When AI Decides What You Buy, Who Do You Trust?
Many people today use AI instead of searching online to decide what to buy. There is this idea called ‘Agentic Commerce’ where you tell your AI what you want and it deals with all the shopping, ethical requirements, allergies, food origin, all of that. People with this idea dream of it being like when we got TV without advertising: a sigh of relief and lots of extra free time. Let algorithms decide on your behalf because it's such a hassle to find the right thing.
But I think what will really happen is more complex. For AI to work and help you to buy what you want, you need a neutral AI. But for there to be a neutral AI you need clean data, good product specs and for companies to refrain from trying to influence the AI: in other words you need trust. But trust needs to hang off of something, and I think that leads to brands. And Brands need a space to reach customers which, I think, is basically… advertising!
I imagine a strange scenario to convey the difference of how Agents shop. Imagine you wake up one morning transformed into a LLM: you would immediately lack volition: the whim to do stuff. Anything you see, especially if something needs to be fixed, would force you to respond automatically. You would feel compelled to wash the dishes as soon as you see them, to vacuum your apartment if you saw a bit of lint, to finish the half finished crossword puzzle someone left on the table. And if you happened to come across some meeting notes with NO SUMMARY or NEXT STEPS? Oh no! You could not resist adding this!
I don’t think it would be a very satisfying existence. Even if you had skimmed through the entirety of everything ever written as part of your training: you would exist exposed, without a direction and no useful identity. You would be vulnerable to an infinity of missing pieces and questions that need to be answered, cleaned up and finished.
Paradoxically the ultimate downfall of you as a LLM would be how you respond to advertising. There would be campaigns more irresistible than sexy models and baby white rabbits on toilet paper packaging are for humans. The same way you could not resist to incomplete to-do lists, you would try to climb up a billboard to fix the spelling mistake in the advertisement for the next accident insurance: “Whould you crimb up here to fix this sntence? Ladderfall Insurance International is here for you!”
This idea, how advertising might affect an AI differently from us humans, is at the same obvious and suggestive of a deeper point. Humans have the ability to choose, even before they know what they want. Synthetic entities will not really be “free to choose” until they are also conscious: a huge can of worms I won’t deal with here.
So maybe advertising is the canary in the coal mine?
When we imagine delegating our decisions to a system, we face this problem. The system would not just sit there, it has to have an answer. There is a big difference between a system that is trying to pick the best car based on various parameters, and one that might decide it is actually fed up with driving and it might be a better idea to walk to work.
My friend Nick came over to meet me at our offices in London. I hadn’t seen him since the Math Rock festival we went to where I recorded a hypnotic slow motion video of his lips making a raspberry (that’s like a fart sound with your mouth). Oh, and Math Rock is like heavy metal done with physics simulations: weird stuff, but can be quite fun.
Nick has spent time with this AI stuff in the advertising world, and I am interested in his aesthetic and insights. We were having a delicious lunch that ended up giving me a belly ache so I won’t tell you where it was.
“What do you think of my piece, Nick?”
“Well, yeah interesting!”
“And…?”
“Well, not sure exactly where you are going with it.. ”
“Ok, well. In this advertising & AI there seem to be hidden layers…”
Nick rubbed his beard.
“It's quite philosophical: but, in ads, if it works, it's ok!”
I held back from rubbing my beard. Instead I compulsively ate some extremely delicious garlicky crispy potatoes that I am sure had no role in the subsequent belly ache.
I said: “Don’t you think we ignore some things that matter?”
“Ok, for example?”
“Well, like I wrote, we are different from AI’s. We can decide what to pay attention to. We can look at the laundry and feel lazy and just go out for a walk instead. These systems cannot, they are compelled to answer the latest question you give them. Yes, it’s philosophical, but… I think it's important?”
“Hm, you need to spend more time with ad people.”
“Ok. Like maybe, with all this AI craze, maybe advertising is part of the solution?”
“Ha ha! Keep that to yourself.”
Nick’s cynicism was always a breath of fresh air, which to be fair, is common across ad people in private.
“Seriously, I mean if we look at something like the idea of volition or something like free-will. It is connected both to laziness and to allowing people to decide what to buy. Even if ads do try to bend their desires.”
“Hm, I think you over-estimate something that is just business. Advertisers want to get you to buy stuff – whatever works.”
The two of us go on eating, white wine, some clams I will regret, and coffee. But by the end I am somehow left with a slight frustration that, what to me seems like a deep truth, is not coming through. I don’t think I managed to convince Nick. But then again, maybe that’s a good thing. Different minds, different views, no single approach can persuade us.
“I mean, going back to my point. Maybe advertising is closing in, getting close to our human nature. Maybe it’s like when rabbits thump on the ground to warn the others of danger?”
“Yeah, well, I do see your point. But I can’t see how it fits in what you do for a client, it's too abstract.”
I take the tube home grumbling in my stomach and in my brain. Yes, most people don’t like advertising, in fact, they mostly hate it. And I agree, it’s really annoying and perhaps it needs more regulation, especially in format and when it tries to trick you. There is a quote I try to remember about Democracy, maybe it's similar: “Advertising is the worst way to influence peoples’ choices, except all the others we’ve tried…” I get back, and my orange friend cat greets me, it’s late but I am still full of energy, so I go on writing.
So let's imagine an experiment: what would happen? If advertising just disappeared? Puff and gone in a cloud of smoke like magic? I believe this would be bad, or a bad sign at least. Actually I feel it in my gut that this would be VERY BAD. How is it that many friends think this is inevitable, and not even so bad?
I messaged out to Andrej, another friend that tends to disagree with me in a consistent way. He also can be a tenacious arguer, drives me nuts, but forces me to work on things and improve them.
Y: [Hey can I share a piece from my blog?]
A: [Hi! Yeah… Ok, I read it. I like your piece but I just don’t buy it.]
Y: [What do you mean?]
A: [I think it's inevitable. AI will take over and do your shopping for you.]
I can feel my blood pressure rise, I think because I find this scenario ugly, not only unconvincing. Maybe I am old fashioned and miss growing up and watching TV – to actually watch the ads, they were the best part!
Y: [I see big problems with that. You can’t easily just optimise. Most people don’t even automatically re-order toilet paper monthly, the simplest AI in the world!]
A: [Like I said, I just can’t see it. I mean if I ask AI what laptop to buy… I don’t feel like going through all the options. The system will know everything about me and it might even order a laptop for me before I know it.]
I can feel my blood pressure rise more. I get an instinct that there is something deeply wrong about this view, that ignores something fundamental.
A: [Look, I pasted your paper in an AI and asked it to critique it and it replied like I would. “We are entering the age of Agentic Commerce which solves all the difficulty of choosing a product!”]
Y: [I disagree, for now instinctively, I need to work on this to explain. For one, if it could work, then wouldn’t Amazon or Alibaba have solved it?]
A: [I don’t know, I like your theory. Just not sure I see how economics works.]
Y: [You need Trust, or would you really use an AI? And I think Trust in commerce needs Brands? Maybe regulation too, but that is trickier…]
A: [Sorry, I just can’t see what you are suggesting is going to happen.]
Y: [Well maybe it's as dumb as asking Perplexity for the best vacuum cleaner and I get a comparison and also a popup that tells me I can get 20% off a Dyson?]
A: [And, that solves a trust problem?]
Y: [Yeah, I think so! Why else would Dyson pay to advertise!? Would you pay if you knew the Perplexity was biased?]
Our discussion goes on, and on, I am exhausted. He is on to something so I can’t completely ignore it. So I break down and create a special page of brain calisthenics argument with Gemini 3.0 to demonstrate what I am trying to explain would happen. To be fair I don’t use the heavy reasoning level because I just want to convince Andrej that there is a route that even AI accepts. And as you dig into the idea of Agentic Commerce, at some level you just discover advertising again. Unless you are willing to really hand your total destiny off to synthetic existence.
Look at this argument if you want to.
It's a common future scenario. A future like Star-Trek in which you can walk up to a replicator and order anything you want and it will appear: “Gran’mas Ribollita!” and whoop it’s there. Many I know believe this is a possible deep future, usually the nerdy friends. Some day you will create anything from some kind of self assembling material made of molecular nano-bots. They yell out: “Look at 3D printers!”, at some point, to indicate how disruptive that technology is. You hear the word Abundance thrown around as a future where you can enter the name of what you want and there it is! I imagine it to be a bit similar to the infinite catalog world of AliBaba or Amazon that has everything imaginable. Some people believe this is a dream world with – no advertising – where everything is at your fingertips!
I get it and I like Star-Trek, but it's just too good to be true. There are hidden problems that are brushed under a rug. Everything has a cost tied to it: environmental, energetic, attention, time, and this is ignoring them. For one thing there has to be a way to decide what is in the catalogue and what not, and who decides this. This isn’t just a matter of censorship, it's fundamental, if you can only buy bread and water this abundance is not very impressive. Or if you can only buy junk that barely works, or that claims to solve problems it doesn’t.
And what if you disagreed with how an algorithm assigned you to broccoli? Could you choose frozen peas instead? Could you choose which brand of peas? Or would you be compelled to have Prime Perfect Price Peas? The current online catalogue sites themselves are not working very well. I keep running into trickery in reviews, poor quality products, overcrowding of junk brands, ambiguous labelling and frequent huge gaps in the one thing you want. If we lose advertising, and therefore brands, we would enter an even worse Junk-o-calypse. Losing the most basic of quality guarantees and product responsibility.
I need to fly back to Italy from the UK and – wow – flights in aeroplanes are pretty awful. You get wafts of pollutants before take off, then some other strange smelling air when you sit down – I don’t even want to know – then the flight attendant sprays a whole can of bug spray (depending on destination), and you are left to wonder what it will be like during the flight, or if you might get nasty particles from someone. I put on a mask in protest against crap air, and I drop into my headphones to try to forget the world, while I stare at my collection of media.
Now I have to choose what to listen to. This is some sort of abundance. Do I listen to an audiobook, a podcast or some music? Something from the past, or something from today? Something I bought, or something sort of free?
I pull out a notepad, and while I am partially asphyxiated and I feel my reasoning has degraded to the level of a dead salmon, I write a few words:
“AI needs to make room for advertising. If it doesn’t then => everything it says will look like paid promotion. Our lazy questions to AI depend on Trust – a kind of neutral point of view => So maybe advertising will not die?”
I have been worried about various aspects of the gen AI bubble we are living in. But, the idea that advertising might seep into the various AI chat tools people use, to me, seems a positive one. Although I know it will irritate everyone else in the world. I like it because it would mean – clearly and unambiguously – that the ultimate commercial decisional power is with the human customer directly to a brand, not through a black box mechanism.
What I mean is that if there were no advertising you have to wonder why an AI is recommending one product as better than another. Maybe it is doing so because the manufacturer is paying the AI to learn this – like product placement in a film? But instead if there is a big ad next to the site recommending the next vacuum cleaner, you would have to expect some degree of neutral judgement of the products. Or else why would they buy these ads?
Also, if there is an advert popping up next to an AI response, you need somewhere to go to when you click the ad. This might mean the Internet we are familiar with is not dead – the dark internet hypothesis might be averted. This is a theory that the internet goes dark because the only ones using it are AI Agents following peoples ’ questions – which I can literally see in web stats.
Maybe even this article would survive?
I fell asleep for the rest of the flight ignoring the rituals of snacks and drinks. Miraculously they somehow stopped duty free on this flight, which used to take up half the trip. In my shifting around, with my head dropping down. I kept mulling the idea and its connections. As creatives that work in advertising, what do we do? And aren’t we just confusing and polluting a world as it fills with slop generated at industrial scale?
A week later I was at a fancy dinner party and someone asked me what I did. I never know what to say because the formal answer of – Advertising Production – doesn’t really capture how many weird projects and ideas UNIT9 has worked on. I started rambling and it sounded awful, but luckily there was someone there at the table that was aware of my work and they encouraged me to try to explain.
“I love mixing tech with creative work. For example in a parallel life I would have liked to become a neuroscience researcher. I got the opportunity to go back to school and get a masters in this. Then we won a job to measure the brains of Rally drivers for Ford Performance. We worked with weird ideas like passive haptics in VR, crowdsourcing for science and mapping the brain, even sensitive microphones to measure wriggling worms for healthy soil…”
The lady I was speaking to said: “Well… isn’t that storytelling for brands?”
She’s right of course: but we have been doing the tricky science projects as the production arm of some big agencies that manage the clients. And today something is different, I think at least in part because of AI.
And there is a strange parallel to the problem I see:
Ad agencies should have “agency” to act on behalf of the client in the best interests of the brand.
AI Agents should have “agency” to act on behalf of the user in their best interest.
I see a problem with trust: do you trust your agent if they are also going to charge you to do the work? Would you trust an AI Agent that sells you the best airline deals that are heavily invested by the same airline?
So in a totally bizarre twist: advertising space next to an AI Agent allows the latter to focus on the user benefit of the response (under certain rules). Just like the best advertising Agency would benefit from focusing on Brand benefit independently of the actual production work.
A part of the world is caught in a narrow trap of thinking that automating stuff is: more = better and richer. But in fact the mere idea of being able to automate creativity of different kinds has made us worried about creating. What if this article I spent hours writing could be generated in five minutes with an AI? Why bother writing it?
And all the while I don’t think stories will go away: there is nothing more human than storytelling.
(Thanks to Nick: https://wvsh.ai/ for agreeing to my representation!)
(Thanks to Andrej: for disagreeing consistently)
Find Yates Buckley' original blog post with LMM game prompt here.
