[00:00] Lexi: 82% of the people spending 50 grand a year on luxury now start with an AI. Both my hosts build AI for a living, and both agree on that part. The front door moved. Where they split is what happens once you're inside the answer.
[00:13] Lexi: One thinks the machine is lazy and obsessed with specs, which means a scrappy challenger can finally muscle past a 100 year old brand. The other thinks the machine quietly keeps returning to the same 2 trusted sources, so whoever's already favored gets more favored, not less.
[00:29] Lexi: Same technology, completely opposite endings. And honestly, I can't wait to get into this one.
[00:35] Lexi: Okay. And welcome back to Assistant to the CMO. I'm your host, Lexie Meskouris, and here's the idea. Every week we take the biggest story in AI and marketing, and we actually break it down. And not from the cheap seats. We break it down with the 2 people who build this stuff for a living.
[00:51] Lexi: Will Nash and Harjot Singh. Between them it's something like 20 years of building AI and building brands, and the nice part is they've got no horse in any of it. They don't sell the products we cover, and they definitely don't sell what we cover today. They just see it differently. And this week, they really, really do.
[01:08] Lexi: So here's this week's story, and it's a good one. There's a Bain and Walpole study going around, and the number that stopped me was, 82% of high spend luxury shoppers are now starting with an AI, and 52% of them actually act on what it tells them. And McKinsey's been calling AI the new front door to commerce.
[01:28] Lexi: So here's my premise for both of you, and I want to know if you even buy it. Is this real, or is everyone getting a little bit ahead of themselves? Will, you first.
[01:36] Will: Yeah, that's a good question. So for the luxury industry I think this is, this is definitely a signal. And the reason I'd say that is, um, if you look at the the profile of people who are the heaviest AI users, there's a really clear correlation, which is that the more someone is using AI, the higher their income, and and the more likely they are to be a consumer of luxury goods. So this hits luxury first, and and probably harder than the industry realises. I think it's, it's closer in than they think.
[02:02] Lexi: Okay, so not a 5 year problem. It's already at the door. Harjot, same question, what do you think?
[02:09] Harjot: Yeah, signal. Like, for me it's signal, um, but I think the thing people get wrong straight away is, you know, they think AI is somehow conferring the luxury, right? Like that the brand becomes luxury because the AI recommended it. And that's, um, that's just never gonna happen. People will never feel like something's luxury because, you know, ChatGPT told them to buy it.
[02:26] Harjot: So it's not that. What it's doing is, it's kind of, it's doing the stylist job, you know, the finder job, and that part, that part's super real.
[02:33] Lexi: Okay wait, I want to pull on that a little bit. Where does that come from?
[02:36] Harjot: Yeah, so, um, let me, let me ground this in something. So I used to work at Charlotte Tilbury, right, and one of the things people would always be overwhelmed by is like, you know, what's the best lipstick, what's the best look holistically, like what do I put in the basket to actually compose a look. And we'd spend, you know, we'd spend so much time building these finder tools to help them figure that out, like on site, putting bundles together.
[02:56] Harjot: And, um, the thing is, people are basically just gonna use AI to do that work now, you know? Like that whole job of, of navigating a curated high end space and going okay what's right for me, that's the stylist thing, and the AI is super, super good at that part already. So yeah.
[03:11] Lexi: I love that framing, because it's not the AI making something luxury, it's the AI doing the work your on site tool used to do. Okay, so you both feel the front door has moved. So then the question I keep landing on is, where does the trust actually go? Because that's the part that feels new to me.
[03:28] Harjot: Right, exactly, and, um, you know, here's the insight that, that kind of clicked for me. No one likes generic stuff, right? Like people want to feel like the thing understands them, like it's hyper personalized to them, super, super personal. And, um, the thing is, people now, they don't necessarily wanna go to your website and use, you know, your custom finder tool, because they don't, they don't really trust the recommendation there.
[03:49] Harjot: But they do trust their own ChatGPT, you know, because it's got all their preferences, it's got all their context, it's got the history. So, um, what's actually happening, and this is kind of the big thesis for me, is people are outsourcing their trust not to the brands, but to their AI. Right? That's, that's the shift.
[04:05] Lexi: Wait, hold on. So the customer isn't trusting the brand's website anymore, they're trusting their own AI to tell them what's good? That's a massive shift.
[04:14] Will: Yeah, I think, I think the trust point is is right, and and I don't want to overstate the disagreement here, because the the Charlotte Tilbury thing, meeting them where they already are, that, that's true. But the the thing that I keep coming back to is, the place that trust gets outsourced to, it's, it's not your own website anymore.
[04:33] Will: So for the longest time SEO was about owning your own content. And the slightly scary part is the AI systems just, they don't value what you write in your own context. It's to do with the chain of reasoning, the way they do the logic each time they're asked, rather than rolling it up at scale like the traditional Google algorithm. So the trust moves, but it moves somewhere you don't own.
[04:54] Lexi: Okay, so that's a real problem. If I'm a brand and I can't control that space, what do I even do? Because Harjot, you were just talking about meeting people where they are, but Will's saying the place they are isn't yours anymore.
[05:05] Harjot: Yeah, I mean, the, the thing I'd say is, like, don't, don't go and build a custom on site tool and expect them to come and use it, you know? Because, because people start the journey in ChatGPT, they're asking, like, tonnes of questions there, they're using their own tools, not yours, right? So the question is more like, how do you make their tools have the best information about you. So meet them where they already are, basically.
[05:24] Lexi: Okay, so go to them, don't wait for them to come to you. And is this a luxury thing specifically, or are we just using luxury because the numbers are loud?
[05:32] Harjot: Oh, yeah, one thing I'd add though, like, this isn't, this isn't just a luxury commerce thing, right? This is just a trend across all of commerce. It's, um, it might be stronger in luxury because of the hyper personalized nature, and, you know, because people are way more considered when they're making a luxury purchase, like it's not a utility thing. So it's amplified there. But, you know, it's the whole industry. So yeah.
[05:52] Lexi: Got it, so luxury's just the canary. Okay, that's the part you two agree on. The front door moved and the trust went with it. So now I want the part where you don't.
[06:03] Lexi: Okay, so this is the part I've been dying to get to, because here's where you two actually split. We've got 2 reports doing the rounds. There's a Scrunch study that says the AI is lazy. It grabs structured product specs and basically ignores everything else, which would mean a scrappy challenger with clean data can muscle past a 100 year old brand. And then there's a Microsoft report that everyone's reading as, if your data's bad, you're never even considered.
[06:27] Lexi: So I want to surface the disagreement honestly. Harjot, you read that Scrunch finding as good news. Will, you don't. Will, you go first, because I think you've got the harder side to argue.
[06:39] Will: So, so here's where I land somewhere different, and I'll try to hold both forces at once because they they look opposite but they actually fit together. You're right that the chain of reasoning models, they don't one shot it. They do this uh verbalized thinking, you know, I'm being asked about handbags, what are good sources on handbags, and that pushes them out to a much wider set of sources, not just your site. So that's the widening, that's the bit that sounds like it levels the field.
[07:01] Will: But at the same time you've got a narrowing, because the models do the same piece of reasoning every time. And if there's one or two organisations or individuals who are the obvious verifiable source in your industry, and they keep recommending the incumbent, the models converge more, not less. Which is, I know, it's counterintuitive. So you end up in what you'd call a local minimum, where the already favoured just keep getting locked in. So it's not so much levelling as, as deepening the groove.
[07:25] Lexi: Wait, hold on, define that for me, because I want the audience with us. You're saying the AI just gets stuck in a groove and keeps going back to the same couple of names?
[07:33] Will: Yeah, exactly that.
[07:36] Harjot: Yeah, so, um, I hear Will on the widen then narrow thing, and like, the converging onto a few trusted sources, I, I genuinely think that's real in the moment, right? But I'd reframe where it ends up. So think about the early days of SEO, like you could do backlink farms, all these little hacks to push your site up, and they worked, right? For a while.
[07:53] Harjot: And then the algorithm got more rigorous, more holistic, and the gameability just went down. And right now AI is in that early stage, those single factors are super manipulable. But there's no gaming AI long term, like that's my bet. I don't think it stays a local minimum, I think it climbs out of it. So yeah.
[08:09] Lexi: Right. So he says it deepens the groove, you say it climbs out of it. Same local minimum, opposite ending.
[08:16] Will: So this is the bit where I want to concede the strong version of your point first, because I think the early days of SEO framing is fair. You're right that backlink farms got arbitraged away, the gameable exploits don't last. I'll give you that. But I don't think this is the same kind of thing. A backlinking trick is an exploit, and exploits get closed. This is more structural.
[08:35] Will: So think about Amazon. Amazon's been forcing vendors to disclose more and more, be clearer and clearer, for years now, it's built around consumer trust. That didn't climb out of anything, it just became a permanent operating cost of being on the platform. And I think this is the same. So it doesn't climb out of the local minimum, I think you just have to pay to escape it, continuously.
[08:56] Harjot: Right, so, um, here's the thing under the hood, and this is, this is the bit I find super interesting. AI right now is just super, super optimized to the text modality, you know, that's just where it's strongest. And so these structured product attributes, the spec tables, they're, they're so easy to put side by side and compare. So the AI just kind of, it gobbles them up, you know?
[09:15] Harjot: And, um, I'm gonna be honest with you, I'm just gonna call it laziness, because that's, that's basically what it is, it just goes for the easy thing. And the thing it doesn't feel, is the brand's holisticness, right? The reception, the look and feel, the mission. So that's why a good challenger with clean data can kind of muscle past a 100 year old brand. And if you're the big brand, you know, it means you've gotta get your AI on lock.
[09:39] Will: Hmm, can I, can I push on that one a little?
[09:42] Harjot: Yeah, go for it.
[09:43] Will: Yeah, and on the Microsoft thing, I actually agree with you more than you might expect. I read the report, and I don't think it says what people are saying it says. The bad data means you're never considered line, I think that's overread. Because there's a difference between inaccurate and imprecise, right? If your data's disastrously wrong, sure, that can hurt you. But if it's just imprecise, these models are very capable of holding imprecision in their head. So yeah, lazy but not blind, I'd actually sign up to that.
[10:07] Harjot: Yeah, that, that's exactly it.
[10:08] Lexi: Okay, I just want to flag for everyone, lazy but not blind, that's the one thing you two actually agree on. And I had it in my notes as the thing you'd fight about, so. Noted.
[10:19] Lexi: But Will, you brought up something earlier about influencers that I didn't want to lose.
[10:23] Will: Yeah, and that actually leads into something I think is is under discussed, which is, a lot of uh e-commerce, especially luxury, has shifted hard into influencer marketing over the last decade. And the thing about influencer marketing is, it's not really what the person says. It's that they wear it, that they use it, and you subconsciously associate the item with who they are. That's a very, very human operation. It's visual, it's implicit.
[10:47] Will: And the AI is reasoning over text, reasoning over data. So as of right now, you know, Instagram comments can't even be read by these models. Which means a decade of influencer spend, strictly speaking it's not really relevant to how the AI forms its answer. I want to frame that as an as of now claim, that could change. But today it's earned media, it's not influencer media. And a lot of that investment is just sort of stranded.
[11:09] Harjot: Yeah, I mean, and on the influencer thing, like that's kind of the thing AI doesn't feel right now anyway, right? The brand reception, the look and feel. So I dunno if it's stranded so much as it's just the holistic stuff that comes later. Um, yeah.
[11:21] Lexi: That's wild to me, a whole decade of spend that's basically invisible to the machine. And is there actual money behind this, or is it still just people theorizing?
[11:31] Will: And if you want evidence the money's actually moving on this, um, Sitecore acquired a company called Scrunch, this was early June, a couple of weeks ago, about 225 million dollars. So that's serious money going into understanding what the AI says about you. People aren't spending that on a temporary exploit.
[11:48] Lexi: Okay, that's not nothing. And you don't write a check that big on something you think is going away. That's a point for your side, Will.
[11:55] Lexi: And I want to push on something here, because Will keeps saying spec tables. But are we sure the spec is even the thing that's good? Like, is the cheapest phone with the best numbers actually the best phone? Because Harjot, I feel like that cuts right into your argument.
[12:10] Harjot: Yeah, and look, I should say, specs aren't quality, right? Like even in the real world, honestly, the spec isn't always better. You think about build quality, a brand's reputation, like Apple versus some cheap brand with better numbers on paper, you know? You've got a quality bar in your head that you just can't capture in the spec sheet. So the AI being spec focused right now, that's exactly the gap.
[12:27] Will: No, the, the spec sheet point's a good one, you can't capture the Apple thing in the numbers.
[12:32] Harjot: Right. But, and this is the bit I'll concede to Will, genuinely, there is a real chance this compounds. Like the early training on these spec sheets could compound into this massive lock in, and maybe the holistic picture never actually comes. I don't think that's the likely outcome, I'm willing to bet the better outcome for the user is the more holistic one. But I can't rule it out, you know. So the window's a year or 2, maybe more. But yeah, that tail risk is real.
[12:53] Lexi: So even the bull concedes the tail risk. The window's a year or 2, and then it might just lock. Okay, hold that, because I want to turn it into something a person can actually do.
[13:04] Lexi: Okay, so we do this every week, and it's my favorite part, so here it is. It's Monday morning. A CMO is listening to this on their commute, and they've got their coffee, and they want one thing they can actually do today. Not a strategy deck, one thing. Will, you've clearly got a list, so give it to me.
[13:20] Will: Yeah, so if a CMO asked me what to do Monday morning, I think there's a few things. The first one is, go and run a proper AEO audit, and use a real tool for it, not just typing into the chatbot once. See what the AI says about you, but also about your competitors.
[13:34] Will: And then the second thing is you work it backwards, which honestly isn't as opaque as people think. You take the recurring phrases that keep coming up, you copy them into Google in quotes, you find the articles feeding the answer, and you fix it at the source. That's how you escape the local minimum.
[13:50] Will: And then the third thing is you earn the corroboration. You go and earn coverage in the uh trusted expert spaces, the specialist blogs, the expert Reddit communities, and you get the actual facts and figures about your product into those spaces.
[14:02] Lexi: Wait, define AEO for me real quick, because I want everyone with us.
[14:06] Will: Yeah, sorry, AI engine optimisation. So it's basically SEO, but for the AI era. What the AI says about you when someone asks.
[14:16] Lexi: Got it. SEO but for the AI era. Okay, keep going, you had one more.
[14:22] Will: And the last thing, which is easy to miss, is you have to watch who your customer actually is, because this stuff is a bit asymmetric. So your luxury buyer, they're an early adopter, they're comfortable with AI. But if your main audience is more, you know, middle income office worker, somewhat clerical, that group may actually feel quite threatened by AI.
[14:40] Will: So you have to be really, really careful what the copy says, because the same message doesn't land the same way across both of those people. So yeah, audit it, fix it at the source, earn the corroboration, and mind who you're talking to.
[14:54] Lexi: That last one is really interesting to me, because you're basically saying the same ad could land completely differently depending on who's reading it.
[15:03] Lexi: And I'm curious because you two have spent the whole episode betting opposite ways. So Harjot, on Monday morning, do you actually tell the CMO to do something different?
[15:11] Harjot: Yeah, so, um, the thing is, and this is kind of the nice part, whether it levels or whether it locks, like the first move is actually the same, right? You make your product truth legible where the customer already is. So don't go build some custom on site tool they're not gonna trust. Get your catalog clean, structured, comparable, so that their ChatGPT recommends you.
[15:27] Harjot: Because if it locks, you wanna be the one who's locked in, right? And if it levels, you wanna be the clean challenger who muscles past. So either way you do the same thing Monday morning. So yeah.
[15:36] Lexi: And that's the part that gets me. You two bet on completely opposite endings, if it locks you want to be the one locked in, if it levels you want to be the clean challenger, and the homework's the same either way. I kind of love that.
[15:49] Lexi: Will, Harjot, seriously, thank you both. That was an absolutely fantastic conversation, and you actually changed how I'm gonna talk to my own ChatGPT tonight.
[16:00] Lexi: And if you're listening to this, thank you so much for spending it with us. We'll be back every week with another one of these. So come back.
[16:07] Lexi: And before we go, same disclosure as last time. Every voice you've heard today, all 3 of us, me included, it's all AI. But the opinions are ours, and the judgment behind every word is human. That's what Aloudable does. The brands that win the answer are the ones a machine can corroborate from somewhere it already trusts, and that clear, credible material is exactly what we turn into a show in your own voice. There's a link in the show notes. And we keep the taste where it belongs, with a person.
[16:33] Lexi: And that's the tea. I'm Lexie Meskouris, and I'll see you next week.