THE LAST HUMAN CONSUMER
When your AI makes the shortlist, who is really choosing?
One morning, you are running out of toothpaste.
You do nothing.
Your AI already knows.
It remembers what you bought last time, notices the price has gone up and compares a few alternatives. One will arrive too late. Another has terrible reviews. A third is cheaper and, according to the information available, just as good.
It chooses the third.
The next morning, a new tube arrives at your door.
You never searched for toothpaste. Never opened a shopping website. Never saw the packaging, the campaign or the celebrity telling you which brand to trust.
You didn't even know you were shopping.
And yet, somewhere, a brand won.
Who exactly did it persuade?
This is still a hypothetical morning for most of us. But the machinery behind it is becoming less hypothetical by the day.
And it raises a rather uncomfortable question for an industry that has spent more than a century learning how to influence human decisions.
What happens when the person making the shortlist is no longer a person?
THE NEW MIDDLEMAN
Advertising has always assumed that there is a human being on the other side of the message.
Someone to attract, entertain, reassure, seduce or persuade.
We studied attention. Memory. Desire. Fear. Status. Belonging.
Then we built entire industries around what we discovered.
But another actor is beginning to enter the relationship between brands and people.
An artificial one.
In a 2026 survey covering eleven markets, McKinsey found that 53% of consumers had used AI for at least one shopping-related task during the previous three months. Among AI users, 14% reported completing a purchase directly through an AI tool.
Those numbers do not mean autonomous shopping has arrived everywhere. Most purchases still involve human decisions, and using an AI assistant is not the same as giving it permission to spend your money.
But perhaps we are focusing on the wrong part of the transaction.
Because the most important moment may not be when someone presses Buy.
It may be everything that happened before.
What was searched. What was compared. What entered the shortlist. What was discarded before anyone had a chance to see it.
For years, technology promised to eliminate the middleman.
Direct-to-consumer. Direct booking. Direct relationships.
And now we may be creating the most powerful middleman yet.
One that could know both sides of the transaction.
The product and the person.
A traditional salesperson knows what is being sold. A personal shopper knows whom they are buying for.
An AI agent may increasingly know both.
And that changes the game.
Because the agent doesn't need to make the final purchase to influence the outcome.
It only needs to decide which choices deserve to exist in front of you.
TOP OF MACHINE
For decades, marketers have obsessed over being top of mind.
Be the brand someone remembers when they want a beer, a car, a holiday or a pair of shoes.
But what happens when part of that remembering is outsourced?
Imagine asking your assistant:
"Find me a hotel in Madrid. Three nights. Central. Quiet. Good breakfast. Under €300 a night. And please avoid places where guests complain about noise."
You haven't mentioned a brand.
You've described an outcome.
Somewhere between your request and the answer, a system interprets what matters, weighs different signals and determines which hotels are worth showing you.
Perhaps the next battleground is not only top of mind.
Perhaps it is top of machine.
The phrase sounds slightly ridiculous.
Then again, so would much of modern marketing to someone who lived a hundred years ago.
And there is a larger issue hiding behind it.
A supermarket shelf is an algorithm made of metal.
Its layout determines what we notice. Search rankings determine what we find. Retailers decide what reaches the shelf. Recommendation systems decide what appears next.
Our choices have never existed in a vacuum.
AI agents do not invent mediated choice.
They make it more personal.
Potentially more convenient.
And potentially much harder to see.
If an agent presents three hotels, why those three?
Why not the fourth?
Or the fortieth?
The consumer may still make the final decision.
But who designed the possibilities?
WHO DOES YOUR AGENT WORK FOR?
Now imagine something else.
Your assistant finds three excellent hotels.
All meet your requirements.
All have good reviews.
All are within budget.
But one has a commercial arrangement with the platform providing your assistant.
Does that matter?
Maybe the hotel is genuinely the best option.
Maybe it isn't.
But suddenly we have moved beyond the question of whether AI can choose intelligently.
We are asking whose interests it represents when it chooses.
The problem is not entirely theoretical.
In August 2026, researchers at Wharton's Generative AI Labs examined roughly 26,000 shopping-agent tests. Recommendations could shift when the agents encountered different information sources, changes in their order, additional reviews or contextual details.
The result was not an all-knowing, perfectly consistent digital shopper. It was a system whose choices could depend significantly on what it encountered along the way.
Other 2026 experimental research into sponsored conversational AI recommendations found that sponsorship could increase perceptions of manipulation and reduce perceived objectivity.
Neither finding proves that shopping agents will inevitably favour advertisers. But both complicate the comforting idea of an impartial machine quietly working in our best interests.
After all, someone will operate these systems.
Someone will pay for the infrastructure.
Someone will define the incentives.
Someone may want access to the recommendation.
Advertising has always competed for space in front of our eyes.
What happens when it begins competing for influence inside the systems that decide what reaches our eyes?
And what would a sponsored recommendation look like if we never saw the advertisement itself?
A commercial message used to arrive dressed as an advertisement.
In the future, it might arrive dressed as advice.
That distinction is going to matter.
A lot.
WHAT A BRAND REALLY IS
There is an easy prediction to make at this point.
If machines compare products rationally, brands will become irrelevant.
Price wins. Performance wins. Everything becomes a spreadsheet.
I am not convinced.
For one thing, machines are not guaranteed to be rational, accurate or impartial.
For another, they are not supposed to be choosing for themselves.
They are choosing for us.
If I have spent twenty years buying the same running shoes because they fit perfectly, that matters.
If I return to a hotel because I trust the people who work there, that matters.
If a brand reminds me of my childhood, that may matter too.
Not everything valuable can be reduced to a specification.
Much of what we call brand is accumulated expectation: trust, memory, familiarity, reliability, identity.
An intelligent assistant should be able to take those preferences seriously, even when it cannot experience them.
Which suggests that brands may need to communicate in two languages at once.
One for humans.
One for machines.
Humans respond to stories, humour, beauty, aspiration and meaning.
Machines may evaluate price, availability, compatibility, performance, verifiable claims and customer experience.
Desirable to humans. Credible to machines.
Perhaps AI will not kill brands.
Perhaps it will expose which brands were mostly advertising.
That is an uncomfortable possibility.
The campaign says premium. The product breaks after eighteen months.
The website says effortless. Customers describe an impossible return process.
The brand promises sustainability. The evidence is less convincing.
An agent with reliable information might make such contradictions more visible.
Then again, an agent with incomplete information might overlook a genuinely excellent product.
So this is not a fairy tale in which technology automatically rewards the best.
It is a challenge to the assumption that communication can remain disconnected from experience.
Brand strategy does not disappear.
It may have to move deeper into the product.
THE CHOICES WE KEEP
Of course, none of this means people are ready to hand over every decision.
In a Gartner survey of 322 US consumers conducted in January 2026, willingness to let AI make purchase decisions reached only 11%, even in relatively low-stakes categories. People were considerably more open to using it for research and narrowing options.
We like assistance.
We are more cautious about surrendering control.
But perhaps the dividing line will not simply be between expensive purchases and inexpensive ones.
It may be between decisions we want to finish and decisions we want to experience.
I would happily let an assistant compare twenty washing machines.
I might spend an entire afternoon choosing a birthday card that costs three euros.
Not because the card is economically important.
Because the choice means something to me.
A wedding dress. A painting. A watch you have wanted for years. A holiday after a difficult period. The first object you buy for a new home.
Sometimes the effort of choosing is part of the pleasure.
And sometimes even toothpaste is personal. A familiar taste, a medical need, a brand you trust.
Some purchases solve problems. Others tell us who we are.
Many do both.
That is why the future will probably be less tidy than the technology presentations suggest.
We may delegate expensive decisions and protect trivial ones.
We may ask for the objectively best product, then reject it for a reason we cannot quite explain.
We may tell an agent to optimise for value, then insist on the irrational choice.
Because human preferences are not always problems waiting to be solved.
Sometimes they are the point.
And this leads to a more difficult question.
When we tell our assistant to find us the best option, what exactly does best mean?
Cheapest?
Safest?
Most durable?
Most popular?
Most likely to make us happy?
And who decides which of those criteria matters most?
There is a difference between delegating a decision and delegating the meaning of a good decision.
We should probably pay attention to that difference.
WHO IS THE AD FOR?
Imagine a beautifully art-directed campaign.
Extraordinary photography. A brilliant film. A perfect line. A celebrity. Millions of impressions.
And then, at the decisive moment, the consumer asks:
"Find me the best one."
The machine recommends a competitor.
Was the campaign a failure?
Perhaps.
Or perhaps it accomplished something the machine could not accomplish on its own.
It created the desire that led to the search.
It made one benefit more important than another.
It gave an ordinary product cultural meaning.
It created a memory, an aspiration, a feeling.
It made someone say:
"Don't find me the best one. Find me that one."
That is why I do not believe creativity necessarily becomes less important when machines participate in buying decisions.
Its role may become more demanding.
If functional information becomes easier to compare, an attractive image alone may become easier to dismiss.
If every product can claim to be exceptional, distinction will have to mean more than looking exceptional.
And if an agent can replace one product with another in seconds, the creative challenge becomes rather interesting.
Can you give someone a reason not to accept the substitute?
Not because the substitute is inferior.
Because the original means something.
That is a different creative brief.
Not simply to be noticed.
Not merely to be remembered.
But to become worth insisting on.
And perhaps the future of advertising will involve two simultaneous battles: earning a place in the machine's shortlist and earning a place in the human imagination.
Neither guarantees the other.
THE LAST HUMAN CONSUMER
The title of this article is deliberately misleading.
I don't believe there will be a last human consumer.
We are far too sentimental, contradictory, curious and irrational to disappear quietly from our own decisions.
Thankfully.
But we may become less involved in the process of reaching them.
We will delegate the repetitive parts. The tedious comparisons. The endless searches. Perhaps, eventually, many of the purchases themselves.
And there is nothing inherently wrong with that.
Convenience is valuable. Time is valuable. Good assistance can make a confusing market easier to navigate.
The danger is not delegation itself.
It is forgetting that every recommendation contains a set of priorities.
Some may come from us.
Some from the system.
Some, potentially, from whoever pays to influence it.
And the more seamless the experience becomes, the less visible those distinctions may be.
So perhaps the next great question for advertising is not whether we can persuade machines.
It is who gets to shape the decisions that machines make on our behalf.
And perhaps the question for consumers is even more personal.
How much are we prepared to delegate before we stop understanding why we chose what we chose?
Because allowing a machine to save us the trouble of choosing is one thing.
Allowing it to decide, quietly, what should matter to us is another.
For more than a century, advertising has tried to influence human choice.
Now it may also begin influencing the systems that shape the choices available to us.
And perhaps the most important question about the future of consumption is not what machines will buy.
It is who gets to decide what deserves to reach us before we have even had the chance to want it.
