Loading summary
Transcript91 lines
- [00:02]
A
Ready to shape the future of digital business? Then don't miss Demexico 2025, Europe's leading event for digital marketing and tech. On September 17th and 18th in Cologne, digital pioneers, innovators and decision makers come together to share ideas, network and shape the future of digital business. Expect top speakers like Carson Meschmeyer, Vanessa Schutzel, and Sir Martin Sorrell across 10 summits. Get inspired. Be bold. Move forward. More@DMEXCO.com hello. Hello, and welcome to another episode of the Digiday Podcast, a show about the business of media and marketing. I'm Kamiko McCoy, senior marketing reporter here at Digiday.
- [01:00]
B
And I'm Tim Peterson, executive editor of video and audio at Digiday Media. What's up, Kimiko?
- [01:05]
A
Tim, did you see the Gap commercial over the weekend?
- [01:09]
B
All I saw was the Gap commercial over the weekend. First I saw someone, like, dancing to milkshake, and I thought, that's weird and seems to be a very specific choreography, but whatever. Swipe. Another person dancing to milkshake. Same choreography. Okay, so here's the latest dance trend on TikTok.
- [01:27]
C
Cool.
- [01:27]
B
Swipe. Then I see the ad and I think, okay, so I don't know if this is, like, what spawned the dance trend. I don't know if this is in response to the dance trend, but whatever. Looks like a. A Gap ad. I don't even clock that. Like, it's a Gap ad for jeans. Swipe. More milkshake, more people dancing, more clips of the Gap ad. At some point, I start to see tiktoks of the Gap ad with people even either narrating or, like, putting text on screen, juxtaposing it with the American Eagle Sydney Sweeney ad.
- [02:08]
A
Naturally, naturally.
- [02:10]
B
This is when I learned, like, oh, the people starring in this Gap ad are not just dancers who were hired for it, but this global pop group, Cat's Eye, that's, you know, known for one, having a bunch of very diverse members, as opposed to American Eagle, which famously had Sydney Sweeney talk about how great her genes are as white woman with blonde hair and blue eyes, eventually I make it to a TikTok that is now. Now we're getting into the realm of social criticism. This TikTok goes, okay, not only is this ad featuring people of diversity as starring in it, as opposed to the American Eagle ad, but did you notice the background dancers? Kimiko, did you notice the background dancers?
- [02:57]
A
Absolutely not. I was stuck on the fact that we brought milkshake back after almost two decades.
- [03:03]
B
So this person, and I wish I remembered you know who they were because I'd love to credit them for it, but it's TikTok I just swiped through. But they go. Did you notice that all the background dancers are either wearing black clothing or brown clothing? As like this is more of a dog whistle or not a dog whistle but like more of a signal of like Gap knows exactly what they they're doing. They are subtweeting American Eagle. I don't know if any of this is true, but I am here for all of it. I am here for social criticism applied to advertising. I am here for drama between Gene Brands when it comes to advertising. I'm just given how this summer, how this year has been. Give me this. This is my junk food.
- [03:50]
A
Yeah, yeah. It's either this or Love Island. And since that's done now we need something else to keep the beast going. You know what I mean?
- [03:56]
B
Yeah. So that was a good sign.
- [03:59]
A
Good shout. Later on in this episode, we've got Seb Joseph, our executive editor of news, and Michael Berge, our senior media buying editor, to talk about how AI if you thought we were escaping AI, you're wrong. So sorry. About how AI is affecting the agency's businesses. Every time I think about, we have a conversation about AI, I think that's seen from holes where it's like, I'm tired of this cray paw. And he's like, that's too damn bad. That's us talking about AI to our audience. Anyways, first we're going to talk about the new AI drama. How the retail media boom may be headed for a burst of its bubble and then the new compensation model for publishers. So lots of juicy scoops.
- [04:40]
B
I mean, let's just keep with the drama because there's plenty of drama this episode. Like you mentioned, also in the featured segment, we have the drama of just our brands going to be taking AI in house. What's that going to mean for agencies? How does compensation work? Just a lot of AI drama. But as you mentioned, starting off with X, AI, the Elon Musk's company that is the now parent company of X, the platform formerly known as Twitter. I can't believe like how many different. It's like a, you know, British dukes, like titles, all the different titles. X AI, owned by Elon Musk, owner of X, has sued Apple and OpenAI, alleging that they have been conspiring to thwart competition for artificial intelligence. According to Reuters coverage of this. Reuters cited, you know, from the lawsuit quote, this is Xai's allegation quote. If not for Its exclusive deal with OpenAI, Apple would have no reason to refrain from more prominently featuring the X app and the Grok app in its App Store. I don't know that that's true.
- [05:51]
A
I know I've said this before, like if you can't beat them, sue them. But that really does seem to be this man's philosophy here when it comes to this type of deal. Because like you had the not to take it back to give everybody PTSD about Garmin, things like that. But this is a, a regular behavior from Musk where if my business is not winning the way that I want it to win, boom. Lawsuit.
- [06:12]
B
I know democracy is under siege. I don't want to act like the App Store rankings are a reflection of hope for democracy. That said, like the App Store rankings, as far as I know, are just a reflection of download. If you are the most downloaded app, you are at the top of the App Store. So I mean, have you downloaded X lately? Have you downloaded Grok at all?
- [06:38]
A
No, I have not.
- [06:40]
B
Same. I haven't downloaded Grok. I downloaded X when it was Twitter when I was a baby in what, like 2010, 2009, after Sully landed in the Hudson. That was when I got on Twitter. So it's been a minute. I'm not contributing downloads for x. Whereas ChatGPT is a newer app and a very popular app and its user base is many times the size of X's. So it makes sense that ChatGPT would be at the top of the App Store.
- [07:12]
A
And also, didn't ChatGPT just get dethroned by Hank Green's Focus Friend app?
- [07:18]
B
Yep, yeah, it did, which I haven't installed. Have you heard about this app?
- [07:23]
A
I have seen a couple of TikToks from Hank Green himself talking about it. I couldn't tell you what it is.
- [07:31]
B
So the idea is. So it's an app that's supposed to get you off your phone for a period of time. Basically. I don't know if like you set a timer or I think you like set how long you want to not be using your phone for. And through the app there's a character that is a bean who's I guess knitting like socks and other things for its room. And the idea is if you stop, if you turn off focus sooner, if you start using your phone before like the timer goes off, the bean looks at you, is really disappointed and has to stop knitting its socks. And so the idea is people can't get off their phones for their own well being, but no one wants to disappoint this cartoon bean. And so this actually is helping people to get off their phone. I think it's like with the Gap cat's eye stuff, it's fun. It's a little dumb, but it's fun. It seems harmless, but yeah, it also dethroned ChatGPT. So it's not like ChatGPT has this reserved position by Apple atop the App Store rankings.
- [08:41]
A
Yeah, I do think it's. I don't know. And I also am curious, kind of like when you talk about, like, AI platforms and where they, you know, perform and whatnot, I don't often hear GROK listed as like, you know, the top. It's kind of second tier along with like, perplexity, you know, and that and that type of deal.
- [09:00]
B
Yeah, it's done well in certain, like, benchmark tests, but there's a lot of questions surrounding those tests. Like, Met has also experienced this with llama of whether some of these large language models are being trained or designed specifically to do well on the test. And so they score highly on the test, but then in like, real world situations are garbage. It's like the kid who is really good in high school on tests, but like, you come to meet them as an adult and they have no idea what they're doing with their life because they were always just really good at tests. But life, not so much critical thinking, not so much.
- [09:39]
A
So is it a good, solid product or do you have a true antitrust case on your hands? I think that's the question.
- [09:47]
B
Yeah. But, yeah, so the timing of this lawsuit seems a little tough because of one, the focus friend dethroning ChatGPT in the App Store.
- [09:57]
D
Two.
- [09:57]
B
Bloomberg reported. I believe it was over the weekend that Apple's talking with Google about having Gemini power Siri. So what? A year ago, a little over a year ago, Apple announced a deal with OpenAI to have ChatGPT be part of like, Apple Intelligence, which still hasn't really fully rolled out and has been very disappointing as someone who was very interested in using it and seeing how that could help my workflows. But as someone who's also used Siri and specifically just, oh, I just, I just set off Siri. I was about to send a text message. That's all I ever use that for. Like, it doesn't have any real value to me. Maybe Gemini powering it will help with that. But it is interesting that Apple's talking with Google about Gemini powering it, because that would seem to take a bit of the wind out of the Sails that like Apple and OpenAI have this relationship that's anti competitive.
- [10:58]
A
Yeah, I think it does take the wind out of the sails. You know, you had a series of tweets threatening a lawsuit that turned into an actual lawsuit being filed. But I don't know how strong of a case you have in talking about anti competitiveness and antitrust and whatnot. If you know one, you had a bean, a fake bean, dethroning chatgpt, you know, and Apple and talks with, with Google.
- [11:22]
B
So yeah, unless like all Musk is really getting looking to get out of this is a deal with Apple to make Grok one of the options for Apple intelligence or maybe like the option in place of ChatGPT. I think if, if Musk was still part of Doge and had that Trump relationship, this would feel more substantial than it is. At the same time, like, given the way things are coming this year, who knows? Another thing that, you know, we'll have to see how things go is retail media. So you've been all over this story, but like, there's just been a boom over the past, what, five, six years? Longer. I mean, you just spoke with Christie, who launched Target's retail media network and is now at Uber advertising for last week's episode. But we have. Macy's has reportedly struck a deal with Amazon, according to adweek, to sell its ad inventory through Amazon. And then what's the other one?
- [12:27]
A
Yeah, Ace Hardware. They're, they're launching a retail media network with Epsilon. You've also got Walmart and Target who have seen like, as they break out their ad earnings, you know, their, their revenue has grown for their ad businesses. All of this to say, I think where I know we've been talking about an inflection point within the retail media landscape for a while now, but I think you're starting to see like the actual footprints of it show up where you've got a conversation where like long tail players are having to either tap into different tech stacks or tap into being a late mover in this space to compete with these bigger ones that are actually making money. But also, here's the question. If these major retailers. I'll start with the Walmart and Target first. If these major retailers are like relying on their ad businesses, right, because you've got softening sales, economics, tariffs, and then, you know, Target by themselves with like boycotts and things like that, the question then becomes if they're softening sales in other parts of the business. Retail media ad networks are your, like, your bright spot. And the whole Thing of this is like endemic advertisers are advertising on your site to sell their stuff. If there is softening sales, how do I then justify that sales are happening from this free telemedia network? And I think that's the question that media buyers and advertisers have been asking for a long time where this is maybe going to be the first time that you see an actual confrontation with that.
- [13:55]
B
Because like in the Target example, so you know, Target's retail media revenue hit 217 million, which is up from 162 million in the second quarter of this year. But at the same time like overall product sales for Target have been flat. So like retail media is growing. The regular retail business, not so much Walmart, it's connect business, which is its retail media Network in the US grew revenue by 31% year over year in the second quarter of 2025, but overall revenue was only up by 5%. And so the idea being product sales not doing so hot, retail media sales doing hot, but if the retail media sales are to drive product sales and the product sales aren't happening in nine to 12 months, are we going to see a drop in the retail media sales? Is that what you're saying?
- [14:48]
A
That's the whole thing, right? Because for a while now, like I said, media buyers part of it was like we, if you signed a jbp, a joint business planning agreement last year, now you're seeing the effects of that, you know, or three months ago, whatever, however long ago it was, this is the tail end of those negotiations. Now I've talked to media buyers and advertisers and the thought pattern is like, we're getting to the point where there's enough data for me as a media buyer to say, yes, this is working. No, this is not working. This is the year over year, month over month, quarterly return on ad spend. And now that it's those things, those building blocks have been in place for a while, you know, there, there is going to start to be more questions about what do I get for, for this. And you can't bully me in a JBP anymore because now I have numbers on my side. So I think you will see that inflection point start to, you know, happen alongside the maturity curve.
- [15:40]
B
Yeah, because I guess another big question is this is all contingent on people going to retailers, websites to look at products in which to then see these ads. But with, let's go back to AI with AI, if people aren't going to websites, that negates a lot of this which is forcing new business models, including new business models within AI, where like business models have already arisen. So our third juicy scoop this week is about publishers looking to change how they're being compensated by the AI companies for their content. So we've had a whole host of publishers sign content licensing deals with OpenAI, with Amazon and in those it's basically, hey, OpenAI writes a check to the publisher, the publisher says, cool, here's my content library. Like go in, grab what you need and we're done. And you know, maybe we'll do this negotiation again in a few years. Thing with that is the value isn't really there for the publishers long term. Yeah, like the money, the payments that they're getting in these kind of lump sum type deals aren't going to cover their businesses. It's kind of, they're in a way, there's like a milk and cow analogy that I'm not going to land the plane on because I'm already mixing analogies. But it's basically like this is not the way forward. This is not going to be bringing in money on a recurring basis. The incentives aren't going to be there for the AI companies to continue to pay for this stuff. And so you have IAB Tech Lab for one who has this working group, the AI Content Monetization Protocols Working group that's developing a framework to enable usage based pricing. This is also something that just Davies are senior Media editor reported on recently how there's a shift towards new types of deal structures between publishers and AI companies that are more usage based. Perplexity is already doing this effectively through its ad revenue sharing compensation model where it has a deal with Gannett that's usage based. If people are using Perplexity and links to USA Today pop up in that. Or like there's content from USA Today that's cited for that. USA Today then gets a cut of the ads that would show in that query as kind of a reward for its content being valuable there. That seems where everyone wants to be moving to. Including Perplexity, which formally is moving towards this compensation model with what it introduced today on Monday called Comet plus, which is a $5 a month subscription program that basically people are going to. It sounds like a newsreader kind of thing, like a Flipboard or an Apple News, but you pay to get Comet Plus. Comet plus is going to be able to like tell you like here's what Digiday just reported on or if you're just like hey, what's up with this like Gap Cat's eye thing and it pulls an article from publishers who explain it. There'll be those publishers get a cut of the $5 a month that you pay. And so Perplexity actually said it's going to share 80% of that revenue to publishers. So like each month you pay $5, $4 is going to go to publishers. Now that's going to be like divvied up based on all the publishers that you may come across when using perplexity, it's still 80% of the money that gets shared as opposed to usually it's like 50, 50 or 55, 45. So it's, it is a new compensation model something that's more favorable to publishers than how this could otherwise go and still could go.
- [19:38]
A
Yeah. And I think that's interesting because publishers have notoriously not had the upper hand in these deals.
- [19:44]
B
Still don't. Still don't.
- [19:46]
A
I feel like this is you know, an inch forward because at least, at least you have more of a say so when it coming to the table. I think there's still you know, obviously depending on like what size publisher you are and like what you're like what is considered usage and things like this all have to be shaken out. But it does bring up one, at least the conversation and what this compensation model could look like. And on the flip side of this, publishers aren't the only one trying to figure out what this looks like. In this space on today's feature segment, we've also got a conversation about how agencies are trying to figure out what the compensation model looks like here for.
- [20:20]
B
Themselves and how usage based pricing could also make sense for agencies. So yeah, we'll talk more about the usage based pricing models and kind of just the state of publishers dealings with AI companies in an episode that we'll have in I believe two weeks when we're going to have Jess and Sarah Guaglioni on the show to talk about that. So we'll be able to dive more into it then. But for now let's dive into agencies and AI and how compensation models are being changed in that respect with this conversation with Seb and Michael. Seb Bergi, welcome back to the podcast.
- [21:05]
D
Thank you. It's good to be here.
- [21:07]
C
Good to be here.
- [21:08]
B
So not to be a impolite host, but I actually want to start with Kimiko because you just published a story with the in some ways comforting headline but with a not so comforting end to the headline. The headline on this story is why generative AI doesn't fit into the Standard in housing playbook EM Dash yet. And that is a human ridden EM dash, not an AI ridden EM dash. But it's very ominous EM Dash. So it feels like today we're talking about how AI is impacting agencies. Businesses from forcing them to change their compensation models to actually stealing jobs from agencies. If brands were to decide to just hand over that work to AI agents, your piece looks at the latter. To what extent are brands handing over work to AI agents instead of to agencies?
- [22:01]
A
You know, interesting. And one I want to set the record straight on EM dashes. Writers have been using those for a real long time and has nothing to do with AI which is scraped our writing from the Internet anyways. Besides that, myself included, now that that gripe is public, I've tried to attack this story from several different angles to try to get a sense of like, if you think of it like a conveyor belt, right, or the assembly Ford assembly line, how far are brands willing to take AI? I've approached it from geography, cost, things like this, and a lot of it seems like how much of a heavy lift is it going to be when it comes to data process privacy and like having to hire the people, the knowledge workers that are going to know how to run this stuff because it hasn't made it to the point where it can run itself. So I talked to a couple of different brands and they kind of fell all over the spectrum. You had Adobe, who has a lot of these products that they've built out themselves to be able to sell. Right. To others to kind of justify the cost. Because if it, you know, if I spend a ton of money on this and it doesn't work out, doesn't matter because I can still sell it to other people. On the other hand, you've got US bank. Their CMO said, yeah, if I build this out, that means that I have to maintain it. And maintaining it costs money. So it just a lot of it just kind of hinges on like, should I pay my agency to take on these challenges and put the cost there or do I do it myself? So it comes down to data privacy and how much cash I'm willing to spend.
- [23:28]
B
Yeah. And Seb, reading Kimiko's story, it feels like there's a lot of parallels to the programmatic in housing trend from like the late 2010s where some brands were taking programmatic in house and even looking to take social and search in house, but that didn't kick off a full blown wave of it. So like to what extent are there parallels now with the potential for brands to take AI in house. And to what extent is this a whole other beast?
- [23:57]
C
Yeah, there are definite parallels that are somewhat triggering, given I covered extensively that kind of in housing phase.
- [24:07]
B
I feel like we trigger you each time you come on the show. Last time it was privacy sandbox. Now it's programmatic in houses.
- [24:13]
C
I don't know what that says about me, to be fair. But anyway, we digress. Yeah, I think as with all of these transformational kind of innovations. Right. We kind of go through or the like, advertisers go through that kind of Gartner hype cycle that I think the in housing trend is a really good sort of distillation kind of. Right. And you get that kind of trigger point where there's a kind of. It comes out and there's a raft of curiosity that kind of naturally leads to marketers thinking how much of this can we kind of control, which will start with maybe hiring an expert that will kind of grow into a team. And I think that's where you'd get, what is it, the kind of peak of inflated kind of expectations. And I think you saw that with kind of programmatic when there was that kind of raft to hire, you know, directors and build out kind of in house teams. Very quickly you kind of go through that trough of disillusionment which I think we're kind of, you know, at the. In the kind of foothills of right now where marketers, CMOs, CEO, CFO start to realize that the technology may not deliver on its initial kind of promises. And I think we're kind of there with the AI kind of bit now where you're starting to see brands that have jumped on this thing early and are testing it in kind of various sort of guises to kind of Kimiko's point, almost like isolated sort of incidences, a local market here, a campaign there. But then they're starting to realize that there's a lot of work that needs to go into it in order to be able to kind of scale it, not just in terms of kind of expertise, but investment and then I guess kind of organizational sort of structure as well, which is where I think we'll get to that kind of, you know, was it the slope of enlightenment and then the kind of plateau of productivity? All that is to say I'm kind of bullish on the fact that I don't think this will be a, you know, a kind of death knell to kind of agencies in any shape or form. I think it will definitely transform the kind of Relationship as we saw it do with programmatic kind of back in the day. But I think the kind of outcome of all of this will be that marketers will hopefully have a clearer idea as to kind of how they want to exert control over this technology and all of the different guises. It's going to kind of manifest itself in, you know, moving forward.
- [27:04]
B
Seb, I feel like I just sat through your TED talk slope of enlightenment plateau.
- [27:09]
C
It's the garter hype cycle in it. I did a bit of prep beforehand, but yeah, and also it came up in conversation when we were reporting on this a couple of weeks ago. So I took notes.
- [27:23]
B
Tim, there you go. That's all we have.
- [27:26]
D
And you remembered them too. You remembered them too. It's even better.
- [27:29]
C
I've got them written down, to be fair. I'll share that one.
- [27:34]
B
Birgit, you talk to a lot of agencies. To what extent are they worried about brands taking AI in house or are they seeing Seb laid out and Kamiko laid out in her story? Actually this isn't really a threat to our business, so we don't gotta worry about that. We got enough problems.
- [27:50]
D
Well, I think because they've, they've already had to worry about brands in housing on a couple of other levels, there's less of that kind of jittery instant response of oh my God, we've got to do something to stop this. Agencies are so busy trying to figure out how they can incorporate AI into what they're doing on behalf of their clients that that's kind of keeping them pretty occupied. I don't think there's much discussion about losing AI driven work or AI oriented work to brands more how to make it work for them and then get into the discussion of how to get compensated or in the negative side, how not to have clients say, well, if you save this much time and have to use so fewer people, you know, where's our money coming back to us from? And I think that's kind of where agencies are with, with brands right now. It's, it's early times for that discussion, but definitely it's, it's going to accelerate agencies moving more towards outcome based remuneration structures. Just like we saw kind of COVID you know, rapidly encourage the use of kind of like commerce and E commerce. This is having that effect too.
- [29:13]
A
I think it's interesting what you're saying here because again, back to the conveyor belt point and kind of like how, how they determine how they charge for these things and. Right. I'm Almost curious about like how agencies then make the case for themselves because part of my reporting kind of revealed that agencies essentially the, the tools that are available to brands themselves right now. So like a Google or a pmax meta AI and things like that. We've reported Crystal Scanlan, our platform's reporter has kind of dug at this already that you've got folks walking away from this because the tools themselves, they're not user friendly for brands. Agencies are kind of able to come in and layer on some of their own tech stacks, their own tools, their own data. Right. To be able to make the case for themselves in this AI landscape would love to kind of like are you seeing from my end that's what I'm seeing. Are you seeing the same thing from your pov?
- [30:04]
D
Well, I would qualify it as. It's a different story when you're talking about holding companies versus larger independents holding companies. Because AI can have such a kind of broad impact across the board and maybe find kind of large scale efficiencies and savings certainly within like the analytics and research area, which is where a lot of it is being applied, although some of that is actually machine learning that's been around for years. But there's other newer forms of generative AI happening. Independence might be a little bit more vulnerable to that because they don't have quite as much kind of countercloud with the Googles and the metas as a holding company does. So that's how I would distinguish it.
- [30:49]
A
Also to the point about like external agencies and how much they're able to do versus brands. I mean like you've also got hold codes to your point, like I said with even like Adobe, which is a brand, but the point still remains the same like an external agency can build out this proprietary tool and then turn around and sell it to again make the case for themselves as a, as a product to bring in revenue as opposed to just a. There's a word for it, cost center. That's what I'm looking for.
- [31:18]
D
Well, certainly the holding companies have invested so much in kind of incorporating AI into their tech stacks and trying to become. Seb, you wrote about this a few months ago. The idea of holding companies, no, you know, being more kind of like trying to be platforms or operating services than what they used to be kind of perceived as. And AI is having an effect on that both on the creative side where there's so many kind of efficiencies being created on that creative side. But as I just said before, also within the media side on research Analytics. It's coming for planning. I don't think it's quite there yet, but it's inexorable that this is going to continue to happen. But I will say one thing, going back to a point made earlier. The compression of time frame from shiny new object to disillusionment to those Great Gartner terms that you used before Seb, like it is so compressed with each new technology. With the advent of the Internet, it took about 15 years for that to really make itself, you know, turn. Turn the media business into digital media. You know, the advent of Programmatic was a little bit more compressed and now this timeframe with AI is even shorter. Like whatever the next innovation is, is it going to take like six months to go from shiny new object to disillusionment to, you know, essential part of just a quiver of options?
- [32:54]
B
It does feel like this has happened. It really sped up when it comes to the agency holding companies because like agentic AI was a term that really just hit the lexicon, I feel like second half of last year. And when I was at ces, the start of this year, I was talking to a lot of agencies and agentic was the big buzzword and they were all trying to figure out those strategies. Within a month of that, one of the agency holding companies demoed for me, their AI platform and how they had all the large language models in there, they had generative video tools in there. They could manage briefs and have that shared across their entire team. There was another agency holding company I talked to a month or two ago. This person was saying they have nearly 30,000 AI agents, which is just. I mean, you could spin up an AI agent pretty quickly these days, but that's still an insane number. Bergi, what are the use cases right now for AI within agencies? To what extent are they using this for client work versus just internal things? When you mentioned operating systems, that triggered me to remember Marcel, which is publicist's supposed operating system from years ago.
- [34:07]
D
That was supposed to be like their first kind of introduction of AI. And now it seems so quaint and antiquated. What I would say is on the internal side, Agentic is being used predominantly to create kind of virtual focus groups. They're creating Personas so that when a media agency is trying to figure out some kind of new direction for a brand, they take all the input, all this data from the client, along with other third party data, whatever internal stuff they've put together, kind of put it in this cocktail shaker. And then AI is like the bidders you add, I guess, shake it up and it comes up with these Personas that can represent multi, thousands of people of consumer focus groups that they would otherwise have to spend thousands upon thousands of dollars and maybe even hundreds of hours parsing through to kind of glean these insights that Personas give them in a matter of minutes or hours. Now, what I don't know, and I think what is yet to be proven, is how accurate these Personas are. But agencies are certainly saying this has been a real game changer in terms of being able to pivot on ad campaigns or even broader directions of clients. That's the predominant way it's being used, certainly more at the holding companies. But as AI gets less expensive, even independents and smaller shops are making these kinds of investments to do that. And what I don't know yet, and I think Kamika, would be maybe this story we could work on together is like how brands feel about agencies doing this kind of shortcutting of focus groups and consumer research. And if they believe the hype. Sorry, I didn't mean to turn this into a news meeting.
- [35:54]
C
No, but it's a really good point. And it came up in conversation I had with someone on the brand side recently, and they kind of wondered out loud whether the explosion of agentic AI that we're seeing across the Holocaust exacerbate the remuneration problem kind of, rather than kind of remedies it. Because what they were saying is that, you know, agencies already tend to measure value by like, outputs, you know, instead of what they call it, like, impact. And so he was getting at this idea that the danger is that, you know, if these Holcos have essentially kind of infinite sort of capacity and haven't necessarily put the guardrails in place, that kind of makes the problem worse. Right? Because instead of focusing on what really works, it just risks turning more into like a standing for better. So you end up with like, you know, teams that are busier, but, you know, they're kind of leaner and subsequently more brittle and just further removed by these kind of tools from doing the work that actually moves the needle.
- [37:05]
D
I wouldn't go so far as to use the word slop, but it gets closer to this idea of AI slop if agencies are just cranking out kind of tonnage as opposed to stuff that, you know, might resonate. And I think measurement is going to be a key there. And I think we're still kind of seeing that, catching up to what the potential is.
- [37:26]
A
Well, that's the thing, right? Because they have to Sell something. In this landscape of the AI hype cycle, right. AI effectively, kind of, well, stands to really threaten the way that agency compensation models work. If you're charging based on how long something takes and you cut it down to 20 minute, you know, what, what then happens as to how I'm charged? So if I can't charge you by the minute because AI has made things faster, right, not necessarily better, but faster, I've got to find another way to supplement that revenue. And to me, what this kind of signals, based on what we're saying here is that now I'm building digital twinning tools, synthetic audiences and things like that, because that I can sell.
- [38:06]
C
But it's that measurement point that Bergi sort of alluded to, which I think is really important as well here, right. Like how can you implement an outcomes based model as a CMO if you don't know essentially what outcomes you want? Right. You know, whether it's kind of ctv, real media, even credits, there's still a ton of to kind of unpack there in terms of, you know, what they're, what they think they're buying versus what they're actually kind of buying and frankly also kind of how much that they're kind of paying for. So it'll be really interesting to see kind of how the, all of this stuff around kind of AI and the, the way it's changing the kind of, the kind of commercial sort of aspect impacts the measurement side of things as well. Because that is a perennial sort of problem that kind of moves in fits and starts.
- [38:58]
D
I would argue that there's both a fine line as well as a yawning chasm between output and outcomes.
- [39:05]
C
For sure.
- [39:05]
D
I think that's kind of what you're saying.
- [39:08]
B
Yeah, I mean that's always the challenge with outcome based compensation or outcome based buying is how do you know it was the agency's work or even this ad that led to a boost in sales or actually correlated with a decline in sales and not, you know, it's just a bad product or supply chain issues, what have you. And so like that seems to be one potential compensation model, but obviously there's a lot of, you know, problems with. Seems like another is kind of usage based compensation or in some ways having AI computer be on the line item at agencies. Like what we're seeing right now with the AI platforms themselves is they've and a lot of AI tools like Cursor, which is a vibe coding tool, they're moving to usage base. Because the big problem with AI is you need a lot of GPUs to use this stuff. There's so many costs. I mean, anthropic OpenAI, they're not making money, they're losing so much money because of how much it takes to use these tools. I imagine agencies would stand to also be losing money because of how astronomical these costs can be. How are they thinking about making up for that? And is that where things kind of land with the compensation models?
- [40:29]
D
I think you just gave me my next story idea because I don't know if I know the answers to that yet.
- [40:34]
C
Well, I don't think anyone does. Right. Again, on the Omnicom earnings call, John Wren sort of essentially said that that would be a sticking point because it does cost a lot of money to kind of support these solutions in the background, like just a lot of overheads. At some point, some of that is going to have to be passed on to kind of clients. And John was basically saying that that might be the sticking point. Once clients see how much these services cost, they may actually think, oh, hang on a minute, we may need to kind of take a bit, take a breath and kind of figure out whether or not this is kind of actually viable at the moment. So it's a really kind of great point, Tim, and something we should kind of definitely look at more as a news team because it sounds like those are conversations in play right now.
- [41:30]
B
And that's actually really interesting because that makes me think one of the reasons why agencies have been able to survive the media buying in housing boom, as it were, is agencies buy in bulk. And so they have a lot of negotiating power from doing that, as opposed to an advertiser, A P and G, a Coca Cola conceivably could go direct because they spend enough money, but from the mid to long tail, they don't have enough money to really negotiate with a Disney, an nbcu, a Google. Similarly, if an Omnicom, especially now with IPG and the full WPP on down the list, if they're using so much AI and so much compute, they would be seemingly in a position to be able to negotiate with Google with OpenAI with Anthropic to say we're buying a lot of compute from you, we need a discount on this that we can effectively pass on to clients if the clients then want to borrow that compute. I feel like now we're just seb, you put us in a warden school of business conversation. It's a really interesting thought exercise to go through though.
- [42:46]
C
I agree. I think at the heart of it, Tim, I Think this is what we talk about, you know, when we sort of write about this shift from the kind of Holco era to the kind of operating company kind of era and the subsequent like kind of economics around that. Right. Because you know, the whole care was much more about, you know, financial engineering essentially. Right. Like you know, inorganic growth, cost cutting used to sort of prop up EBITDA and so on. And I think, you know, when you look at what Lesser is trying to do with WPP Media, you can definitely start to see how they're thinking about, you know, kind of riding on the coattails of the kind of platform businesses as they, you know, kind of spend ridiculous eye bulging amounts of money on AI. And you look at, you know, who, you know, WPD Media's most recent partnerships are with like all the biggest players in AI and kind of data and I think it supports your kind of point, you know, about how we might start to see some of the commercial relationships with those companies kind of play out. There is probably something here when it comes to how what this means specifically for the AI companies, the perplexities and the OpenAI's of the world, who I know are, you know, working with these companies now. But it's, it's not, it seems like it's more of a consulting type of kind of model than a, you know, your typical kind of media sort of arrangement. So definitely lots to unpack there.
- [44:27]
D
When it comes to the Google and the meta side of AI, those kinds of negotiations are going to be so much more complicated because you know, especially with Google, agencies are so kind of connected into Google through so many ways for them to try to kind of counter negotiate some sort of discount. You got to take into account the use of pmax and all these other services that Google has. And I think that makes it a much more complex 3D chess kind of negotiation process than it would be with let's say an OpenAI or some of the other kind of pure play AI providers.
- [45:06]
B
Yeah, it feels like the agencies would have more of a position of strength with a perplexity with an anthropic like Anthropic is famously losing money and to use its API costs so much more on a token basis, on a per token basis compared to OpenAI. Like OpenAI just released GPT5 and OpenAI if I'm not mistaken, it's undercutting everyone on price. Whereas Anthropic. But anthropic opus 4.1 is so much better than all the other a large language models when it comes to coding. Right now that there would be an opportunity for an agency to be like, hey, anthropic, I see you could use some money. We got money. How do we work together? And that can kind of preserve some competition with a Google, a meta, an Amazon.
- [46:01]
D
Yeah, good point. To play them off each other might be the way to do for an agency or a holding company rather to maybe just save a little bit there. Look, they're saving a lot on personal being let go. One, one analyst told me, you know, that CMOs at the holding companies are privately acknowledging that they're using natural attrition to just not replace people. And you know, kind of AI can kind of fill those gaps and cracks where people leave. So when John ren talks about $750 million in savings, yeah, a lot of that's going to get eaten up on AI related costs. But they are saving a lot on just shaving down the number of personnel.
- [46:49]
B
And I imagine that also that affects people who have been in the business, been in the workforce, but also entry level employees or the potential for entry level employees.
- [46:58]
D
I think we're seeing that across the board. In the economy, where I've read stories in the last two, three weeks of entry level jobs are just harder and harder to come by because that seems to be where a lot of AI can kind of be applied. But yeah, what do you do when you're trying to get into an industry and you can't get in at the entry level? How do you become seasoned enough to be able to get that mid level job? It seems a little bit like a catch 22.
- [47:28]
B
Or do we see people just decide, actually, I'm going to start my own agency because all I really need is a Cursor subscription, an OpenAI developer account, little bit of budget, but I can make this work on my own.
- [47:44]
D
Maybe you can get that local bakery to be your first client, who knows?
- [47:47]
B
It's funny because coming into this conversation I was kind of expecting that's where I would land because that's where my head's been at lately. Kimiko and I have been talking about this on the show for the past few weeks and I've kind of been of the mind of like, okay, sucks that people are losing jobs, but because of these tools that are out there, there are ways to kind of create your own job now thanks to these tools that can automate a lot of the work, handle a lot of the heavy lifting that you would otherwise need to hire people for. It's that whole idea of there are people who are paying for a ChatGPT Pro subscription at $200 a month because that's cheaper than actually hiring a person and they can use the functionality for that. But as we were talking about compensation models and the cost of compute, it feels like. And I think both can be true, but it feels like there is a real case to be made for the sustainability of the agency holding company model because they can assume these large costs.
- [48:51]
D
Yeah. Look much like kind of for 30 years I've heard the upfront is dead and is never going to happen again. All the death knells of agency holding companies. I just, I don't. There may be fewer in two years. I mean, certainly we got Omnicom and IPG becoming one, but I think agencies are going to always remain kind of essential to this process because the process is complicated, the market is complicated. I don't mean to sound like I've got too much Stockholm syndrome because I cover agencies all the time, but you know, I just. It feels like they always find a way to still be essential in some way or another. So. And I hope all those entry level people don't just become influencers. We have enough of those.
- [49:37]
A
Oh, I was just about to say we got to get some more plumbers and things in the mix. But I do think, I do think there's something to be said about kind of the changing model here. And I think the basis of this conversation is that there's going to be a lot of ripple effects that we're going to see kind of shape up in the same way I can almost. I would be so interested for somebody to make a comparison to the dawn of the Internet, like you mentioned earlier, Bergey, and like those parallels, not me writing that, but somebody else writing those parallels between what the Internet meant when it was finally adopted and what those job shifts meant and kind of what that looks like for, for AI as well.
- [50:15]
D
Hello, Harvard Business Review.
- [50:17]
A
Yeah, exactly.
- [50:19]
B
Yeah, I guess to borrow from Kimiko's headline, the M Dash. Yet on this whole conversation is something Bergi, you had mentioned earlier, this idea of AI based focus groups and it does feel like we're still on this trajectory where if we do move to an agentic web, if people are primarily just interfacing with ChatGPT, with Claude, with Gemini, that who are you really advertising to? At the end of the day you're advertising to the AI models, in which case do you need the agencies as much in there? How much human intervention do you actually need at that point? Which is the big question. We're not going to answer.
- [51:03]
D
Certainly not. I just don't want us to end up being the movie Wall E becoming a documentary.
- [51:10]
B
Oh, I feel like it already.
- [51:12]
D
We're kind of there.
- [51:13]
B
Yeah, I feel like we're watching the prequel to Wall E. Fair enough. I feel like we made a lot of Headroom today, but obviously there's still a lot to be sorted out. Seb Bergi, appreciate you coming on.
- [51:27]
D
Thank you so much.
- [51:28]
C
Thanks for having us.
- [51:35]
A
Well, that brings us to the end of this episode of the Digiday Podcast. Thank you to everyone for listening. And please don't forget to share this episode with someone who you think would enjoy it. You can even rate us and leave us a comment on Apple Podcasts. We'll be back next week with another episode of the Digiday Podcast. Thank you so much for joining us.