The 'two-audience problem': brand building in the age of AI

Hello and welcome to the walk podcast. I'm Lena Roden, content director at Walk. Today's conversation is one that is front of mind for marketers. How to build brands in the age of AI. While AI is not new, the introduction of generative AI and LLMs like

Chatchy PT and Claude have had a huge impact on all aspects of the industry. And while this is all work and progress, we thought it was time to take stock of what we know about brand building at time when discoverability and customer journeys are changing. We'll be launching the walk guide to brand building in the Gen AI age in early April. The report looks at how to use AI effectively across consumer research, creative, media

and measurement. So it's going to be a big report and a great piece of work.

So if you're not already a walk subscriber, there's never been a better time to become one.

One of the contributors to that report is Oliver Feldwick. Oliver is Chief Innovation Officer at TMP. And he also helped set up WPP Open that was embedding the AI tools across some of their major clients. Before we move on, let's hear from our sponsors. If you're listening to this podcast, you'll know that walk is the home of leading effectiveness

thinking.

“And now, we're teaching you everything you need to know about the fundamentals of strategy.”

Walk strategy fundamentals is walks new digital course for junior or inter-level strategists. You'll learn from award-winning case studies and hear directly from top strategists and walks on expert. Giving you the latest industry insights alongside proven, practical frameworks. Chopping up your skills to excel in 2026 with walk strategy fundamentals. Available now at walk.com/learning.

Ollie's no stranger to walk. He's written a ton of great pieces for us. He's also been on the podcast. So Ollie, welcome back to the walk podcast. Thanks for having me. As I said, you've got lots of experience in AI, which is why I was super keen for you to write this piece for the report. So for the benefit of our listeners, can you share a little bit about your background and your expertise in the world of AI?

Yeah, of course. So we're actually coming up on it. It's about 10 years ago. I did the IPX and its diploma where I wrote about the early, at that point, the early potential of AI and how I thought it might change the world of creativity and brand building. So I wrote a piece called unleashing cyborg creativity.

“And I think that was kind of the culmination of, you know, my background of, you know, working”

and living in the world of integrated agencies and sort of brand building and more sort of

traditional creativity, but also always having been a nerd and being part of, you know,

sort of the digital industry and digital agency world and seeing those two kind of forces come together and and I think what I've always been interested about is bridging the gap between, you know, what's crossing the chasm between new technology that's appearing and how's that actually filtering through two real people and real brands and the real impact that it's happening and that kind of translating between the potential of technology, which often runs a bit ahead of the reality and

trying to match that back to what's really going on and how that will affect consumers. And so since writing that piece, my focus has shifted, you know, I have a sort of strategy and brand building background, but it's been increasingly focusing on how does the technological context that we do that in change the way we should think about it and one of the themes that I kind of come back to is that, you know, the fundamentals of a lot of what we do stay the same.

We know so much about how to build brilliant brands and we shouldn't throw that away, but we also shouldn't pretend that the environment we do it isn't changing, you know, the way that people research categories shop the web, you know, the even the whole idea of the web as the place where we live so many, it's so much of our lives. 15, 20 years ago, it was a completely different landscape, so it's obviously changed a lot of the considerations that, you know,

marketers and brand builders have. So if the fundamentals stay the same, that the landscape

we're doing and it's changing and that's something that we always need to be staying on top of

and it seems to me that, you know, AI is one of those once in a decade, big changes that could really rewrite quite a lot of different aspects of what it means to be a brand and which is why

“it's worth talking about it now and over the next couple of years, I think we're still very much”

in the early days of seeing that impact playing out. Yeah, in your piece, do a kind of like

and then a bit of a podded history to where we are now and let's just dive into some of the

“themes in the paper that you wrote for us. Only one of the most striking concepts in the pieces”

what you call the two audience problem. Now this is really fascinating and this is where marketers you say markets must simultaneously appeal to humans as well as machines, so the AI agents and algorithmic gatekeepers that will, will be making purchasing decisions on behalf of humans. We're talking here about two very different audiences which you outline in the piece, one is the markets are used to appealing to the messy nature and emotional nature of humans but the other

one is the urban logical and structured audience that is the machines. Can you tell us a bit about these two audiences? How they differ and why this is such a change for brands right now? So this is actually just kind of an evolution of where we've already been, so the digital environment had already introduced quite a lot of machine into the customer journey which was, you know, would be affecting the way that people discovered our brands and we had to learn learning

SEO and understanding how Google works and how algorithms work has dictated the last 10 years

“of marketing as well. So understanding the nuance that has been important but I think that's moving”

from being, you know, that was something that was kind of tactically important and is moving

something that's now quite strategically critical because the role of intelligent agents

is going to just increase in our lives and as because there's going to be ever more trust and decision-making and sort of depth and breadth of questions going to these these new sort of stakeholders in the process. So we're already seeing this with, you know, the amount of traffic that's shifting from more traditional search to generative search and to sort of large language model chat experiences. So for a whole bunch of categories where previously,

you know, the people would trust their brand characteristics and go with the brands they've were familiar with and then go and find them in a retail environment somewhere. Now instead, people are turning to chat GPT to go, okay, what should I be thinking about when I'm making this

“purchasing decision, where should I go and look to buy this thing and their rather than search”

is about trying to connect you to a sort of fairly singular answer in quite a sort of direct way.

This can be much more expansive and qualitative so it can be much more kind of going. Let's let's properly drive the sort of inspiration and depth around your holiday planning or your car purchase decision or what do I need to buy now that I'm becoming a new parent. It's just spreading it's reach a lot further and we've seen, you know, things like the the direct search referral traffic to people like WebMD is as drops sort of 30 or 40% since the introduction of AI

summaries within Google and of chat GPT. So that shows that, you know, in these categories, people are going to shift more and more of their behaviors to this sort of digital aggregator, this sort of that the large language models helping us make those decisions. And so as brands, we need to make sure that we turn up there in the right way and how we get talked about by these models is a complicated and emerging thing that we're trying to understand.

Obviously, you know, what motivates humans is is a rational and as actually they've talked a bit about, you know, we know we've got system one and system two thinking, you know, we're different different styles and speeds of thinking. We have our sort of intuitive slightly more subliminal and are slightly more rational thinking. People have now started talking about system zero thinking. This idea that actually there's a new type of thinking that is largely

done or out out out sort of outsourced to the machine that will make those decisions for us. And understanding the patterns of that decision making is quite an interesting shift that we're starting to kind of understand. And we know that there are different modes of thinking. So the way that we influence people in the ways we influence machines are going to be different. Hopefully, you know, the ambition is that there's still generally a line, you know, people want

to make the right choice and a choice that makes them feel good. And if they're using a large language model tool to try and help them make that choice, it will help them make the right choice for them rather than, so it's not a case of suddenly you've got to do too, you know, but split what you're saying into two radically different things. But you just need to have a concept of how is my brand discoverable and understandable to the machines in the right way? How do the

models make most sense of my brand? And how do I continue to entice and engage and, you know, persuade and seduce my consumer? And those, yeah, those will take slightly different tactics and slightly different approaches. You know, you want people to want your brand with their sort

that we've done to the two parts of a human brain. But we've now, we've kind of split that out even further. And the sort of the machine element is even more, you know, rational and potentially all knowing, you know, we can do a fairly wide reaching piece of research and kind of factor all of that in. So how do we total turn up and what do one our consumers to do in that environment becomes a really interesting challenge? Yeah. And it's a wide ranging research that the machine can

do, but also it's the speed that it can do as well. If we're optimizing for two audiences, so the human mind and the machine mind is there a risk that the output is fills a bit soulless. If we're

“optimizing for a machine? Yeah, I mean, I think, you know, in most instances and it will probably”

vary a little bit by channel and by specifics of what you're trying to do. And this is again, is something, you know, when you're looking at your, you know, product description pages and your

metadata and how you're doing search, you're thinking machine first rather than human first,

but you want to make sure it works to humans. And when you're thinking of your, you know, big video narrative storytelling, you might be thinking a bit more human first, but you're still thinking machine in the back of your head. So I think there's going to be different different moments when your primary audience is one versus the other. I think ultimately, you know, thinking through what that customer journey is and the role that the human

and the machine are going to take in that decision process is something that we just need to think about and think about in depth, the going, how do we want consumers to ask for this product?

“What do we want them to ask if they're using chatGPT or if they're using generative summaries and”

search to assist their decision-making process? What sort of new behaviors do we want to try and and sort of instill in that? Are we sort of similar thing, you know, in the early days of digital where it would go search this to find out more, we're to establish searching as a behavior that people might do off the back of seeing and add. And there's an analogy I find quite helpful, which is if you ever worked in the, in the drinks and alcohol category, there's a thing of

getting the bar call of it's going, how do you, you know, you don't want people to go, I'd like a vodka and tonic, you want to go, I want to smurn off and tonic because then your brand's been chosen before the bartender can choose for you. And what we're adding in here is we've now got, you know, AI agents kind of acting as the digital bartender for everything. So whenever you make a request, if you get people to make the brand call and go, I want, I want to smurn off, I don't want

generic vodka. If you make that, that's one way to win as a brand. So you go human first, go for the heart and get them to request your brand by name. Or you go for the one way, you go, you know, what, I want to just, I know my product is the best possible product. I'm going to structure it in a way that if anyone asks for a vodka and vodka and vodka and tonic, that I'm going to have the best possible vodka that will be the answer to that, of going on the best quality versus

cost versus whatever. The answer is usually going to be a bit of both, but but you have this idea

of going, "Am I trying to, am I trying to get people to request me by name?" Or am I trying to get so that I'm so ubiquitous that I become the default answer anyway? The net result should end up being of going, having good clear distinctive brand communications that knows what you want to stand for and communicates that in a way that works in different channels and to both these human and machine audiences. So I hope it doesn't end up that you're being pulled in two different directions

trying to aim for the heart and the head and missing both. But I think that that kind of dynamic is something that just needs to be more aware of, and we did it to a small scale with, you know, the news search, search dynamic. And again, you know, we've always known that with the way that people shop price comparison websites and aggregators is they trust the machine, but when they're choosing from the short list, they pick the brand they recognize. And so you've got that dynamic of

kind of going, "I want to know, I've got the cheapest, but I'm balancing cheapest with something I'm familiar with that makes me feel good to choose that one." And, you know, even in low interest

categories, you know, that's where brand will play a critical role, but you've also got to understand

how to play the environment that's happening in. So would you say that in the AI age,

“brand is more important than ever? Yeah, I think in a world where there is more access to information,”

more access to content, it's easier to create content than ever. The ways to cut through and stand out and be chosen are going to become more important and more limited. So being, you know, a truly preferred, truly distinctive brand that, you know, people don't want to substitute. It's going to be incredibly powerful when, you know, in previous times when, you know, it was, you were limited in your choice anyway and you're trying to nudge people between two or three things, they might

be able to choose at the local store to now going, you've got, you know, intelligent assistance

that is going to massively help you stand out. That said, we've seen a similar, you know, a new

“type of brand emerging. So we could end up with a kind of direct to agent brand, a bit like”

direct to consumer brands where we've seen brands that are designed just to perform really well within the models. So being the most trusted, the most clearly explained, the most discoverable brand. And in a similar way, you know, it's some of the way that sort of marketplace brands have started to, to challenge a lot of categories. So in a lot of categories, people will choose a brand they don't recognize just because it has the most reviews and it turns up well on Amazon,

for example, a no sort of built for Amazon brands perform really well because they understand the environment they're in. So if you're, you know, if you have a bought kind of utility things of wherever it's, you know, batteries and power tools and mystery kitchen items, where, you know, there's probably one or two known brands that you might quite like. But also, if you've got 10,000 reviews for a spatula and you don't recognize the brand name, but it's got a lot of good

reviews and it's the cheapest one and it comes up high in Amazon. You're going to choose that. And I think we're going to see even more of that kind of behavior where if someone recommends it to you

and it's got and it's a tool that you use consistently and always gives you good recommendations,

you're going to just trust it for more and more recommendations. It sort of becomes your satnav for life. You know, I don't double, I don't double, double check my satnav when I put it in there, the weather that's the best route or not, I trust it will steer me right. And if this ends

“up going, you know what, there are lots of decisions I really don't much care. And I think that's”

a challenge. If you're in a brand where you think, you know what, honestly, I don't want to spend a lot of time thinking about my car insurance purchase. If I truly trust my agent to go, look, I've checked everything. This is the best one. You know, choose it. And I've done the homework, I've checked the small prints and you're not missing out. It's better than your current policy. That extra context and knowledge makes it that much better than what you got

previously from sort of simple aggregates is to now smart recommendations where you go, yeah, that's great. And there you have really sort of disaggregated the brand from that journey, where you go, I'm sort of going, I'm trusting that they've had due diligence done by the agent. And that's one way, and then, but then of course you do have a new brand relationship that we haven't talked about here, which is the relationship with the AI agents or the AI companies themselves.

So, past the reason why I'll trust the number one product and Amazon is because I trust Amazon. And similarly, if I'm using chatGPT or I'm using Claude or Gemini, there's then the power of the brand that sits behind that versus using an unknown AI agent or an AI agent, you know, if your AI agent gets its wrong all the time, you're going to stop trusting it and your brand relationship with that's going to change too. So, it's sort of shifts, you know, it's sort of a new

a new, a new portal for all sorts of customer journey moments is appearing and people are already voting with their feet and using it a lot. So, it's one, you know, a lot of, you know, there are lots of, lots of this is probably a couple of years ahead of where we are right now, but we've already seen with what's happened with AI summaries that people are willing to trust this for a huge amount of, you know, conversations and decisions that people make.

Yeah, the idea of direct to agent brands is really, you know, really captivating actually. But the numbers that I imagine are still quite low, the people that are actually relying on those agents to make decisions for them. That's still very nice and just to be clear. Yes and no, so, you know, at the moment people aren't briefing an agent to go out and do complex multi-step like research and to go and negotiate with a car dealer on your behalf. And that's

far the vision of the future might be that that happens more often. But what is happening already is that any purchase decision that the involves search now involves generative search and some

“generative nudges along the way. So, unless you have to be one of the few markets that don't have”

AI summaries, I think France is one of them, for example. Basically, it was at 13 billion or so

searches a day through Google, have a generative summary at the top. And we've seen that now about sort of 60% of searches as zero click searches, which mean that people find what they want in the generative summary at the top. And don't need to actually click on any of the search results. So, that's actually a pretty radical shift in terms of behavior already. So, people, you know, billions of, you know, people asking what brand of toothpaste should I buy for my,

you know, one year old and, you know, what car should I buy? Are looking at the generative summary and that's in fact impacting their decision journey already. So, that it is already happening in quite a mass way. And that's actually, it's a big kind of experiment sort of happened without us really clocking it. If you ask people explicitly, do use AI to assist in your decision making. I think, you know, a lot of people wouldn't, wouldn't necessarily call that out to something

already, that, that ship is sailed doesn't work.

Good news, decades of benchmarking, insights and vistan class case studies are now available at the push of a button. Lines Intelligence is the new operating system for creative marketing excellence. With state of the art, AI technology and established marketing insights from the work contagious IQ and war. All in a single, adaptive workspace that evolves with your thinking. Save hours of manual work as she or your thinking with clarity and confidence.

Lines Intelligence is where groundwork becomes great work. Insight, strategy, creativity, execution, all in one place. To learn more, check out intelligence.lions.co. Oli, I want to talk to you about something that one of the more technical parts of your paper, you talk about how LLMs work on embeddings and how brand embeddings can encapsulate what a brand stands for based on all known data about the brand. Getting a handle on this will help

“a market to understand how AI personifies their brand. Is that right? Explain that clearly?”

Yeah, so there's probably a couple of bits of going on. I talk about it a bit in the opening of the piece of the Jeremy Bull more quote that people build brands the way birds build nest out of the scraps and straws they chance upon. And that forms the embedding that we have

of a brand is basically all of the associations and the weights and the experiences that we've

had of it based on the available data to us. And the way that large language models work is they basically take as much data as they can from every possible avenue and they basically look at it from a plies of weights to each different thing which then creates that embedding of so any one object or concept or word or token has a series of weights to how much it's associated with other things. And so then you embedding of your brand. If you can imagine it is almost like this,

it's a unique shape of going that's your brand in this kind of liminal concept space that the model has created. And as the model gets updated, that will change slightly. But if you ask a model, you can do, you can do yourself qualitatively, you can also there are tools that help you track, you know, how your brand is seen and how your brand and category entry points are seen by different models. But if you just I've started as a call researcher, so I did lots of brand

projective exercises. And when you ask your focus group, imagine your brand came to life and walked into the room today what would it look like. If you are some model to kind of do that, do a really rich description and explanation of the brand, how it acts, how it's seen in the world. That's going to be a pretty accurate reflection of the brand because it's based on every possible piece of data that the model can build around it. And so as the models get stronger

and get more and more data in them, the more that's getting close to the kind of the ideal,

“that's what your brand actually really stands for in the mind of the collective knowledge of the”

world. So if you ask going what's my brand stand for and that's different to what you've written on your brand, your brand guidelines. The problems you not the model and that shows the difference between your consumers and your vision of the brand. And it also punishes fickle brand managers more. If you change what you think your brand stands for every six months and the models kind of go, well, I think it stands for this, this, and this. That's just a good representation of the same

kind of confusion that your consumers had as well. If you think of the model as kind of like the all-modage consumer that has the perfect complete picture of the brand and then every consumer has their own version of that based on their own experience of it as well. And it just means as well, if you want to change the way the brand is perceived, you can change the scraps and straws you put out there. But whether the model picks that up or not, that depends on how well you're able to put

your distinctive asses out there to put your story out there in a coherent way to get it out there in a world. So it's played back in the model and in the world that it lives in. So for me from a sort of philosophical point of view, I quite like that this embedding is the kind of platonic ideal of your brand and you can use that to assess whether that matches what you think your brand stands

“for and what you want it to stand for. And all you have to run this exercise on any particular”

brand yourself. Yeah, I've done not on a couple of brands and it's, yeah, it's a very interesting tool. You get the same or, you know, a similar level of interest to what you get when you do a do it in a focus group where you get those moments of realization where you go bring this

brand to life. You know, I was doing it with Argos first as John Lewis, for example, recently, where

act to each other? And it actually had quite a lot of really interesting insights about, you know,

the way the brands are perceived in the world, the kind of assumptions that you would make around them.

“And, you know, I think both looking at, you know, brand data and sort of intuitive understanding”

of the brand, it kind of captured it really nicely. And you kind of go, that suddenly goes, what's, what are the strengths and the weaknesses of your brand are, are really kind of sort of a spotlight by how the models see them. So I think you can now use that as a really useful diagnostic tool. You know, obviously it's great to also do focus groups and talk to real consumers. But it's a really interesting way to try and analyze, you know, how does my brand fits in the world?

And it's one of those times when you go, oh, that's wrong. That's not what it says in my brand guidelines. That's a good way to diagnose the gap in your comms between what you think you stand for and what the world sees you as. Interesting. So Ollie, say I'm a brand manager. I'm listening to this right now. How do I actually audit whether my brand is machine readable right now?

What is the first thing that I should do? Yeah. So again, there are a couple of sort of technical

tools that you could use to do this where it would, you know, go through and analyze this sort of discoverability of the data, you know, you want the key information about your brand and your products to be easily readable. And some of that is just, you know, how your website structured, how your content structured, things around making sure that your metadata is structured and, you know, structured well video performs well, but in particular video with transcripts because again,

transcripts then make it faster for the models to pass. So there's quite a lot of like hygiene things about how you structure your data and, you know, get it, get it, get that out there. And things like not having a very convoluted website structures, paywalls make things complicated. You know, if you're a content based brand and a lot of your content sits behind a paywall, that will affect what the model is able to know about you. So which again, might be a strategic

“challenge for you if you want to have the paywall, but you also want the model to have to know”

you as a, as the most coherent and trusted source on a topic. So there's the kind of the technical side. And then I think there is also that slightly more qualitative side where you can do it a little bit like having a conversation with your brands. So actually just going in a structured way, if you work through, and again, some of the tools will help you do this, but you can also do it yourself as a, as they're going to see what happens when you ask about the most common, like category,

entry points and trigger moments. And you ask that question, I want to buy a new luxury SUV, what card should I think about, what factors should I consider? If your brand doesn't appear there, and if the thing is that your brand is strong on, I don't, don't turn up as features, you know, that's a diagnosis of a problem that you could then start to try and fix. And there are differences between the models, you know, there are, you know, our sort of three major models, but there are

versions within those models, and there are other models as well. So, you know, to do this properly, you would look at lots of different wordings of similar questions. And, you know, the difference between buying a premium versus a luxury SUV will change slightly how things turn up. And that might affect how you talk about your brand in the world. If you kind of go, actually, we perform brilliantly well on premium SUV, but badly on luxury SUV, do we want to then fix the luxury

problem or double down on premium? It is the sort of thing that you can start looking at kind of

“going, where do we perform better and worse and what do we want to try and try and fix around that?”

Sure, that's really, that makes sense, that's really, really helpful. I only want to move on to talk about the three tensions that you highlight in the paper. So, you say there's like three tensions that Mark has must navigate to reach this dual audience.

So, the first tension you mention is that between consistency and adaptability and

consistency, as we know, is a brand-building fundamental. It helps build memory structures and makes it easier for audiences to recognize and choose a brand. But you point out with, with AI, there's going to be infinite variation in that consistency is no longer about repetition, but it's about recognized ability under variation. So, can you just explain what does recognition under variation look like and has any brand nailed this?

Yeah, so I think, you know, when a big part, this is about passing kind of the trust and the control about how a brand is going to be brought to life and expressed over into other environment. So, it's a generative environment. So, every time a language model or a generative tool creates something about your brand, there are new things being put out in the world around it. And so, within that comes a certain loss of control. And it's probably most similar to,

as a brand manager, you know, it's always nice to have feel like you have complete control over

your brand. But actually, always, for, for all time, it's actually some of those happy accidents of going, well, actually, this brand expressed through someone else's perspective or point of view, brought to life in this particular way, is still true to the brand, but is a unique and new

“expression. So, coherently different expression of it. I think, in computer gaming, as a bit”

of nerd, there's this idea of kind of procedural worlds and generations where it's some, there's a game called No Man's Sky where, you know, every, every different place you go to in that game is unique and brand new and appears as soon as you get there. But it's done based on a core set of rules. So, there's a kind of coherent logic and narrative to everything, but the way it's expressed every time you go there is different. And so, I think that again,

that idea of, you know, what your, what the Coca-Cola brand stands for as a sort of ideal sort of, core means it can then adapt in almost infinite ways as it comes to life. So, if you, and I think the good test of this would be if you, you know, if you got the brand to just generate a whole bunch of different scenarios and environments around that brand. Could you make it so that each one is still consistently, you know, looks, looks and feels like the same brand, even if it's

different every time. And I sort of feel like this, you know, it's slightly exposed to things

that's always been somewhat true of brands of going a good brand can adapt and survive in all of

these different contexts. And so, wherever you're turning up in a new channel, a new publication,

“you know, brands have had to do, if you want to turn up in TikTok, you don't just look like”

you do in other ads, you have to adapt and fill out what it is to be authentically you in TikTok is different to being elsewhere. And so, if you then have generated models creating content around your brand, talking about your brand on your behalf, you have to trust and but also equip the models to do it well and then trust that what it does is, is going to work well. I quite like there was a thing that the Coca-Cola did recently around all of the

unofficial interpretations of their brand by kind of street art and, you know, food stalls around the world. And I think a more nervous brand manager would try and, you know, get those forbidden and paint it over, go, these are all unofficial brands that don't count. And instead, they did a thing where they acknowledged all of it to go, these are all Coke even if they're, you know, the wrong size, the wrong shape. They're distinctively thematically Coca-Cola, even if they have

“a lot of variation. And I think that confidence and acceptance of those different different interpretations”

of the brand, I think actually strengthens the brand rather than weakens it. So I think there'll be a similar thing where it's that kind of confidence and flexibility that allows the brand to live under sort of that kind of, that, that recognisability under variation. Oli, I want to move on to the next tension that you discuss in the paper, which is the tension between efficiency versus meaning. Now, we all know AI is official and it can create months of content

and enough to noon, but more is not necessarily better or faster is not always better. So

in the paper, you argue that if everything gets easier to make, meaning becomes harder to earn. So how do brands earn meaning and avoid drowning in a sea of forgettable or even soulless content? Yeah, it's tricky because, you know, now there was always been the signalling effect of branding has always been quite an important part of it. So we can afford to make a good looking TV, add and we can afford to buy a TV spot. It was like a big signalling thing that goes,

this brand is proper and you can trust it. And now if I can generate video that looks very slick and elegant and professionally made, really quickly, and I can get it to appear on, you know, appear somewhere, you've got to go look, that's acting like a proper brand, but does it really mean anything and does it really connect? And so I think that's where there's a kind of that tension of kind of going. If everyone can create stuff that looks pretty good, the importance of

actually having something that connects and it's actually interesting and distinctive and meaning becomes more important. And that doesn't mean it's not necessarily a sort of a gatekeeping thing. And we've seen a similar thing again when, you know, when everyone got their smartphone, everyone gets a camera, editing software, suddenly anyone can make video. And as a brand, you're suddenly like, well, what does that mean? It means that, you know, there is there

are more people creating more content. It doesn't mean that only professionals can keep making content. It means professionals need to up their game and know what they're uniquely good at. And it means that creators can come and bring in new ways of telling that story as well. So I think it's a continuing to kind of open up new ways for brands that have something to say and brands have an interesting point of view, have more ways to say that, but being able to crank out more

content efficiently isn't necessarily going to get you to a better result. So I think there is

I think doesn't end up paying back, particularly when everyone else can do that as well. I think

in the past, if you were particularly good at making content, and others weren't, that could be

“an edge. And I think as that bar is getting lowered, if it is easier to create average content,”

then need to create something genuinely interesting in lots of instances. Not necessarily in all instances, but in lots of instances is increasing. Yeah. So it's remembering, you know, efficiency is useful, but it's making sure that what you're doing is effective. Oli, the final tension that you discuss is personalisation versus coherence and personalisation, as we know, personalisation scale has been marketing's holy grail. And as you say, in the penis, AI offers

this potential on steroids. But without without a coherent identity, there's a risk that the brand

becomes a million different things to a million different people, as you put it. So how do

marketers balance creating infinite AI driven one to one expressions while maintaining that shared cultural reference points and brand identity? Yeah. I think it comes to one of those, you know, as we get more tools to do more things, we need to continually remind ourselves that just because

“we can, doesn't mean we should. And I think having an idea of when are we creating a shared”

brand experience versus when are we creating a meaningfully personalized experience? Each time is a choice. It's not necessarily saying going to personalisation is going to be the right answer. It's just a tool that's now much more available to us and can work. And I think being able to personalise from a brand that means something at a cultural level is that bit more

powerful. You know, it's that much more useful to be able to do it. So it's an added tool and an

added capability that we have. We did something recently with Jose Marino and Snickers where we create something where we took a, you know, the universally known platform. You're not, you're not you when you're hungry platform. But we could then use it where we could create a personalise take on this where friends would submit their mates own goals around the the Champions League and an an an agent that we configured to think right and act like Jose Marino would deliver a

funny bespoke response to that. And that for me is one way you've got, you've, you've created a creative system that understands a big brand idea and can then express it in tens of thousands of

“unique ways. And that's a way where I think a brand can deliver meaningful personalisation at scale”

where it's talking from a central point of view and a coherent, you know, world view but it's bringing it to life in a meaningfully personalised ways. So that individual gets something that's unique to them and it's, but it's all part of a bigger whole. Yeah, the Snickers is a great example of that. We've spoken a lot about the challenges that this new world and not so new got, you know, this generative AI is, is posing for brands and challenges and opportunities. But I wanted to

talk about, you do talk about some solutions. And one of the, one of the solutions you talk about is moving from traditionally not from a campaign focus to a brand system model. Now can you explain what exactly is a brand system and how does it help resolve some of the, or the three tensions that we just discussed? Yeah, so I think when we think about, you know, campaigns and assets as these kind of fixed things and fixed points in time, you know, we are, we're limited in how much

we can create and get out there. So that, you know, we, we have these, you know, one off pieces or these set pieces that run for a finite amount of time. And then we, we start again. We go and do something different. We create a new asset. And that, that limits, I guess, the ability to, to connect and evolve what we mean and to, you know, adapt over time. What we think is interesting of actually as a brand system is going rather than rather than our focus being on how do we

make things? It's how do we make a system that can make things, you know, and it's the same shift of, you know, when they, you know, creating a mechanized loom was about creating a way that we could, you know, create on a bigger tap, you know, tapestry on a bigger canvas. We can create a bigger scale and create in a way that reformulates it for lots and lots of different moments, situations and environments. So one of the things that we've been looking at and developing recently

is kind of content engines or content systems where we have an understanding of the brand, we have an understanding of the audience, we have an understanding of sort of data inputs of, you know, what's happening in the world, what are the different moments and triggers that might

based specifically on what's right for the brand, what, what the current data signals that could

be informing that, what's the audience want to see? And that creative system can keep creating in different contexts, different formats, different needs, different use cases. So that, you know, the brand is expressed coherently, but it lots of different channels and different moments and different contexts. So it can, you know, you have much more of a kind of on-demand content relationship. So you find an audience who wants to connect with, it can create an asset that's

designed to connect with that audience based on everything we know about it in that system. Without having to go back to going, well, well, actually our campaign's not due to start until next month, but we didn't shoot that so we can't show that. So it gives you just a much broader

“kind of content variance to bring your brand to life. I think in the past, you know, our ability to”

produce content has been the limiting factor on our ability to just really build out brands of going, well, we, we've got four static assets and we've got two video assets and we can edit them a couple of different ways. But now we could go, well, we can generate every possible variation that we might want or need to bring this to life for different audiences, different moments and different environments. Right. Oli, I've got a big question for you. If we look say two years ahead,

what do you think the metrics will be that matter most in this new landscape? I think the metrics that are going to matter are going to be around. Potentially, I think model cut-through could become, you know, share of model and model cut-through could become a really good powerful new kind of proxy metric that we could be looking at. So it's a sense of going, what impact is our brand having in the world and how is that being

played back in these models that reflect the world? So in the same way that share of search was

“established as a, as a, as a, as a good metric, share of model. I think could be a really good one.”

And then I do still think. And if share of model is a little bit looking at sort of mental availability in the models, I think then sort of mental availability and the humans will still matter just as

much. So I think, you know, looking at those two in parallel, I think will be kind of kind of critical.

And I think it's, it's probably one of those ones are actually what metrics will matter the most. It's one of those things where the fundamentals will stay the same. The ways you achieve those metrics will change. Or, you know, I think it will be kind of fame and distinctiveness and, you know, meaning at, particularly around category entry points is still going to be really important. Like we still ultimately brand will still want to be the most wanted most thought of thing as someone

enters the category. And whether that's, you know, the default thing that people think when they want to want to choose a new catch-up versus when they're researching a car versus when they're planning

“a holiday, like the journey and the context that plays out in more change. But I think the”

kind of considerations and metrics people will be going after will potentially stay the same. That's really helpful. Really interesting. Oli, we're, we're a nidia at a time, but before we close up, I do want to ask a little bit about how all this applies to say start-up brands. Because you talk about start-ups having clarity as an advantage over established brands that have built up memory structures and heritage are their examples of new brands that are already doing

this well or is it too soon? I mean, it's possibly too soon, although I'm sure there will be, you know, I think it's the benefit and the downside of having no baggage. The baggage means that there is already a lot of weight in the model, which is going to be helpful. And this will be a thing that does make a difference if people are asking for recommendations. If there's decades worth of content

talking about that brand, that's going to give you a head start. But if you find a really key point

to cut through on, you can quickly come through and design a brand, as I said, with clarity that cuts through on that specific thing and find an area to win. So it is a bit like when, you know, the D2C brands came along and they spotted a specific channel or a specific gap where there was perhaps some complacency in the market and came in and owned that. So, you know, I think I think a start-up brand can come with a really clear point to view and go, we're going to win in this narrow part of

that. That could be rejecting AI, you know, it could be that you won't find us on price comparison

It could be around trying to get, you know, get a very specific bar call of going, you know,

ask, ask the model for this, knowing that you'll perform well in that or saying, you know,

“don't trust this trust, you know, whatever we bring to the party. So I think there are ways that”

you can kind of box clever around that to cut through and not having that baggage and not having

that, you know, legacy of what you stand for means that you can stand for anything. The risk obviously is that if you stand for nothing, you know, you've got a lot of catching up to do.

Ollie, for you go, I could chat to you for ages, but if you could give one piece of advice to a brand

“that's done, nothing yet prepared for the AI mediated world, what would it be? I think take a good”

honest look at how consumers will start to find your brand in this environment. And I think it's easy to say, oh, no one actually does this yet, but people already do. And so if you just start going, looking at your generative summaries and your, uh, how you appear in the models for the classic kind of questions that you think people will be asking about your brand and your category, that should be a really helpful learning point and possibly a bit of a wake-up call to kind

of go hang on. This is happening and I'm not sure, I'm not sure what what I'm doing about that.

“And I think that's a good sort of starting point to kind of realize, you know, where the gap”

potentially is. Great stuff, Ollie, that's a really good place to leave it. Thank you for that advice. And listen, Ollie, as ever, thank you very much for sharing your time and expertise with us today. If you're interested to learn more about this topic, walk subscribers can access Ollie's article on walk.com and look out for the report, the walk guide, brand building in the A to A I, which will be launching in April. If you haven't done so,

already, you may want to subscribe to the walk podcast on your favorite podcasting platform. Thank you for listening.