Skip to main content
Smith Business Insight Podcast | Series 4. . Episode 4 AI Reality Check

Marketing Unbound

Smith Business Insight Podcast

Generative AI is getting so good that the human touch in marketing may be a thing of the past


It’s hard to imagine what advertising and marketing in general will look like given the speed of development in the AI constellation of technologies. Today, wholesale automation of marketing functions, hyper-personalized targeting and predictive forecasting. Tomorrow, AI bots on sales calls? 

This episode takes a decidedly analytical look at AI-enabled marketing with guest Ceren Kolsarici, Ian R. Friendly Fellow of Marketing and director of the Scotiabank Centre for Customer Analytics at Smith School of Business. Dr. Kolsarici talks about the speed bumps that could slow AI adoption and shares what marketers need to do now to lay the groundwork for what’s to come. She is joined in conversation by host Meredith Dault. 


[Music playing] 

Meredith Dault: Way back in 2018, which in AI dog years is several lifetimes ago, Lexus created an advertisement completely scripted by its AI system, IBM Watson. Watson was good enough to analyze 15 years of award-winning car ads and then identify which elements resonated most with audiences. Well, that was five years ago, so just imagine what can be done today or even five years from now. 

Frankly, given the speed of development in the AI constellation of technologies, it’s hard to imagine what advertising and marketing in general will even look like. Today, the wholesale automation of marketing functions and hyper-personalized targeting. Tomorrow, conversational marketing with ultra-sophisticated chatbots. If you’re in marketing or even just a consumer, you’re either giddy or terrified. 

Welcome to this episode of AI Reality Check. I’m your host, Meredith Dault, a journalist and media producer at Smith School of Business, and today we’re talking about AI in marketing and sales. And we are in the best of hands with Ceren Kolsarici. Dr. Kolsarici is an associate professor, the Ian R. Friendly Fellow and Scotiabank Scholar at Smith School of Business. She is also the director of the director of Scotiabank Centre for Customer Analytics at Smith. Welcome, Ceren. 

Ceren Kolsarici: Hi Meredith. 

1:20: MD: Ceren, you’ve often talked with Smith Business Insight editors about the need for marketers to rely more on analytics and less on their instincts when making spend or channel decisions. And you’ve shown that marketers base 70 to 80 per cent of their decisions on heuristics, essentially easy-to-apply rules of thumb. Is AI just a smooth transition from the discussion around analytics or does AI represent something more revolutionary for marketers? 

CK: I think that’s a great question. When I look at AI, when I teach AI, when I talk about AI, I perceive it, or I discuss it as analytics on steroids, because AI really deals or knows how to deal with big data, which is the type of data we have now, the data that has a lot of velocity, the data that has a lot of volume and that’s coming from lots of different sources and has a lot of uncertainty. And with typical analytical methods — the econometrics methods and the statistical methods — we wouldn’t be able to handle the complexity of data as we have today, as companies have today. I think that’s a good way to perceive AI. Also, one thing to consider would be which disciplines AI draws from. So, AI heavily draws from statistics, which has been around for 360 years. 

AI draws from computer science, and, you know, computer science deals with talking to computers and translating our wishes to computers so that they can understand and execute. But a lot of the times, computer science problems are very focused, and they are small, manageable problems — very specific tasks. So, AI also draws from machine learning, which is probably the black box of AI. Whatever happens in terms of the technical sense, most people would call it machine learning. What machine learning does is takes a computer science problem and makes it much, much broader. So it can handle a really difficult computer science problem with the help of statistics. The ultimate goal of machine learning is that systems can learn from their environment and function on their own. 

So how does AI fit into this puzzle? AI is really relying on these three disciplines but making things more accessible to the outsider. Someone who doesn’t know any technical details, when they hear it’s part of AI, it’s an AI application, they are less intimidated than when they hear it’s a machine learning application. There are several memes on the internet. Sometimes I show it to undergrads and sometimes my professional students even. It’s really the same thing, statistics, and when you frame it differently, it’s called machine learning. And when you talk about it to the audience, it’s called AI. So, to answer your question briefly, it has a lot more impact. AI has a lot more impact because it’s domain agnostic. It impacts health care, it impacts production, it impacts the creative side of marketing, it impacts workforce. And all of that is very unique to AI. That’s why when we discuss AI, we discuss it as a part of general purpose technology. It changes the way people live. It changes the way we operate in our day-to-day lives, as well as in the workforce, as well as how companies operate in terms of internal functions and how they interact with consumers. 

5:08: MD: So what would you say to this question: Does AI represent something revolutionary for marketers then? Are you seeing that? 

CK: Absolutely. And not only for marketers, for the world. We have had so few general purpose technologies. I think the last one was the computer. Right after that was the internet. Maybe before all of that there was electricity. So AI is going to have an impact as large as these, even larger. So absolutely revolutionary for the world and marketers as well. 

MD: So let’s pick a product category. Let’s say we’re marketers and we’re picking a category — cars, yogurt — something like that. What will marketing these products look like in five years? 

CK: You said five years, but in your introduction, you mentioned it’s in dog years. Five years is an incredibly long amount of time in the AI space because the changes have been exponential. If you think of ChatGPT, I think it reached a hundred million users in seven weeks. So we’re really looking at an exponential curve. And, as people, we’re not really good at exponential curves; we really didn’t have to use it at all in our evolution. It’s kind of tricky, right? What’s going to happen in five years? We can look at a smaller time frame, I think, because even I don’t know, as someone who is an expert in this field, what’s going to happen in five years. If you start with this closer timeline, let’s talk about the metaverse. 

You probably heard of metaverse which is, yes, related to marketing, but also related to several other domains. What is metaverse? It’s an alternate digital space, digital universe where companies can present themselves, consumers can present themselves, individuals, workers, employees can present themselves in their digital identities. Brands have already started using this notion of metaverse to reach consumers in this alternate universe. So that’s one thing a lot of brands are dealing with. The big question is how much should they invest in the metaverse, because we still don’t know whether it’s going to stick or when it’s going to stick. One of the challenges with technological innovations like AI, disruptive innovations, is their diffusion pattern is not the way we predict this bell shaped curve. A lot of products diffuse with a bell shaped curve where, you know, innovators adopt a product and then early adopters, and then the early majority. 

With AI, and not specific to AI, but disruptive technological innovations … Geoffrey Moore published his really famous book, Crossing the Chasm, in 1991, way, way before AI. But he talks about this chasm between early adopters and the early majority. So for these kinds of innovations to reach mainstream populations, the population needs to be ready. One challenge is how do you make consumers ready to adopt products in the metaverse? 

MD: I was going to say, does the metaverse work if I don’t go to the metaverse? 

CK: As a consumer? 

MD: Yeah. 

CK: No. So one example of how products use the metaverse or brands use the metaverse, for instance, I don’t know if you have kids, I have an 11 year old and he loves video games. He has very limited screen time, but he loves it, and he interacts with his friends in this video game space. So that could be an example of metaverse. 

And McDonald’s is trying to launch a metaverse store where consumers can order their food while they are playing a video game, but then the food comes to their home. One of the ways we can have consumers warm up with this notion of AI or brands in the metaverse would be linking the digital brand in the metaverse to the actual consumption of the product. And what McDonald’s is planning to do is a great example of that. I guess the big question for brands then is to figure out how much to invest in these initiatives, how much to be a market leader in these initiatives and to know when to pull the plug. Disney just removed itself from the metaverse recently. I’m sure it was a big decision. So, it’s like trading off revenue generation and the cost, and AI can benefit increased revenue if you create lots of consumption opportunities in the metaverse. That’s a good example. It can also benefit from firms by cost-cutting, but it’s also risky because of the uncertainty. When is the population going to be ready? 

10:18: MD: Right, and you’ve talked about pretty big brands, McDonald’s, Disney. How many of these kinds of marketing tools are going to be accessible to average businesses? 

CK: That’s also a wonderful question. When you think of brands who have a lot of resources, resources both in terms of human capital but also in terms of financial resources, they would be at the forefront of these innovations. And the small brands, startups, brands with restrictive resources, they will be watching and seeing whether any or whichever ones of these initiatives are going to stick. 

We use a terminology called “MarTech”. So MarTech is the collection of firms that serves companies when it comes to the technological side of marketing. AI is one piece of that. And if you look at the MarTech landscape — and this is pretty shocking for somebody who’s not in this space — in 2011, in the MarTech space, there were 150 vendors. And if you fast-forward to 2022, now in the MarTech space there are 9,000 vendors. So it is more than a 5,000 per cent increase in a matter of 11 years. So, to your question, you can outsource some of these functions to these MarTech firms. But you have to be more careful because you have less resources to spend. So you have to make sure the ROI of the investment that you choose to take on is going to be worth it. That’s why the watchful waiting makes a lot of sense for these little companies that don’t have a lot of resources. 

12:15: MD: Got it. So how much of this do you think is going to be felt and experienced by consumers? I think about your 11 year old who’s growing up with the idea that you can order McDonald’s while you’re playing a video game in the metaverse. How is that going to trickle down to consumers’ habits generally? So think about someone, they want to renovate a house, a procurement officer wants to buy office supplies, machinery for a business. How is it affecting consumers? 

CK: It’s going to affect consumers … Well, you talked about different segments of consumers. So, obviously, generation alpha, our kids are going to be a lot more open to these changes, but they don’t have the income, right? They don’t have the financial resources to spend. So companies need to go after us, people who actually have the disposable income to spend. But I perceive AI to impact the whole consumer journey, the whole funnel, so from the awareness stage to consideration intent and the purchase stage. I can give you several examples. One piece of AI that brands tried using and still are using is called NFTs, non-fungible tokens. So NFTs are digital merchandise, digital items that consumers purchase. In fact, there is a platform for just NFT trading, it’s called OpenSea, and a couple of years back, NFT trading in OpenSea was about a few billion dollars. 

So there’s a lot of revenue potential for brands there. I’ll give you just one example of NFT. This is sort of an extreme example, but it will, I think, give you a hint of what kind of revenue opportunities are there but also give you an indication of how much brands need to jump to convince consumers to make these kinds of spending decisions. So Jack Dorsey, Twitter’s founder, his first tweet, just this one-line tweet, the NFT version of it was sold to a Malaysian businessman for, I think, $17 million a few years back. So, what can this businessman do with that NFT? He can present it in his home. Definitely it’s going to be a conversation piece, so he’s going to have some social capital gained from that purchase. But really, it’s hard to grasp or wrap our heads around as normal consumers. 

MD: Yeah, can I stop you there? 

CK: Of course. 

MD: So literally he owns that tweet now. He has a visual, digital ownership of the thing. 

CK: Yes. 

MD: So he’s sharing an image of the thing that he has purchased for millions of dollars. 

CK: Yes. And remember I mentioned combining the virtual world, connecting the virtual world to the real world. There’s something we call digital twinning. And in this Jack Dorsey tweet example they created an actual physical version of the tweet through digital twinning. So this Malaysian businessman can, you know, exhibit that in his office or in his living room so at least there’s this connection to the physical reality as well. But still, you know, if I purchased an NFT dress, a designer NFT dress, it’s going to be one-of-a-kind because that’s what’s unique about NFTs. They are one-of-a-kind. 

MD: Can you wear the dress? 

CK: Yes, I can wear the dress, but as my digital self. So I cannot wear the dress during this podcast, unfortunately. But if we were doing this podcast in the metaverse, I will be able to wear the dress, or, you know, when you ask me about how AI is going to impact consumers, many brands are now doing their advertising shoots and print ad campaigns using AI models. So those AI models would be able to wear that dress or would be able to wear any brand of any dress in NFT versions. And as long as we can connect our digital selves to our actual selves through these kinds of creative means, I think it’s going to speed up the diffusion of these next new technological innovations, metaverse or NFTs being the two of them. 

16:50: MD: Wow, that’s fascinating and, also, I have to confess, a little bit disturbing because it just assumes a certain amount of our life living in this digital realm as a consumer. 

CK: Yeah. But we’re going there. I mean even I am, I’m kind of a traditional person and I prefer in-person, I don’t really prefer Zoom meetings. I prefer in-person meetings over Zoom, for instance. But even I can see that that’s where the world is evolving. And one of the key things brands can do is acknowledge it, embrace it, and see how they can improve their operations from both an efficiency perspective and effectiveness perspective. 

So that’s another benefit of AI: efficiency. It can help us do things faster. It can increase our productivity or brand’s productivity or effectiveness. It can help us do things better. Right? And AI can, it’s not either/or with AI. AI helps in both. A lot of times we do it faster and we do it three times better. If you think about it from an advertising campaign perspective, you will do your photo shoots or campaign production in a more cost-efficient way, but it will also be a lot more effective in reaching the target market because you’re going to be able to show it at the right time to, to the right consumer with the right crafted message. So it’s kind of a win-win if you’re able to do it, right 

MD: Right, so we’ll know how many eyeballs saw that NFT dress, whereas you don’t know how many people looked at the magazine ad. 

CK: Absolutely. And I think one of the other benefits, because again, going back to your question about how this is going to impact consumers and marketeers, one of the other examples I can give you about AI technologies is blockchain. NFT is part of a blockchain technology as well. So blockchain impacts a huge part of how brands and advertisers deal with data. So it has positive impact on data collection and privacy, ad fraud and ad delivery frequency and timing. It really increases the efficiency and effectiveness of digital advertising campaigns. 

And you mentioned how many eyeballs saw it, and I’ll give you one really surprising statistic. When I saw it, it surprised me as well. Ad fraud amounted to a hundred billion dollars last year. So a hundred billion dollars, I’m going to put it in a scale so you can understand how big it is. It is about two-thirds of Facebook’s advertising revenue. So what does it mean? It means a hundred million dollars worth of clicks on digital advertising messages display or pay per click, they were made by non-humans. They were bots. So brands are losing trust in digital advertising. They, they’re explained ROIs are much smaller than the actual ROIs. And then they, they stop investing in these channels. But with blockchain technology, we can actually verify each of those clicks through consumers’ journeys. So the ad delivery will be verified, confirming it’s a real person, not a bot that saw the ad. So that’s a huge benefit of AI, but consumers are not going to … it’s kind of behind the scenes, but they will see the impact of it with the exposure to ads they want to see at the time they want to be exposed to those ads and through the channels they want to see the ads. 

20:31: MD: OK. So this is a great transition to this question about data, because we have heard about the challenges around quality and quantity, and we’ve also heard about black box AI systems that spit out decisions based on, like, nobody knows; we don’t know what their decisions are based on. So what are the speed bumps that could slow down AI powered marketing initiatives, in your opinion? 

CK: So I’ll answer that question, but I’ll start with the three, three big technological innovations of our lifetime. Your lifetime, my lifetime as well. What were there? The two big ones were mobile technology and social media, right? So if you look at the diffusion curve of these technologies, they are plateauing now, they are saturated. We haven’t seen huge changes in terms of innovative capabilities in social media or mobile technology in the last decade. So these technological curves are going to be now replaced with AI-powered innovations. What are they? Augmented reality, virtual reality, Web3, composability. And one of the questions you asked is how are these innovations accessible to an ordinary firm or an ordinary consumer? Composability ensures that we’re moving towards a non-coding environment. 

So, a regular person in a regular firm would be able to apply these AI technologies through an interface they’ll be able to manage. They don’t have to code it. And all of the information on the data will be stored in the cloud. You don’t need the software or hardware requirements that you would have needed otherwise to store and analyze big data. You don’t have to have the coding skills through composability. These are AI powered innovations but also speeds up the diffusion of all of these technological innovations. 

Coming to your question about how much we can trust AI or, you know, the speed bumps in front of the AI diffusion, I think the biggest speed bump would be how much we can trust it, because I’m sure you have experience with ChatGPT. It does incredibly well in certain things, but sometimes it’s also incredibly stupid. It makes so many easy mistakes. It almost feels like it lacks common sense. And I think this is one of the challenges AI needs to overcome or AI scientists or computer scientists need to overcome. How do we deal with AI lacking so much common sense? Because when we talk about AI, yes, it, it passes the medical licensing exam, it passes the bar exam with flying colors and all of that, which is wonderful. But if you consider an AI powered lawyer or AI lawyer, but then he’s at the court and he lacks, you know, common sense, how much would you be able to trust that lawyer, right? These are hurdles in front of AI, and I don’t see an immediate solution to these problems, even looking into the field of computer science. This is not an easy problem to solve. 

They usually talk about the common sense and AI mismatch as shooting for the moon. How do you shoot for the moon? You obviously don’t do it by adding one more layer to the tallest building, right? So you have to have some drastic thinking. And some of these mistakes AI makes, they are so easy, even like children can do them. One the things was … I asked AI, I’m drying my clothes outside in my in my backyard and it takes five hours for five pieces of clothes to dry. How many hours would it take for 30 pieces of clothes to dry? And the AI tells me 30 hours, right? 

MD: Right. 

CK: If I ask this to my son, he would be able to say it’s five hours. So then a computer scientist will tell you it’s … you just have to scale it. You have to train it with larger data, right, so it can learn. But it’s so unsustainable. It has a really large carbon footprint. It’s really expensive to train these AI algorithms. Somebody might wonder, why would we even attempt to teach AI common sense if our children can do it so naturally. I think once we resolve that issue of how AI contributes to humans, I think AI is there to augment human intelligence, not replace human intelligence. At least that’s the way I see the future of AI. So the paradigm of human versus AI or human or AI versus human and AI, the winning part of that paradigm is definitely going to be human and AI. I think that’s where the future of AI lies, and we have to just recognize it. 

26:09: MD: Right. So, given everything you’ve just said and bringing it back again to marketing and sales, what do you think, are we seeing people just handing over marketing sales functions to AI at this point, or we’re not? 

CK: Absolutely not. So I’m a firm believer in AI augmenting the human function. And yes, AI can’t do common sense. AI also can’t do a judgment. And I talk about this in my class, there’s a website called “Will Robots Take My Job?” And if you go there and if you write your title of your job, it gives you the probability of your job being taken over by AI. And if you write “marketing and sales directors” or anything managerial, the chances are less than one per cent. Why is that? Because of the issue we talked about; it lacks judgment, it lacks common sense. AI is going to increase our productivity and effectiveness, but it has to be within the framework that’s defined by humans because humans are going to be the part of that duo that’s going to be asking the questions. So what are the questions that you want AI to target? It has to come from a human. And then we also have to make sure the results are accurate, and it sort of aligns with some sensibility check. 

MD: Right. I suppose you can’t have brands just sending stuff out into the universe willy-nilly and getting it wrong. 

CK: No, absolutely not. 

MD: A little risky. All right, thinking about the future, I have to tell you, Ceren, I feel more reassured. I mean, I just feel like I’m not going to have to live in the metaverse forever. 

CK: No. We will be in person for the foreseeable future. 

27:58: MD: Oh good. Speaking of the future, a last question I want to ask you is about advice you might give a young person who, maybe somebody sort of 17, 18, they’re graduating from high school, they’re really excited about a career in marketing, but of course they’re looking ahead and anticipating a great deal of change. What advice would you give someone looking ahead at a career in marketing? 

CK: That’s a difficult question. And I get that question a lot. I’ll give you what I would tell to really anybody, any parent or any young human who asks me that. We have to acknowledge these technological changes. We shouldn’t fight with them and we shouldn’t be scared of technology. So there’s this huge concern about AI being a threat for human’s identity, AI being a threat for humans in the workforce and all of that. But I’m confident that your job is not going to be taken over by AI unless it’s a monotonous labour kind of job. But your job might be taken over by somebody who knows how to use AI or how to make their job better, make themselves more productive through AI. So, embracing these changes is going to be important. 

Also, for marketers, it’s very important to understand that AI is a really powerful tool. And it’s you, the key is you because you’re going to be the person who’s going to determine the scale of impact that AI can make. So I always give this example in class. AI could be a disruptive technology. So it doesn’t have to just increase your predictive performance by five per cent. It doesn’t only have to increase your, I don’t know, number of ad clicks by 15 per cent. Those are tactical improvements, which are great, but AI can do so much more than that. It can be disruptive. 

There’s this example with Amazon where they use recommendation systems on their web page. When you log on to the Amazon web page it will show you, Meredith, you purchase these things; we think you’re going to like these things. So, let’s say the predictive performance of the recommendation engine is 20 per cent. And AI is working, or the data scientists are working, to improve that. So a very low-hanging-fruit example is if I go to 80 per cent. Meredith is going to buy four of the five things, which is great. It’s going to increase our revenue for Meredith. But you can do so much more than merely using it as a recommendation system. 

So what Amazon can do then is to change their whole business model from shop and ship to ship and shop, meaning they know you’re going to buy four of the five things and they will ship it to you before you order them. You’re going to walk out of the door, you’ll see the box, and you will keep four of them and send the one you don’t want back. It’s even possible that you’re going to keep the fifth one because you know, you’re on the fence. 

MD: Because you’re too lazy to take it back. 

CK: Or you’re too lazy, right? So this is the disruptive potential of AI and a lot of the companies don’t think about outside of the box enough to see these disruptive potentials. And that’s one of the other recommendations I would give any young person, not just go with the low-hanging fruits, but try to see the potential these technologies can have for your company, for yourself, from a sustainable competitive advantage perspective. 

MD: It sounds like a really imaginative moment, which I would think appeals to marketers. 

CK: Yes. Absolutely. It’s a really exciting space to be in, to teach, to research. So I couldn’t be happier. 

MD: Well, it’s been a pleasure speaking with you today. Thank you so much. 

CK: Same for me. 

[Music playing] 

MD: And that’s the show. I want to thank podcast writer and lead researcher Alan Morantz, my colleague Julia Lefebvre for her behind the scenes support and Bill Cassidy for editing support. If you’re looking for more insights for business leaders on AI and many other topics, check out Smith Business insight at Thanks for listening.