with the new mattress will never know which brands lost that four-minute pitch. Neither will they. They were not beaten. They simply were not in the answer.

And in the customer intelligence system, that is the only way left to disappear.

Notes on the science

The frameworks Fiske & Taylor (1984), the cognitive miser. Kahneman (2011), Thinking, Fast and Slow, System 1 and System 2. Wegner (1985), transactive memory. Sparrow, Liu & Wegner (2011), "Google Effects on Memory," Science 333, 776–778. Risko & Gilbert (2016), "Cognitive Offloading," Trends in Cognitive Sciences. Iyengar & Lepper (2000), the jam study, Journal of Personality and Social Psychology. Scheibehenne, Greifeneder & Todd (2010), choice-overload meta-analysis, Journal of Consumer Research; Chernev, Böckenholt & Goodman (2015), Journal of Consumer Psychology, on the moderating conditions. Logg, Minson & Moore (2019), "Algorithm Appreciation," Organizational Behavior and Human Decision Processes 151, 90–103. Dietvorst, Simmons & Massey (2015), "Algorithm Aversion," Journal of Experimental Psychology: General. Hovland & Weiss (1951), source credibility. Tversky & Kahneman (1973), the availability heuristic. Cialdini (1984), Influence, on authority and social proof. Goldstein & Gigerenzer (2002), the recognition heuristic, Psychological Review. Johnson & Goldstein (2003), "Do Defaults Save Lives?," Science; Thaler & Sunstein (2008), Nudge. Sharp (2010), How Brands Grow, mental and physical availability. The data ChatGPT weekly users and daily query volume from OpenAI's February 2026 figures. Share of people starting search with AI, and the B2B generative-AI usage figure, from 2025–26 buyer-behaviour surveys (6sense). Zero-click search rates and AI-answer click-through from Similarweb and Ahrefs analyses. Amazon's recommendation-driven revenue from company disclosures. Consumer trust figures from a January 2026 AI-shopping survey (PartnerCentric). Most-cited AI sources (Wikipedia, Reddit) from AI-search citation analyses, early 2026.Share