D (3:26)

I'll just pick up because I'm going to say something similar and then different. Absolutely. I mean, there's a lot of talk about really sexy things in AI, but I think the things in CTV that are really winning right now are somewhat more boring. Audience and bidding optimization, I think who sees the ad is a really important piece. So smarter lookalikes, predictive models, incremental reach, all of these things are really important. Jonathan, you mentioned creative. I think in CTV there's still a lot more creative selection going on, but I'm hoping that it goes more in the AI direction, like testing multiple cuts, versions, translation a lot on the automatic versioning of taking it to 15 seconds and 30 seconds and all of that and the last part and again, boring. But it's nerdy and awesome. The measurement and attribution piece, like closing that loop on outcomes, I think is really important in this space and I think AI is really helping to do that effectively.

A (4:38)

Yeah. And Carolyn, I would say I think one of the great things about being able to test all those different versions is being able to create them. And yeah, in creative development, we're definitely using AI, we're leaning into it. But one of the things is we now have the ability to not just create one ad for many, but many ads for many. Right. And we can even dial in, we can get way more targeted and we can really see what's working on on a bigger scale. So we have an ability to lean into different times of day or different targets or different, you know, how weather. I think we can really get into being able to look at ads in a whole different way than we have before.

B (5:19)

So you guys are trying to tell me that the sexiest development in the history of technology, the most important part of it to see tv is going to be the not sexy stuff.

C (5:32)

That's where we have the most pain point. Nate, like we like operationally and you know, we talked a little before this, right? Todd, like the challenges with all these creative renditions is that it's operationally heavy to distribute all of that Media and to make sure that it gets trafficked and tagged correctly. That's human capital making all of that happen today. Because there is no industry standard that unifies all those operations across, you know, supply up to the demand side. So dropping AI in there, that's a massive cost savings for anyone that's trying to traffic creative and operate their network. If you're a publisher or operate your campaigns of your agency are all the way down to the brand.

A (6:12)

Yeah. And it's, it's cost, it's time, it's efficiency, it's being able to react. And I think we talk about all these things like we've, you know, we've wanted to do. I'm like, oh, what about if we're able to know, you know, what, what this, this person's doing last week or I brought up the weather earlier, like rainy days, like, but being able to actually have the ability to create different work and to do kind of real time, actually adaptations of the work. It's a big deal with production. And it's not about we're going to use production always to create the content itself, but it's almost using it for editing, using it for tagging, using it for a lot of things like that on top of the ability it has on the creative side.

D (6:58)

Well, and I also said this the other day. I think Nate knows this, but maybe not you. Like both of you, Jonathan and Todd, I started in radio a very long time ago. So I actually sat in a studio and I cut commercials, I made all those end tags. And to some degree, what we're talking about is a highly optimized and efficient version of like locally localizing that to the nth degree, but without giving away my age, because the lady never tells. Let's just see, it's in and around over 15 years ago. But like today, we're talking about versioning and personalization to a level that is by far more exceptional. What I heard from you, Jonathan and Todd, was business outcomes. And what the CFO and the CEO really cares about is business outcomes. And not so much the sexy press release. Yes, it's boring, but it's business outcomes.

B (7:51)

All right, well, I am going to give away my age, but people can probably guess anyway based on the fact that my beard is mostly white at this point. Point, you know, you guys are talking about personalized digital ads or localized digital ads. And I've been around the industry long enough that I've been hearing about these for more than 25 years, almost 30 years. And it, it seems to be One of the key things that people are talking about for AI powered ad creative as well. But then you see case studies and it, I mean, it doesn't necessarily look that different. Right? It's. I see, you know, I see dynamically created AI powered ads saying stuff like, we've got a, an allergy update alert for Austin, Texas. You know, how much of this is really new and how much of this is just sort of AI powered digital creative optimization that we've had around for a while.

A (8:45)

I would say it's. It's the ability now, I think, to do it on, on a different level, because I don't think it's new. It's what a lot of people have been talking about for a while. I mean, going back to like, you know, the most. I don't say the most boring thing. I'm sorry to all clients, but like the local, like local tagging things. Tagging things. That's in, oh my neighborhood. Oh, like, you know, you're seeing an ad in California, I'm seeing in New Jersey, but it's tagged more to me. I think it's the fact that it's not just that, but it's also being able to swap different elements of ads to make them feel more personalized and having the ability to if just say it's something and there's sports in the ad, and I know Caroline likes soccer more and Jonathan maybe is football more. There's ability to do things where you can now create many different versions of one of, of one ad in that way.

C (9:40)

Yeah, I was thinking that same thing. Just in the sense of elements of the advertising that may resonate with different cohorts of individuals. I think the challenge, Nate, is like when you try to do dco from the perspective of truly being dynamic down to the end user is it looks terrible if you get it wrong. Because imagine you're directing it to Nate and it's actually Sharon on the other side of that just going, I'm not Nate. Well, who's Nate? So you've got to have a really strong confidence in your deterministic identity. And maybe social platforms could do that. But then there's a whole nother slew of does that become a dark mirror episode when we focus on how do we align creative renditions to different cohorts? Because you know that someone may respond differently to say, a deep male voice versus a light, airy female voice. Those can be elements that can be dynamically swapped in on creative and determined as you're running media. Because the machine can learn and say, oh, okay, this cohort responds at a different rate than this cohort response. That's all possible. And it's not going to end up in a situation that feels creepy or that feels like, you know, who's Nate?

D (10:50)

Yeah, I think that increased risk is central to how specifically the advertisers think about that level of personalization. Because I always say on the one hand you've got the dream of what AI can do, right? Like DCO took pre approved assets and remixed them with AI, but people felt comfortable with that because they had approved the original building blocks. Right. And the discomfort comes when you're talking about the level of customization that Jonathan just mentioned. But then you could also do real time optimization because of course, Nate's a person and he changes second by second. And so what ad he's going to want to see three months from now might be different depending on whatever's going on. Maybe Todd, the weather. But that level of personalization, to his point, increases the risk and to a brand that that risk, especially in a TV environment, which is why I think CTV will be slower to really maybe like address the dream of AI personalization as faster as some of the other channels will, is what they're trying to mitigate. Because one really large CTV example and mistake could have large ramifications.

B (12:07)

Yeah, yeah, that's a great point. So, you know, really what we're talking about here is, is using AI to make ads higher quality or more effective. And obviously that can happen on, you know, on a response perspective. But. But it also could possibly happen on brand advertising as well. There's been a lot of chatter in the last week or two about Coke's latest efforts at AI generated Christmas advertising. And I just makes me ask the question, when, when are we going to get to the point where this stuff actually looks good and has a positive impact on consumers? Because right now a lot of this stuff feels very much stuck in sort of the uncanny valley.

A (12:54)

Yeah, I mean, I think the original Coke holiday, It was like 2024, right. When it, when it first was like two years ago, I believe the original. And you know, it was, looking back on it, I think it was pretty amazing what they did in the technology of where it was and where it is now. But I think the quality, there's a couple things going against is there's a lot of people who have an aversion to AI, like they just, they're, they're nervous, scared, they don't want it, whatever it is. So I think brands really need to Tread with the idea of someone like Coke. Coke is. Was the real thing. Do you want to be in the AI world? Brands like, like dub, like. So I think there's going to be the idea coming from the brand. Do you want to play in this world? Is this something that you think your target and the people that love your brand and your loyalists want, but the quality of it itself is it's getting better and better and it's. And it's going to be where it's most effective. Right. I think if you're doing things that are, we're using AI like we're using cgi, I think it'll be more accepted. I think when you start getting into the idea of AI people, I don't know. I don't know if people are ready for their brands to be living in the world of AI people yet. But the quality in so many aspects of AI is really getting good, where you're not going to be able to tell the difference between what CGI and what's AI in terms of, like backgrounds and worlds like that.

B (14:29)

I mean, I agree with you that even for what they did in 2024, the quality was surprisingly high. But it doesn't necessarily mean it was an effective ad or something that people enjoyed watching. Right here in New York, there used to be these guys in Times Square who would make paintings in front of your eyes using nothing but spray paint cans and they'd make these amazing moonscapes and you'd go, that's a lot better painting than I thought you could do with nothing but spray paint cans. But it doesn't mean I want to take it home and hang it in my house. And I feel like that's sort of where are with a lot of the AI brand creative. It's amazing you can do this.

A (15:06)

Do I want.

B (15:06)

But also.

A (15:07)

Yeah, yep, I.

C (15:09)

And I would. I just build on everything Todd said, which is, you know, when we think about what is the role like, it sometimes just helps us take a big step back, which is what is the role that the CTV marketing is really trying to play. And this is. I want to get in touch with my prospective customers. I want them to remember my brand. I want them to take action. And if you have a social environment where you think about where we see AI content popping up more and more now is in social media. And so we as a society start to become more sensitive to that. We say, okay, is that AI? Like, I saw that baby fight a, you know, a bear. Like, is that really a thing? Like, it looked real you know, and, and the quality has really improved. Like Open OpenAI Sora 1 vs Sora 2 is. It's incredible. I've seen some of the generated content there. You wouldn't know that it was AI except for the fact that producer said it was. But if that damages authenticity, then the branding message has failed. And the goal here is to get that outcome. As we said. Right. As Caroline pointed it pointed us back to, then this is where I agree with top like brands have to be careful and say, does me going full tilt into having a artificial influencer promote my product because it saves me having to pay some, you know, some talent rights. Is it worth that if it becomes well known because that AI avatar shows up on another brand? And then how do you protect that intellectual property? And like how do you make sure that if that spokesperson is supposed to be exclusive, but somebody else could just pay for the license to get them to show up on their marketing? Like there are all these challenges that are going to come from pure Gen AI, which is why I really agree with things that Todd and Caroline have said, which is we can use these skills to really improve post production, to cut down on costs, to change what evergreen means to content production. Because it used to be evergreen was something that you could always fall back on because it was produced and you can cut it up a couple different ways and it looked great. Now evergreen can be little components that AI can stitch together in any number of combinations and still tell a compelling and authentic story.

D (17:16)

I just want to pivot for one second about this Coke ad because something that's so funny to me about this whole thing is how much we talk about the technology versus like the ad itself. Like, was it a good ad? And it kind of reminds me I'm a big cyclist, hence tattoo about. And I'm sorry for the men on this panel, but it's usually men talk all about their bike. So like for a two hour ride, all I hear is what gear shifters and brakes and everything. And I'm like, I don't care. I just want to. I'm just out here to ride my guy. So it reminds me of all this chatter around the Coca Cola ad. I'm like pulling back, actually, the AI should be invisible. Is it a good ad? Do we think Coke and holidays and family connections. And so that kind of made me laugh because there was a lot of chatter about like, oh, Santa's hand wasn't exactly right and the truck. I'm like, who, who is analyzing these ads like this? And so to your point, Todd. I think it's because people have a feeling about AI, but then they forget that. Also, there's a human behind the AI, like Geppetto with Pinocchio. You know, the human decided what fuzzy creatures to put in that ad. So when people are like, ooh, the AI is terrible, I'm like, you do remember there's like a whole bunch of humans that are behind it, like pulling the strings. Yeah.

A (18:42)

Well, I think a lot of people think at the time someone just pressed a button and the ad came out. Yes, I think knowing like everything that was behind it. So, you know, one of the things I think a lot of people talk about that Coca Cola is it lost the emotion that a normal Coca Cola ad has. And in the end, that's what we're talking about. It's connecting. You're not going to get credit for using technology. We saw, we saw the metaverse where all brands were like, look at us, we're in the metaverse. It's like, great, what are we doing here? What's the purpose? What's the value? What am I getting an emotional connection with your brand from this? And AI is the same thing. It's a great tool and resource. But you cannot replace the human emotion. That's that real connection between ad and viewer.

B (19:35)

So, Caroline, I love your point about if it's done well, the AI is invisible. But I also know that there is at least some conversation around the industry and at the IAB about making sure that the AI doesn't become completely invisible. You guys, if I'm not mistaken, are, you know, are receiving requests from folks that we maybe create standards for disclosure on AI creative or AI in other parts of the advertising space. I'm curious where, where you guys are with that right now and what that would look like. And then after that, both Jonathan and Todd, I'm curious for your take on whether it's necessary and enforceable.

D (20:18)

I can't wait to hear Jonathan Todd's answer on that. So we have convened a working group of platforms and advertisers and agencies to really have deep discussions around this. But first off, I want to answer the question that comes to a lot of people's minds when we talk about do we disclose AI usage and advertising Creative. The question is why? And the answer is a few different things. One, consumer trust. We looked at a lot of studies that were talking about, do consumers want to know about AI usage? Also, it's a bit of a fragmented regulatory environment. Specifically in the US There are many state by state regulations versus full Federal guidance. And then of course in other territories like the eu, there's more, let's call it conservative approaches, but there's no what does the advertising industry think is the right way forward? And so we thought that it behooved us to think about a balanced approach to labeling. And when I say balanced, I don't mean any usage of AI anywhere from text to image to video. No, no, I mean a thoughtful approach about when a consumer could be potentially misled by not disclosing AI. And the one example I like to use all the time, because it's funny, is if I'm on a social platform and I see a virtual influencer selling me a skin cream because it's going to improve my skin. You know, obviously I'm a woman, I think about my skin and, and I buy that skin cream and then later find out it's a virtual influencer, I will be disturbingly upset. So that's the sort of why the how is a combination of to consumer message and also utilizing metadata to carry the message forward. Who's responsible? We're thinking about that. When to disclose. It's actually quite complicated, but I think it's something that needs to be figured out because when people say oh well, the regulators will figure it out. Well guess what, the policy folks always come to the business folks here and they say, what do you guys think? Well here we're going to give it to them what we think.

C (22:44)

I think that's the right approach, which is that we do better as an industry when we lead through self governance versus relying on regulation to come in. Because the reality is anyone that's in Washington or in a state capitol, they don't know the intricacies of our business. And so if we haven't set the right standard, then we'll end up with a standard that doesn't actually work well for either side, including the consumer. I think the question Nate was like, is it enforceable and is it necessary? So I think your example Caroline, is a great one. However, the, the contrarian in me says, well, how many of the humans that have promoted skin cream have actually used the product long enough to really be a representative of said product? But I get the sense like I think to start it might be necessary. I think as the industry evolves over time, we may find that the necessity goes away. I also think that we will see effectiveness of ads that have disclosures are going to be lower than ads that don't. And that's going to put pressure from the buy side to basically say look, you know, I'm only using a little bit. I don't have an avatar here. Why do I need to get disclosed? I think this is where having that discretion that you mentioned, Caroline, is going to be really important.

A (24:06)

Yeah, we're in an area now where we shoot things, we create things on sound stages and then we have them appear in the real world. Or you know, if we have somebody that's on strings and then we, you know, we, you know, Photoshop out the strings and then the ad appears like they're flying. So I think there's a lot of things now that we're, that we're currently doing that don't need these kind of, these, these, you know, legalities to it. I do believe when you start to over promise or you start to do things where you have AI people that are creating a reality that doesn't exist, that's, that's where I get a little, a little concerned and where you might want to have some, some type of legal lines or whatever it is to, to let people know. I mean, especially when we're dealing with kid advertising and things like that, it's going to be even more. Sometimes I feel like adults, if, look, if you don't know, then that's, then, you know, maybe, maybe that's on you as an adult. But I'm also a person that finds it funny when they, in a car ad when it says it's a stunt driver. No kidding. You know, so I, I just, I think it's, it's going to be in an area. But if, if we do make this, I think it has to be enforceable because I think, you know, then you have an unfair playing field. If, if, you know, some people are going by the rules and others aren't.

B (25:35)

But how do we do that? I mean, you guys acknowledge that the likelihood that ads that are labeled as AI generated are probably going to take a hit in performance. And we also know that the platforms and media sellers who are running these ads are not involved in the creation of the content. It's going to be challenging for them to know. Right.

D (25:56)

Who is responsible. And you can imagine who doesn't want to be responsible because of that reason. And I think that's fair. If you, you know, if you're Todd and you know exactly what was put into the sauce, then you know what's in the sauce. But if you're eating the pizza, how do you know what somebody put in the sauce? So I think what's interesting too, I actually heard both of you say this and we thought a lot about it. Even when I would say something like, like traditional tools like Photoshop and Pro tools and et cetera, just the amount of AI that is in all of those, I'm just using air quotes here. Traditional tools makes it difficult to even distinguish between, okay, what was used and what's used now. And so to your point, Jonathan, I think all of this is going to change in a hot minute. In fact, I have anxiety all the time because I'm like, the minute we click Publish, it's like. And it's old.

A (26:54)

Yeah, yeah.

C (26:57)

Jonathan, I was just going to comment too on what Nate was saying, like around or just generally like how do we enforce these things? I think the same way that there are tools now to help teachers detect when someone has used ChatGPT to write an essay, we will see more AI driven tools to detect when AI was involved in content production. And it's a necessity, it's going to be a necessity for the social platforms, it's going to be a necessity for this industry because ultimately the publisher is the most at risk from the FTC's perspective and they've got to protect themselves. So they're going to have to put these safeguards in place. And with the proliferation of all these different renditions and who knows, potentially gen AI creative on the fly, some point off in the future you're going to need to have some form of governance and censorship in between what's coming out of the demand side and what's arriving the consumer's eyes.

A (27:50)

Yeah, and a big thing too is also like we, you know, I know we chatted about this and previously is like talking to your, your clients, you know, you need to be aligned, you need to know how the clients feel about the use of AI, where it goes. We know some client, you know, when it comes to their products and everything. So it definitely needs to be a lot more out in the open and, and not kind of, hey, we're, we're sliding this in as AI. I think it's. And it's going to become more normalized. Just like. Was the same conversation happening when you mentioned. I think, Jonathan, I think like Photoshop, when Photoshop first came was this conversation. Do we have to let people know that this image was slightly Photoshopped a couple of years from now? I think this will be that same kind of world of. It's another tool that helps enhances the work.

B (28:40)

We're running out of time, but just before we wrap up, I don't know how closely you guys read the contracts we sent over. But you have all agreed contractually to come back here in exactly five years on November 14, 2030 to do this panel over again. When we meet back here in five years. What will the biggest impact of AI on CTV advertising have been?

C (29:04)

I'll put two bets. I think one is the continuation of what we're seeing so far, meaning I've yet to see native video advertising really take off. I think Gen AI potentially can make that more possible. We've talked about like alignment on content, so you could potentially exit out of a content stream, go into seamlessly go into a marketing promotion that feels still like look and feel to the stream that you're just consuming. Like if I'm watching, you know, just coming out of the. Or maybe we'll say going into the holiday season, maybe there's holiday themed creative right after that. That, that I could see being an interesting application of Gen AI in native and CTV space or dynamic brand replacement on objects in the room or something. The other bet I'll make is I think if we do see a proliferation of AI generated content, there's going to be some of those marketers out there. They're going to be like, hey, we shot this with real people. This is on set, real people, scripted, we edited it with scissors. They're going to really lean into the retrospect.

D (30:11)

I think that makes a lot of sense. I also have two. Mine is that AI will be invisible. I think we won't be having this conversation on did the hand get properly orchestrated? It'll just be a part of production. Nobody will talk about it. This is bullish. But I generally am a dreamer. I feel like CTV will hopefully finally catch up to digital and being as personal personalized as digital content is and as shoppable and Nate, I'm sure you have a feeling on that, but I have been talking about Shoppable video since like 2007 and I'm just dreaming big that AI can make this fully happen.

A (30:57)

Yeah. And first off, Todd, AI will be sitting here. I think we'll be at that point. I'll. I'll be. I'll be probably off in the other room. But I agree with both. I, I think it's going to be. I think it's going to be a middle. I think the, the big dreams we have now, it's going to fall somewhere a little more in the, in a. In a reality where it's going to be more personalized. I think there's definitely going to be a lot of like the goal of shop right then and there. And it's going to be very easy and it's going to be very seamless. And, you know, I think we'll be talking, we'll be looking back at this like, oh, can you remember what we thought I was going to be? And it's just be normal.

B (31:35)

Yeah. All right. All great insights. I don't know if you guys were here at the start of the summit today, but our co host, Jeremy, promised that this summit would be so good that people one day would tell their grandkids about it. I think we've accomplished that today. Caroline and Todd and Jonathan, thank you guys so much for your time and for this great conversation. We appreciate you being here.