Fast, Wrong, or Both? Dr. Thomas Ramsøy on Using AI in Advertising | TMRE 2026
AI is making advertising faster, cheaper, and easier to produce. But is it actually making advertising better?
For this special episode of Signal & Noise, Lucas Longacre, Head of Product at Inlightened, joins us as our on-the-ground correspondent at TMRE 2026 (The Market Research Event) in Denver, where he sits down with Dr. Thomas Ramsøy, neuroscientist, author, and Founder & CEO of Neurons, for a fascinating conversation about neuroscience, artificial intelligence, consumer behavior, and the future of advertising.
After attending Dr. Ramsøy's presentation, Lucas was so intrigued that he arranged an impromptu interview to dig deeper. What followed was a wide-ranging conversation about how AI is transforming advertising, why marketers may be measuring the wrong things, and what brain science can teach us about creating advertising that actually works.
Dr. Ramsøy shares a startling statistic: brand and advertising memory has fallen from roughly 30% to just 5%. Meanwhile, consumer attention to ads has dropped from several seconds to less than two. We're producing more advertising than ever, but people are remembering less of it.
And generative AI may be making the problem worse.
As marketers rush to generate more content, test more variations, and optimize for clicks and impressions, Dr. Ramsøy argues that the industry is overlooking something fundamental: good judgment and a scientific understanding of how people actually respond to advertising.
But he's not anti-AI. Far from it.
Through Neurons, Dr. Ramsøy has developed technology that uses neuroscience and machine learning to predict how people respond to creative content. Rather than simply using AI to produce more ads, he believes we can use it to understand which ads will actually work—before spending millions putting them into market.
In this episode, Lucas and Dr. Ramsøy explore:
🧠 The Collapse of Advertising Attention: Why brand memory has plummeted and consumers are becoming increasingly selective about what they remember.
🤖 The Generative AI Paradox: Why producing advertising faster doesn't necessarily make it more effective, and how AI risks accelerating existing mistakes.
📊 The Problem with Vanity Metrics: Why optimizing for clicks, impressions, and view-through rates can undermine long-term brand building.
🎯 The Four Powers of Effective Advertising: The neuroscience behind stopping power, persuasion power, transmission power, and locking power.
🔬 Predicting Consumer Behavior with AI: How Neurons combines neuroscience, predictive models, and generative AI to evaluate advertising and improve creative before campaigns launch.
⚡ The Future of Creative Optimization: How predictive, suggestive, and generative AI could transform how brands develop and test advertising.
🏢 The Innovation Dilemma: Why smaller organizations may have an advantage in adopting AI, and how established companies can avoid falling behind.
👀 The Two-Second Challenge: What advertisers can learn from TikTok, Instagram, Snapchat, and YouTube about capturing attention and making every second count.
The conversation also explores AI-generated content labeling, scientific validation, the future of market research, and why human creativity and judgment remain essential.
The big takeaway? The future of advertising isn't about creating more content. It's about creating content that actually works.
A special thanks to Lucas Longacre of Inlightened for representing Signal & Noise at TMRE 2026 and bringing us this terrific conversation. His background in product development, technology, and storytelling makes for an especially engaging discussion.
About Our Guest
Dr. Thomas Ramsøy is Founder & CEO of Neurons, a company applying neuroscience and AI to predict consumer attention, emotional responses, and advertising effectiveness. He is also the author of How to Make People Buy.
Learn more: https://www.neuronsinc.com/
Subscribe for more conversations with the people shaping the future of marketing and technology.
#SignalAndNoise #ThomasRamsoy #LucasLongacre #Inlightened #Neurons #TMRE2026 #ArtificialIntelligence #Advertising #AdTech #MarketingAI #Neuroscience #ConsumerBehavior #GenerativeAI #BrandBuilding #MarketResearch
Transcript:
speaker-0 (00:01.719)
Is it?
speaker-0 (00:10.857)
Tell us a little
A little rocky. Well, we'll stabilize it. Well, thank you for joining me. I'm recording here at TMRE in Denver. I just finished actually going to a a very interesting talk here with Dr. Ramsey. And do you mind just giving us a little background of of who you are and why you're speaking here today?
Sure. so I have a background in psychology and neuropsychology, many years in clinical sciences. And then I did a PhD in neurobiology to study kind of what happens in our brains when we make decisions. So all the FMRI scans and all that, I I knew my way around and
So to speak, and all the statistical analysis of it. and then I got really interested in kind of moving stuff out of the lab and into the real world to study kind of everyday decision making. That got me into onto this track that you know with the Copenhagen Business School, University Hospital to start a kind of research lab back in the days. And that eventually turned out to be, you know, contacted by a a huge you know American company, Lowe's, for example, the Home Improvement company that
they realized that they needed the science, neuroscience. you know and understood that both as an inspiration, so how can we improve stuff? But also how can we measure some of the things that because you can't just ask people questions all the time because we say one thing, we do another thing. But
speaker-0 (01:30.84)
They realize that maybe we can measure some of these things and you know when things break and when things work, right? So I was invited in two thousand and twelve to start kind of my you know, my company back in the days. So in twenty thirteen I started Neurons, which is a company that you know initially started to be a kind of consultancy. We tested how people respond to anything from completely new innovation, product testing, unboxing
Kickstarter campaigns, X Prize, the whole thing. All the way over to testing advertising and also downstream kind of loyalty, for example. So and then the big pivot came in twenty twenty, actually twenty eighteen, when all of a sudden we could we realized that we could predict the stuff that you usually would have to measure every single time. I was al always on this trajectory, really kinda inspired by Isaac Asimov and psycho history and the whole thing. So is it possible? So I kinda hand coded, you know
back in twenty twelve kind of hand coded these kind of algorithms to try to do it.
I mean, you're speaking my language. I I grew up reading Asimov and Foundation series, but
Exactly. So history is like can you combine psychology and mathematics to predict human behaviour? And it turned out you can do that. So we and it incidentally at the same time there's been a new type of neuroscience that's called neuroforecasting that shows that you can take a handful of people, twenty people, and the brain responses they have that it is orthogonal to individual preferences, idiosyncratic responses if you like.
speaker-0 (03:01.762)
They there's actually some deep responses in those people that predict some market responses. So that's been shown for music, TV and movie hits, it's shown for financial markets, it's been shown for the whole well every single time we look at it, it happens. So with that in mind, I kind of turned in twenty eighteen we realized that we can turn this into kind of AI machine learning algorithms to improve and predict improve the predictions basically of of
tension, emotional responses, connect responses, memory. And we now have been able to do that. So today you can show an algorithm an image or a video and it can very accurately predict exactly where people are gonna look, how they're gonna respond, and the likelihood of them remembering the ad or the message or who was talking. So that's like you know very short kind of nutshell raise for me kind of who I am and where I
I mean that's incredible. As a product guy, I'm very used to this thing where people say one thing and then you look at their behavior and you're like, Are you sure? You said you really like this feature and I see in the data you never touched it. You never used it.
Very early on as well because we the company was basically born and raised in the US. So straight into Silicon Valley, worked very early on with you know, Google.
So you're doing a lot of machine learning stuff with Google, I imagine, with their stuff, right?
speaker-0 (04:24.106)
But also just the neuro testing. So we don't call it neuro prototyping. Published a book a few years ago where we actually explained that, you know, how you can use the science to use we call it future KPIs. It's like you can't just like you can't you say, Okay, I'm gonna I'll make this moonshot, right? How do you steer in the right direction towards that moonshot? The end goal is to land on the moon, but you need to have steps on the on the journey. Just asking people questions is not enough. So we we did this testing with you know anything from
Casey Yeah, it's it's awesome.
speaker-1 (04:47.427)
Definitely.
speaker-0 (04:53.048)
Google was in the developing their tango, the augmented reality solution. These are all publicly available information. When meta acquired the populace, for example, and wanted to study social interaction, every time we saw that, you know, people say one thing and then they behave in a very different way. And we can kind of pinpoint exactly where the breaking point was.
I mean, I would love to focus a little bit about what your your speech that you gave because that to me was I wasn't even planning on doing any interviews today, but I was so intrigued with what you were presenting. So you opened up with this number that kind of like stopped me in my tracks, which is brand and ad memory has dropped from thirty percent to five percent. Yeah. You know, because of what is it, TikTok, Instagram, whatever, just new technologies and and mediums have now kind of broken this idea of like
brand recognition, right? So can you just speak a little bit about that, like what's happening?
I think I think that if you when we do the measurements and we've done these measurements for twenty years at this point, I think. So and other people have done that, CamTar, Ipsos, all all these companies have done that as well. We are measuring both app memory and brand memory. So we are even though we are kind of an AI company that provides these predictions, we still base that on actual measurements of people. So we constantly update our models, and that means that we constantly also get data, right? and we've seen just implumets. Initially it was like there's something wrong about the data.
There's something wrong, right? But then we saw that from other companies as well. They saw it's like just incidentally, it's just same drop, same drop, same drop all the time. They thought the same thing, there's something wrong with the data. But it turns out it's true. And that's of course a problem. And the question is like what should how should we explain it? One is that are people forgetting more or are they remembering more selectively? And I think it's more selective memory that we're saying.
speaker-1 (06:38.45)
Is it a volume issue too? Like because we process so much stuff, we're gonna really only hold on to the real winners? Something like that? Okay.
And another thing I talk about well is that we've also seen that ad memory is plummeting as well. sorry, ad attention is plummeting as well. We see that it's gone from a few seconds, you know, three and a half, four seconds, down to less than two seconds now. That also means that it's not because people have poor attention skills, probably better. They're probably better at discerning this is an ad, I want to move on. And that decision is just getting faster and faster.
Well, I'll say so. I'm a video guy. I come from TV and film. I had a food and travel show when I was younger, a documentarian. I can't do TikTok. It's just for me the how the algorithm works and how the for what works in that format is so destabilizing for me. And the thing is that like fire hose. Whereas I see a young person on there and they're like flipping through so I do feel like and I'm I think I'm the outlier now because I know grandparents who are on TikTok and they're thriving. So I do feel as though like
the medium is changing how people behave and process things. And so that to me is yeah, like is it a so I guess my question for you is like how much of this is a a collapse of a technology problem, like phones, feeds, fragmentation versus like a creative problem. So is the industry making like bad ads for the medium or they're just we haven't learned quickly enough yet what is working in that medium, I guess I would ask.
It it's both. I I think that on the one side definitely there is a technology problem because we see that technol technology is inspiring or pushing people to consume information in different ways. We see that, you know, some platforms are very comparable to each other when we look at their you know, how people respond to it. Instagram and TikTok in our data sets tend to be very comparable. But, you know, Snapchat and YouTube are very different first. So there definitely is a kind of a platform and how things are manifesting on those platforms.
speaker-0 (08:28.856)
But the big change we see is is how the creative is formed and shaped. there's a lot of wind and wins and and losses on how you create that creative. what we see is that people, first of all, in the old days, you know, before Facebook came, we had TVCs and that was it. Then Facebook came and the first thing people tell said was just take that TV C understand why that didn't work, and then they realized, shit, we need to fix that. and then now we're in this space that hey I I don't have
Throw it on there.
speaker-1 (08:52.62)
Yeah, yeah, definitely.
speaker-0 (08:57.262)
three platforms. I don't have six platforms. I have a gazillion platforms. And even within that single platform, I have reels and feeds and maybe a third version as well. So that also means that I need to understand exactly how should I communicate on that platform and that sub platform as well. And the problem that I talked about in my talk today is also that we see that people don't even have the time or the budget or the the sophistication, the knowledge so to speak to understand how what works in
Because we tend to work very much on behavioral data, but the neuro data is the causal data that gives you that aha moment of that's what's working. It's because that people see it but they don't respond to it, or they don't see it in the first place, or they see it, they respond to it, but it's just too much information, so they're moving on. So that can help cut out we call it microdiagnose. So it's like, okay, if you get that in understanding, then you can you can have a much deeper understanding and appreciation and you know moving into those different
And I think from what I gathered from your talk, it's like being armed with that data, then you should be able to craft better content from that, knowing what people are responding to, how their habits are shaped.
And they do and they do. We see that they're doing it. The problem is that getting those data traditionally at least has taken time and a cost. You know, every time we do did this test, the classical problem was the clients would say, I need the responses yesterday. Yeah, you know, it's you have a sixty K project and it's gonna run for three three weeks or something, and they're like, Yeah, that's that's great, but it's gonna be used for the next round of the next round.
Yeah, I do feel like everything's accelerating. I mean I I think this is across disciplines is people want they're like, I want this faster and cheaper now, like that and like and accuracy I think always was kind of a moving target when it came to that 'cause people just want something to act on. And I I I do think this is now accelerating that 'cause we're awash in data. We're awash and being able to turn things around quickly. so it just leads to just a deluge of data and I think a lot of people misusing how to interpret it.
speaker-0 (10:57.922)
I think one of the things that we see as a result is that people tend to get hyper optimizing for view through rates or some kind of vanity metric or we see there's kind of we need click through rates or whatever, right? Impressions or cost per impression. And it's does that really tell what you want to do? And I think that and there's a lot of pushback these days towards going towards branding as well because if you put hyper optimize for myth funnel metrics, you're gonna be at the cost of brand communication. And we see that kind of pushback today, the CMOs and even all the way into the boardroom.
is that people must start to have a much higher appreciation of branding and the value that gives.
I yeah, I see that. so one of the things you said in the speech was that generative evaluative AI doesn't fix advertising and it actually accelerates the same mistakes, which I've seen, you know, maybe not in advertising, just in across disciplines again, like whereas because it's a powerful tool that can create quickly, it just almost instead of fixing anything, it just gives you ten times the amount of problems, a hundred times the amount of the same problem. So can you elaborate on that?
Yeah, so I think that we we tend to have this idea that we could just brute force our way out of it. I mean a gazillion different versions of the same app, we're still gonna test on the audience, right? And I think that having some we could we can call it pre flight intelligible systems like even before you start doing your A B test, what about fixing someone taking out someone that's trash? Take out some of the bad stuff first, right? Is it overloading? Okay, let's fix that. Yeah, do we have a low likelihood of being seen as a brand? Okay, let's fix that.
Is it likely to have low engagement? Okay, let's fix that. So if it's possible to predict those things, that's one of the things that we have that we need today. And the technology allows us to do it. And I think that the problem today is that people they tend to be so hyper optimized on hyper focused on LLMs, so attaching Ts and Clouds for generating copy and ideas and stuff like that. And then generative AI for, you know, image and video production. And it's so easy to do. so execution has become really easy.
speaker-0 (12:59.724)
That the actual evaluation and judgment of it has not kind of followed trace.
Yeah, I mean I see that across my work as well where remember when emails used to be three sentences and now you get like paragraphs because they'll be fed it into an LM that did a good job of like explaining it, but again, don't have time to do all that. I'd have to feed that into my LM to give me the three sentence bullet point, you know? Exactly. so another thing you brought up is ninety-three percent of marketers are using AI to generate content, as you mentioned, you know, videos, still images.
And ninety percent to speed up production, right? So that's like the vast majority of people are using this to be faster better, right? so if the speed isn't the problem, then what really is the problem?
is good judgment. I think that we've kind of been parking judgment more and more. It's it's it's almost like we think that because it's it looks good, right? That's the evaluation criteria we usually have. And I think that when you look at how marketing works, it's it's a round table discussion. It's like, I like this one and I like that one and then it ends up being, you know, the CMO or someone comes in with a very kind of cla clear voice has to say it has to be that one. It's hard to argue against it. And I think that's the problem today that it
When you have the time, if you have the time at your disposal to get some data, we all want to be data driven, but it's really, you know, it takes time to get those data. But now there's actually no excuses anymore. And I think that we see that shift today is that people starting to realize that we can get these data earlier on, but we definitely need to kind of need to incorporate it much easier and much easier into our process. And that's part of the hardware.
speaker-1 (14:36.662)
Yeah, for sure. And one of the so I asked you this question during the the your discussion, which was about how Europe is now r requiring AI generated ads to be labeled. it's funny because I have mixed feelings about this. I do think there is a huge space for transparency, especially when it comes to advertising. for example, if somebody has a like and I've seen this already online, this is a product. It doesn't exist yet. You know what I mean? They're selling something that isn't even real. And so like
Obviously there needs to be some guardrails, some legislation, some restrictions, but like, you know, tell elaborate on that a little bit of how you think like the problems that it it raises, because one thing you did bring up is and I think s a lot of the solutions that are great with with a using AI is in the development architecting process stage to save time and money, right? Down the pipeline. So would you then have to label that on your generative AI? And I so please elaborate on that.
I think the problem today is that we we can't tell the difference. The assumption here is that we can tell the difference. Almost like the Pepsi Pepsi versus that people can tell the difference, but the moment you tell people that they're giving you know you're giving them Coca-Cola, I mean you can keep them Pepsi Cola. They will like it more. So so I think it's basically the same thing today. Is that people, first of all, they can't tell the difference. I think the assets themselves, most of them have some components of AI along way. It's like it might be the
The Pepsi Challenge?
speaker-1 (15:52.236)
Yep, yep, yep.
speaker-0 (16:03.63)
It might be the brief, it might be the actual visual or part of the visual. There's a lot of different things that could be influencing this to some extent, right? And I think it's a moving target as well. I think that you get AI as part of Adobe, Canva, or wherever you rank it, as part of Figma. So even today, you know, putting guardrails up for what works, what is AI generated or not, it's just gonna be really difficult. And I think the problem is also that the moment you tell people this is AI generated.
You have this backlash information, you backlash effect that people like it less. And you could do this. I don't think there's any studies doing it, but we'd be a cool study to test, you know, give people either AI generated or not AI generated, but then half of the time to label it AI generated or not AI labeled. And some say that it's just driven by people's expectations, not the actual fact, right?
See yeah, see how they wrote.
speaker-1 (16:55.912)
I highly recommend you check out a subreddit called Is This AI? and because what it'll do is people will they're not sure. So they post on there and then there'll be comments. And occasionally you'll have somebody who's very skilled, like a technical person who understands it and will be like, It is because of X, Y, and Z. And s oftentimes though it's just opinions. Like and it's for the craziest reasons. I love that subreddit. Hold on, just want to make sure we're recording still and everything's good. All right, we're awesome. so where was I?
So we did Europe. Okay. so this is more into like the brain neuron side of it. but if AI generated content is accelerating the wrong things in some ways, like what are the right things? So what are the neurosci what does a neuroscience say about like what actually works on a human brain?
So we know that, you know, so we we have this I've always, you know, even though I kind of s stopped my tenure position at the university and the business school, I kept being an academic. I'm still publishing papers, editorial boards, I'm still doing the whole thing. I still work very closely with academic institutions. We we're running an FMR study with Yale now, for example. work very closely with Stanford, for example. So we actually developed together with Stanford a we call it a four par model.
good, yeah.
speaker-0 (18:08.214)
Because have all this complexity from neuroscience, there's so many things you can dive, dig your head into and it's like it's it's it's a never ending thing to go into, right? So how can we make something that is so technic technical deep and so scientifically deep into something that is meaningful and useful and applicable from marketers? So I worked very closely Professor Baba Schiff at Stanford and we came up with this four power model to say what are the four powers, the four things you need to work on if you want to succeed in branding permanently. So we said, okay, stopping power.
You need be able to stop and sustain attention for a certain time amount of time to even generate some kind of response. The second is we call that persuasion power. So you need to persuade or you need to convey something that resonates with people. And the th third is what we call transmission power, which is cognitive responses. They need to be understood. You need to connect the brand to some kind of association and finally have locking power. You need to be able to lock down memory so people remember it so they can apply that mem that memory later on as you know.
As you stand on you know when if my wife asks me to buy some toothpaste, I'm not sure g gonna pick some random one. the first I'll thing I'll think about is probably Colgate and then she probably would say, No, no, no, it has to be Colgate, you know, simple fresh and you know organic and fair trade or whatever it's like.
Yeah.
speaker-1 (19:23.797)
Minty fresh, yeah.
speaker-1 (19:28.822)
Y my wife and your wife would get along splendidly. I could say that.
Side will not come back with the Zendio.
Yes, exactly. So that's fascinating 'cause this is essentially a a theme that's been in every session I've been in, which is creative quality keeps coming up. and it's essentially the human perspective and you know tastes are always paramount. So like we have all these new tools. It's amazing. Yeah, but it still comes down to the that what you just said, everything you just listed there is so subjective. Yeah.
Denominator, right? Yeah.
speaker-1 (20:01.718)
Right? It has to have a person, a human who has good taste in these things, a good understanding of the domain to be able to take that and execute.
Yeah. And I think that's so we've kind of been able to show that and we've built that model with with the Sanford and that four pound model and that was kind of what we lived out as consultants. And but then when we were able to predict certain types of responses, it kind of followed very neatly that response pattern as well. So it's like okay, first you need to get attention, then you need emotional responses, comment to responses, and then you need to land the cement.
It's like the jazz, right? You have the structure and then you can just gotta go all you know, improved but jab around. Yeah. Yeah. That's amazing. Cool. so this is kind of forward thinking about where the industry's headed, because I do think we're in such a fascinating time. and so if you could change one thing about how the marketing and insights industry uses AI right now, like just one practice or one assumption, what would that be? You could do two or three if you
a tough question. I think here's a picture how kind of emerges for me, especially just over summer. Just over summer. We I I internally at neurons we say that the earth is shifting under our feet right now. We it feels like it's it's a it's a sort of earth slide or it it's it's changing. And it's rapidly rapidly changing. And two things gonna happen with that. First thing people are gonna freak out. That's one thing.
Noted.
speaker-0 (21:27.372)
The second is that there will be a big variance in how much people freak out. Some people will use a you know be able to turn this around to something constructive, and some people will kind of dig down their you heels and just stay where they are, right? So I think when we talk about technology adoption in general, you know this kind of curve of technology adoption. You see the the innovators, the early adopters, you see the then there's a break to the early majority and the the la late majority and the leggers, right? So it's like people are different in how they typically adopt new technologies.
And it might also be that, you know, you are very an early adopter when it comes to audio technology, for example, but very late when it comes to something else, right? so when it comes to this type of technology, it kind of it just exacerbates and explodes that field currently. What we're seeing is that companies on the one side they understand that it's moving fast, that's a lot of opportunity, and that all their competitors are using it, or maybe you're not considering using it or expecting to use it. So there's a lot of kind of stuff going on.
At the same time we see that people have a very poor overview over you know which solutions do we have available. They don't really have a great solution to how should I try stuff out? How should I get an overview? How do I vet things? How do I onboard things? How you know how can I make sure that in within my company that people are adopting and using and stuff.
And then also how do I sunset and you know look for something else? So the big ask for me would be is there a way that we can take some of the learnings we've had in the past for technology adoption and help companies be better at doing that now? Because now is the time more than ever that we need to do.
speaker-1 (23:11.106)
Yeah, I I and I'm watching it firsthand roll out across and it's interesting because smaller companies are so quick to adapt and because out of desperation in some ways, new things that come around. Will this make me more productive, quicker, smarter, whatever? Whereas larger institutions, you know, changing a process is such a disruptive, you know, event. And so like I'm watching it roll out firsthand for I'm up to my eyeballs in like
innovations and and new things that I'm adding constantly. Whereas I have to now have meetings with leadership constantly to be like, What what are you doing? Like please keep us in the loop. Because they, you know, they're not because they I guess what I'm saying is the innovations come so fast and furious from smaller teams now that the big organizations have a hard time having any policy in place to how to integrate it. So I do think I agree with you. I think that there there needs to be some kind of like instructional hand holding
on how to bridge the gap between like bringing this stuff online. And from my experience, like s this is the advantage smaller companies have right now, is like if you can adapt fast and quickly and get there fast you know, sooner, I do think it's gonna be very destabilizing.
think that we look at kind of Clayton Christensen's kind of old kind of innovation dilemma problem. It's like you have this disruptive innovation of, you know, smaller companies can be on that exponential traject trajectory, but you have the bigger incumbent of the the bigger companies tend to be on a more linear path, so to speak. We call that kind of the linear path of doom if you like. Yeah so and they they you know they they are so blind to that uptake on the exponential curve that by the time they recognize it it's game of
Yeah, so I started in the market research industry two years ago, did not come from that world. It was interesting, so I had to learn a lot, like you know, while I was building, right? While I was flying the plane, I was building it. But one of the things I was warned about from people who are in the industry for decades were like this is a dinosaur of an industry. It's like change happens slow here. So it's fascinating for me to see kind of it colliding all at the same time. 'Cause I do think some in amazing innovations are happening.
speaker-1 (25:11.63)
And but meanwhile I'll I'll have calls with clients who are like everything's still done on Excel spreadsheets. You know what I mean?
Yeah. And they they want SharePoint.
Yep. Yep. It took that long just to get SharePoint up and running. They don't have to redo it.
And I think that, you know, for for what it's worth, it's like that has that has intrinsic value. And I think that the way I see it is like one of the predictions I have is that we'll see a bifurcation or just see signs of it now, but it's a bifurcation in the way that advertising will work. You know, nobody is getting really excited about making just another banner head. I think very few people. So that would be problematic. That would be a very kind of AI speaking to AIs, basically. And that would be kind of a quick learning curve. I see that as becoming more like
That ship has sailed.
speaker-0 (25:55.99)
in the you know fifteen years ago when trading was I don't know twenty years ago maybe when trading became AI driven. Yes.
You never could do it quick enough.
Or about how how short the port could be from Wall Street to your UI, right? I think we'll see the same thing here. Basically see the same kind of betting going on here, the same algorithm kind of developing going on there as well. And then we'll have the big campaigns, like the the big ones that I want to go I want to stand on can and get a price. I think that would be the bifurcation.
Yes, yes.
Yeah, I that that actually makes a lot of sense. I mean that kind of big value is never gonna go away, but like that's quick the the quick ones is gonna be super accelerated. Yeah. That's actually something that comes up across disciplines too, is the you're you're now conducting an orchestra, right? so you said insights are arriving too late, too costly, too complex. So what does a better model look like? Like what does the research process look like when everything's working properly?
speaker-0 (26:40.462)
Orchestration.
speaker-0 (26:59.362)
This is the optimal kind of the the the Goldilocks zone here is the ability to have high quality predictions because you need to be able to do this pre-flight, you need to be able to do it in the moment you need it. And don't wait a week or two for for getting results. So it needs to be something that can predict some outcomes that are valuable to you in the moment you need it. And it needs to be at the you know the speed you need it. So high quality, high speed. And that's kind of the breaking kind of the trajectory we've been on so far. Whether you have to go for high quality and low speed.
I
Yeah, it was you had the the balance, it was the C Salt, either one or the other. Yeah, exactly.
It's like going through it, single versus noise. I think that's part of the problem here as well. and I think that with these algorithms, the it all comes down to the quality of the predictions and the things you claim you can predict. That you know, you would like to predict everything in anything, right? But I think that being really honest about what you can predict is really important and demonstrating it. I actually just published a paper in actually pretty a pretty pr prestigious scientific journal, which is called Frontiers and Human Neuroscience.
Where it kind of lay out the groundwork for, you know, how do you actually build models like this that are high on validity, high on reliability, so you can trust them when you make those predictions. And it basically comes up with a kind of a recipe of five stages that says, Okay, first you need to build good models, good metrics. You need to show that they're valid, show that they're reliable. You also need to be able to show that they're kind of I call it able to kind of differentiate. So you can't test five ads and they all score fifty, right?
speaker-0 (28:31.734)
Need to be sure show me something that is differentiating between that as well. And then you also need to show that you're able to predict those responses as well. And that's kind of a five step step ladder if you want. So you create those models. and if you don't, well they have you've kind of failed on the that kind of framework, if you like. so I think that being able to predict, show that it's trustworthy, you can document it. You don't need to spill all the beans, you just show that it's it's actually done the right way. then all of a sudden you
are in that corner that says high speed, high high accuracy. You can't predict everything, but you can speed you can speed up the accuracies of the things you want to predict that are possible to predict first. So there's attention for
That yeah. I I really loved your presentation 'cause you sh had the visual examples and I wish I could show it on the podcast, but it was of how you're using this technology to do like, for example, generative AI with your prediction model to say like, okay, here's the ad, this is what we predict. Like you're not gonna get too much attention. Obviously the person's face instead of the product is gonna get looked at first, that's where their attention goes. So then you generated all these versions of it that would improve it with the score that's like based on real data. And I was like
There you go. You just save so much time, so much money. And like to me, that is essentially how I when I come from the film world, right? And in the film world you had storyboards and you would have scripts and script notes and breakdowns and you would do all the prep work before you showed up to do the production. And I approached it the same way when I build product, right? I'm like I s I scope out, I engineer, I d I meet with all the stakeholders and I get as much down predicting w how it's gonna work before we even start building anything. And I feel like that to me, good use of AI.
Right. So I would love just explain a little bit of how that model works with what you built. Because I think this is truly like a great showcase for how you can use this stuff properly.
speaker-0 (30:18.594)
Yeah, I think there's there's three steps to make it simple. We call it no for lack of you know better I use several times different type of word, but let's call it predictive, suggestive and generative. Generative we already know. suggestive is basically an insights LLM if you like that knows and understand all the knowledge, that understands
what you're trying to achieve and comes up with an insights and recommendations. The predictive one is basically what we talked about so far. It's like is it possible to predict certain types of responses for people? And what people are paying attention to and what they're gonna r how they're gonna respond to everything. And you have some benchmarks on that, what is a good score for people? So that's kind of the first step. It's like being able to accurately predict that. So it should be as accurate as testing a hundred people, hundred and fifty people, whatever makes sense, right? So if you take a hundred and fifty people and you run the same ad through
the AI should get basically the same results. That means that you can predict predict stuff in seconds if it's an image or a few minutes if it's a video, as if you're testing people, which would take a few weeks. Stuff that's the
That's crazy. That's the craziest that is. Yeah, I was gonna say the Harry Seldon, whatever.
Yeah. Minority report or whatever we want to call it, right? But it's possible. The second is that is then, as I mentioned, the suggestive AI, that's basically an LLM with a lot of context. So you can now feed it with the context of your brand, your brand design guide, your preferences, your you know, it could be your audience information, whatever intel you have on that. It also understands the scores and understands the neuroscience, and it's also pre-fed and pre-trained on you know behavioral science, psychology, neuroscience, and stuff like that. So it comes up with recommendations.
speaker-0 (31:57.878)
recommendation for how to improve the assets based on that. And then you can take that. You can take that brief, treat that as an updated brief to your creative designer and design team. But you can also bring that to a generative AI that understands this and it comes up with suggestions for things that can be improved based on that. And it shows you all of the examples that
And then you can take those examples, give them to your designers and say like that's what I mean. Like I just feel like it's how I use AI in my day to day, which I see I'm like, wow, the fact that you were able to implement this in a in a a pipeline workflow is just to me like a chef's kiss of like really how to integrate this in and make it valuable.
To the point sorry, to the point of kind of this programmatic loop that we'll see as well, you can imagine this becomes an iterative, almost like evolutionary loop as well, because it learns constantly as
It'll get sharper and better as you the more you use it. yeah, that's so cool. so if you were gonna tell like a CMO or a head of insights or something who's like walking out of your talk or maybe this entire, you know, endeavor we're on here, to do something different, like starting next week, is there any advice you'd give to do to get started doing something?
I think you know CMOs are are
speaker-0 (33:09.42)
pressured very much as he says by the board, by the CFO for example, to show value. I think that more now than than ever to progress. And I think that the first thing they can do that is actually becoming very, very mobile is one to to utilize some of the proofs that are out there that shows the value of doing AI and certain types of AI solutions that brings, you know, ROI for mid funnel metrics that shows it for branding, for example. The second I would recommend is the insights we've gotten from
Kids in report from Warg, for example, that shows the added value of branding. If you talk to a CMO today, it's like they have a very hard time understanding the actual value of doing branding. Because it it's hard to do the attribution for whatever campaign you do on branding to whatever sales number you have. That's such a kind of long delay and but there's a research now to show them actually build the work. and and then what I try to hit on as well is that there are new metrics available now for showing how you can do branding correctly.
okay.
Is it seen? Is it going to be remembered for example? All that is now becoming available through these new technologies.
Yeah, that's huge. Before it was like like you said, how long would it take to go from like, we think we're a pretty solid brand and then like what, three years go by and then it proves it? Like
speaker-0 (34:24.12)
Yeah, it's like fluctuation self-source model, right? So it's like it's it's a long and short.
Interesting. Well let's talk really qu actually I have a couple quick ones. let's see.
a lot of these are a little just brand I'm not iffy. Th well they were just thing like some tangents we could have gone on.
Yeah.
this is interesting. Well, the the idea that we brought up earlier of like consumer immunity to advertising, because you said like and I have to admit I'm very guilty of this. if it's like I'm I'm creating podcasts where we have like you know, pe people advertising on it. Yes, I'll skip Ford fifteen and go past it. If I'm on Instagram, like I scroll faster past just what I know are branded advertising, you know what I mean? So like is there is this like a reversible trend? Or is it just you think it's a constant battle between consumers who
speaker-1 (35:23.456)
need this stuff to f fund the the products they like, but at the same time refuse to even participate in the marketing side of it.
I think marketing is kind of less than cursed at the same time we're we're so optimistic that people love our ads, right? We're never gonna die on that up. We're just gonna w everybody is gonna love watching ads and then we don't really kinda pay heed to that the actual numbers and that. I think that what we should be much more aware of is how we can show who we are as fast as possible. Because if you only have two seconds at your disposal, you need to show who you are and make it worth staying a little longer.
So I think that first of all, being shown within the t first two seconds actually has value. It's just a reminder, hey, we're Colgate. Here, it's a tooth. so that's one thing. The other thing is like if I am Apple, if I'm someone else I have typically doesn't have a following that then I should show that as early as possible. Don't play too much and that kinda doesn't speak well to your kind of your video videographer that really wants to create a narrative arc and everything. Yeah, yeah. But you need to make it sure sure sure that people actually
pays tribute to it. It kind of shows value very early on, so they want to stay longer. And especially with when it comes to advertising, we have this almost like immune response that it's not always advertising. I'm just gonna flip off. Unless it's my brand. If it if it's something that is relevant to me, people will stay longer time. If that's two seconds longer, if it's ten seconds longer, that is less relevant because people already wanted to say their
yeah, okay. Yeah, take the win. Two seconds is fine. Is that we're here?
speaker-0 (36:53.944)
Well you can.
speaker-0 (36:58.254)
Example, we tested, for example, across all these media channels, and we see that Snapchat has the lowest attention by far. It's like one and a half seconds. People are just skipping through it. Even the the Snapchat users. But brand memory is at an all time high. Where Yes. I think two things. First of all, people have realized in doing advertising on Snapchat, we need to show who we are in the research system very quickly. The second is that they're very good at it.
Why you can you explain that?
speaker-0 (37:27.066)
is that people who look at it they're already like, Yes, I saw that. So people who are used to using the week being Snapchat they're recognizing as those sooners. It's not about kind of skipping always, it's just making quick decisions.
It goes back to what you said earlier about like yes the medium is changed in some ways, so optimize for that and you're golden. So even though you only have the two seconds, if you make a count it's a great two seconds.
People to stay for fifteen. People are happy with staying at two.
Okay. That's thank you. That's really good advice.
speaker-1 (38:03.086)
Yeah, it was better than fifteen, right? Better than thirty.
wasn't because the ninety nine percent of the time was watching the skip boat. Mm-hmm. So people didn't pay attention to the eye.
So no a quick five second ad that you can't skip is probably more priceless than you know than a thirty second one you can skip in five. Okay. Well tell us a little bit about this book. What is this? How to make people buy
My what is it? Third book. Yeah, so this is a book that basically is all about some of the learnings we've had in from the past in marketing and advertising. So this is based on both the case studies we've had, we've tested tons of stuff for you know mobile marketing association, advertising research foundation, all the big social media companies, for example, all the advertisers as well. And what have we learned? And I think that it's kind of a walkthrough of the four-part model I mentioned before, it's like attention, emotion, recognition, and memory. So, how do you measure it?
nice.
speaker-0 (38:52.3)
How can you understand it? And how can you actually do, you know, what are the do's and don'ts for each other's programmers? So that's basically it. But I try to I love I love reading and I love writing as well. And I love these books that allow you to kind of get a sense of not just the factuals, but also embedded in a part of the story. So it's like one of the stories I I mentioned. you know, I can ask this as a question. Would you buy scented water from Kern in Germany?
Scented water? Probably not me, no. From current, right? Yeah.
Yeah.
No, totally. That's hilarious.
It's just it's just the same. Yeah, yeah. So this is the kind of a story, it's like it's the same. Yeah. What we do we use this for? and I think this is like there's a longer story in the book about this. It's like where does that come from? It's an Italian guy that moved to the the city of Cologne and he came up with a new fresh order and he wanted to pay tribute to the the city that had greeted him and welcomed him. And so he called it Oidaton. So water from Cl from
speaker-1 (39:53.048)
Turns out great branding. That's fascinating. Well where can we find the book? Is it online?
Wonderful. And how about yourself? people want to get in touch with you, they have questions, how can they follow you, find you?
The easiest is also to go on neuronsync dot com. Okay. that's where you find both my company and myself and there are also links to my web pages and my books and the chain and easier to get a hold of me.
Wonderful. Thank you so much. This is awesome. Let's stop recording.
All right.
Comments