What the agentic AI era means for ad agencies, with Omnicom’s Jonathan Nelson

Summary

The Digiday Podcast: Exploring the Agentic AI Era for Ad Agencies with Jonathan Nelson of Omnicom

Release Date: January 14, 2025

In this insightful episode of The Digiday Podcast, hosts Kamika McCoy and Tim Peterson delve into the transformative impact of the agentic AI era on advertising agencies. Featuring a comprehensive interview with Jonathan Nelson, CEO of Omnicom Digital, the episode navigates through recent industry upheavals, technological advancements, and strategic mergers shaping the future of digital marketing.

---

1. Setting the Stage: Industry Dynamics and Recent Events

Wildfires and CES Impact

The episode opens with Kamika and Tim reflecting on a tumultuous week marked by devastating wildfires impacting Los Angeles, where the annual Consumer Electronics Show (CES) took place. Tim shares personal anecdotes about witnessing the intersection of pressing real-world crises and high-tech exhibitions:

"At CES, a lot of people who live in LA were dealing with evacuation concerns, coordinating with loved ones amidst wildfires. It was surreal and made conversations often pivot from wildfires back to AI and tech innovations." (02:54)

Corporate Shifts and Layoffs

The hosts transition to discussing significant corporate news, including layoffs at Publicis and strategic shifts at major companies like Meta and Disney. Notably, Meta's rollback on fact-checking measures under the guise of enhancing free speech sparks debate:

"Mark Zuckerberg is positioning these changes as bolstering free speech, but it's clear there's a nuanced narrative at play, especially with the incoming administration." (05:04)

Similarly, Disney's acquisition of Fubo and the subsequent cancellation of Venue Sports highlight the volatile nature of media conglomerate strategies:

"Disney initially planned to launch Venue Sports but faced legal challenges from Fubo. Ultimately, Disney opted to dissolve Venue altogether, marking a significant shift in their streaming strategy." (12:08)

---

2. Introducing Jonathan Nelson and Omnicom’s Strategic Vision

Omnicom's Legacy and Recent Merger

Jonathan Nelson brings a wealth of experience as the CEO of Omnicom Digital, accentuated by the company's recent strategic merger with Interpublic Group (IPG). This merger is portrayed as a pivotal move to consolidate data assets and enhance technological capabilities:

"With the IPG acquisition, we now sit on the largest data set to begin with. This empowers us to reinvest heavily in technology and expand our platform's reach across more clients globally." (23:21)

---

3. The AI Revolution: Transforming Agency Compensation and Operations

Shifting from Billable Hours to Outcome-Based Models

A central theme of the episode is the urgent need to overhaul traditional agency compensation structures. Jonathan emphasizes the unsustainability of billable hours in an AI-driven landscape where tasks are automated and completed more efficiently:

"We're moving towards outcome-based compensation. Clients understand the immense investments we've made in platforms like Omni, and it's only fair that our compensation reflects the tangible outcomes we deliver." (25:14)

Challenges in Outcome-Based Models

Transitioning to outcome-based models isn't without its hurdles. Jonathan acknowledges the complexities of attributing success solely to agency efforts amidst variables beyond their control:

"We need to work on a client-by-client basis to define desired outcomes and have transparent conversations about how we can be compensated based on those results." (27:40)

---

4. Embracing Agentic AI: Redefining Marketing Strategies

Understanding Agentic AI

The conversation delves deep into the concept of agentic AI—intelligent agents that interact with consumers autonomously. Jonathan explains agentic AI as sophisticated assistants capable of handling complex tasks by leveraging extensive data and personalized preferences:

"An AI agent remembers user preferences and handles tasks like booking travel itineraries, making decisions based on individual likes and dislikes, thereby acting as an intermediary between brands and consumers." (34:25)

Marketing to AI Agents: The New Frontier

Jonathan draws parallels between current SEO practices and the emerging need to optimize for AI agents. Marketing strategies must evolve to target these intelligent intermediaries effectively:

"Marketing to AI agents is akin to SEO but exponentially more complex. It's not just about keywords anymore; it's about understanding and influencing the decision-making processes of these agents." (34:57)

Navigating the Complexity of Agentic AI

The hosts discuss the multifaceted challenges of this new marketing paradigm, including the necessity for advanced data integration and adaptive strategies:

"The world's getting more complex, and so are the AI models. Teams need to adapt by understanding the distinct capabilities of various AI models and leveraging them to meet diverse client needs." (35:34)

---

5. Data as the Backbone of AI-Driven Marketing

Leveraging Extensive Data Sets

Jonathan underscores the critical role of data in powering AI-driven marketing initiatives. The merger with IPG has fortified Omnicom’s data repository, enabling more sophisticated audience targeting and outcome tracking:

"We have tens of petabytes of behavioral and transaction data. This comprehensive data set is the foundation for our AI initiatives, allowing us to deliver precise audience activation and measurable outcomes." (26:53)

Integrating Diverse Data Sources

The conversation highlights the importance of federated data models over isolated data silos, advocating for a more integrated approach to data utilization:

"There is no single data set to rule them all. It's about creating a federated model where multiple data sources work cohesively to inform our strategies and drive results." (38:33)

---

6. Future-Proofing: Anticipating AI Innovations and Challenges

Advancements in AI Models

Looking ahead, Jonathan discusses the evolution of AI models, particularly the emergence of "chain-of-thought" capabilities that enhance reasoning and problem-solving:

"Chain-of-thought models break down problems into subsets, enabling more complex reasoning and accurate computations. While they're still developing, we expect these models to become more reliable and cost-effective by 2025." (41:45)

Practical Implementation and Training

Implementing advanced AI requires substantial training and adaptation. Omnicom is proactively equipping its workforce with the necessary skills and tools to harness these technologies effectively:

"We're developing Omni AI, which integrates multiple foundational models. We provide tutorials, office hours, and documentation to ensure our employees can navigate and utilize different AI models based on task requirements." (29:32)

Economic Considerations and Sustainability

Jonathan touches on the financial implications of deploying advanced AI, acknowledging the high compute and operational costs involved:

"AI development is expensive, involving significant investments in compute and data infrastructure. However, our strategic mergers have positioned us to absorb these costs and continue advancing our technological capabilities." (24:23)

---

7. Concluding Insights: The Road Ahead for Ad Agencies

As the episode wraps up, Jonathan reinforces the necessity for agencies to embrace data-driven, outcome-focused strategies in the AI era. The merger between Omnicom and IPG serves as a testament to the industry's shift towards consolidation and technological integration. The future of marketing lies in effectively leveraging AI agents, comprehensive data sets, and adaptive compensation models to drive meaningful outcomes for clients.

---

Notable Quotes

• Jonathan Nelson on Outcome-Based Compensation:
  "We're moving towards outcome-based compensation. Clients understand the immense investments we've made in platforms like Omni, and it's only fair that our compensation reflects the tangible outcomes we deliver."
  (25:14)

• On Marketing to AI Agents:
  "Marketing to AI agents is akin to SEO but exponentially more complex. It's not just about keywords anymore; it's about understanding and influencing the decision-making processes of these agents."
  (34:57)

• Jonathan on the Importance of Data:
  "We have tens of petabytes of behavioral and transaction data. This comprehensive data set is the foundation for our AI initiatives, allowing us to deliver precise audience activation and measurable outcomes."
  (26:53)

• Future of AI Models:
  "Chain-of-thought models break down problems into subsets, enabling more complex reasoning and accurate computations. While they're still developing, we expect these models to become more reliable and cost-effective by 2025."
  (41:45)

---

Final Thoughts

This episode of The Digiday Podcast offers a compelling exploration of how the advent of agentic AI is poised to revolutionize the advertising landscape. Through Jonathan Nelson's expert insights, listeners gain a nuanced understanding of the strategic shifts required for agencies to thrive in an AI-dominated future. From reimagining compensation models to harnessing vast data ecosystems, the conversation underscores the imperative for adaptability and innovation in the digital age.

For those navigating the complexities of modern marketing, this episode serves as an essential guide to the transformative forces reshaping the industry.

Transcript

[Jonathan Nelson]

Foreign.

[Kamika McCoy]

To another episode of the Digiday podcast. I'm your co host, Kamika McCoy, senior marketing reporter here at Digiday.

[Tim Peterson]

I'm Tim Peterson, executive editor of video and audio at disjay Media. God, Kimiko, it's been a week. I almost forgot what the hell my job is. How are you doing?

[Kamika McCoy]

I can imagine that I may be doing holding up just a wee bit better than you. You are coming straight off of ces. I'm sure you're hearing colors at this point. How was it?

[Tim Peterson]

It was a week. I mean, it's been a week and a lot of people have had rougher weeks than me. That was in Los Angeles, the Los Angeles area now with all the different buyers. And that was the one of the surreal parts of ces. I'm from LA originally. My parents, my sister, aunts, uncles are all in la and so getting all the texts and photos of everything coming in. And then just at ces, a lot of people who live in Los Angeles were at ces. They're having, you know, one person I've met with was up until 2am the night before on the phone with his wife because they, they weren't in an evacuation zone, but they were near enough one where she felt like she needed to stay up just in case they had to make a move. Because with these winds have been so wild that you don't know when things are going to switch. So. And then people needing to fly back to take care of their homes or take care of loved ones. Yesterday I was writing an article in just like one of the hotel lobbies and the person, you know, sharing a bench with me was on the phone trying to coordinate evacuation routes for their parents who were in one of the zones. So it was just a wild week. But I think for a lot of folks there was almost a CS was almost a welcome distraction. It wasn't really a welcome distraction because with this kind of thing you can't really be distracted from it. But I know again, I had a few meetings with folks who live in la and we would start off by talking about the wildfires because how could you not? And then it was just like, okay, so should we talk about flying cars and robots now and AI? So yeah, yeah. Bizarre week, long week. But I don't know. There was a TikTok I saw yesterday when I was at the airport waiting for my flight or a person who's just like the seven day free trial of 2025. I would like to cancel my subscription.

[Kamika McCoy]

I was going to say 2025 is starting off with a very strong 2020 vibe. Not. Not great at all. We're only, like, less than a month in and shit's already hit the fan, so. Yeah, absolutely. Hoping, you know, for the safety and everybody that's been affected by that. It's absolutely insane. Glad that you guys were able to have CES through all of that, though. Welcome distraction or not. Talk to me about some of the highlights besides the flying cars. I'm sure AI was a big talking point there.

[Aaron Luber]

AI was a big talking point.

[Tim Peterson]

Also, this. It wasn't a big talking point, but. So for the episode this week, we have Jonathan Nelson, who's the CEO of Omnicom Digital and a CES veteran. I think he said he's been going to CES for 30 years at this point, so he's kind of seen it all. And so he hadn't made it to the show floor yet, but he's very plugged in, so I asked him what he was looking forward to seeing. I spent a little bit of time on the show floor. Not a ton. Did not see this. Human drones.

[Kamika McCoy]

Oh, great. It just keeps getting worse.

[Tim Peterson]

He just drops it casually into the conversation. Listeners will be able to hear it when we get to the interview, but he kind of just like, throws it off, like, on a list of things, and I was just like, wait a second. Did you just say human drones? Excuse me? I guess it's like, just big drones that can transport humans. So, like, quadcopters effectively sounds to be the thing. It's just like, okay, so another thing for New Jersey to worry about at night.

[Kamika McCoy]

Yeah, that'll be great. Super exciting. Also, you. You. You know, you did the podcast on the ground this year. This is ces, so this episode will be exciting. But before we get to the actual guest interview, we've got a busy, busy Newsweek. That happened. I feel like so much happened this week, and so much of it got crammed into the beginning of the week. You had layoffs at Publicis, losing their chief diversity officer. Free speech, I guess you could say, is in free fall with Meta pulling back on its. Its.

[Tim Peterson]

Although Meta would, I think, make the argument, like, no, we are bolstering free speech. Or the. The term that Mark Zuckerberg seem have really embraced now of free expression, which, you know, free expression as opposed to, you know, truth and facts.

[Kamika McCoy]

Yes. Oh, the fact checkers. That's what happened. That's done. And then we've got Disney and Fubo, not Fubu, as I'm sure. I'm gonna butcher it at some point this episode, but we'll talk about those things.

[Tim Peterson]

The number of meetings I had this week in which, you know, I was talking with folks about Disney buying Fubo and, you know, kind of settling the venue lawsuit and that people like couldn't help themselves but call it FUBU instead because I think everyone just had CES brain at that point.

[Kamika McCoy]

Oh, yeah, absolutely. Damon John, who was the founder of Fubu somewhere, is having a heyday with all that free SEO, I'm sure. So let's start with talking about, like I said, the free speech in freefall. Meta, obviously, I think, made some really big headlines. This, not that they haven't before or that they regularly don't, but talking, you know, Mark Zuckerberg got on Facebook video and essentially told the world about some rollbacks that they're going to do, kind of in line with what's happening or what Elon Musk did at X. So fact checkers are being pulled back on and the company is, as you said, saying that it's bolstering free speech. That is, you got to read between the lines there a little bit. But that's, that's kind of the message there. That's, that's happening. And what's interesting is that this happening at a time when, you know, companies are facing so called conservative backlash, aligning themselves with the incoming administration and whatnot. So I think there's a greater narrative happening here for sure.

[Tim Peterson]

Yeah. And it seems like there's a level of confidence that because, you know, a few years ago there was all the talk about a Section 230 repeal, which, you know, was effectively what, you know, protects platforms like Google, Facebook on down the list from being responsible for what people say on their platforms. That never came to pass. And now given how all these platforms are acting, it seems like it won't come to pass because they just want to not be liable for what people say on their platforms. It also seems like they very much want to cater to the right these days because of the new administration, how the election shook out.

[Kamika McCoy]

I'm actually very curious as to how this is going to shake out specifically for advertisers and their spending here. My gut tells me that this changes nothing. But given, you know, some of the pull, the pullbacks that we saw in previous years, like the mass exodus, if you can call it that, that happened with X when Elon Musk first took it over. Now that Meta is kind of aligning itself with some of those same values, what then then happens to, you know, the ad dollars that were being pumped into that space. I say they continue. But you know.

[Tim Peterson]

Yeah, or if I think if there is any change, ad buyers are going to be very quiet about it and not tie it to this because they're not going to want to get sued. Like when X went after Garm last year.

[Kamika McCoy]

That is a really, really good point. And Carm actually dissolved in the face of that. So you know, I think there's validity to the, the fear around that which kind of ties to the next topic, which is this, you know, I guess you can call it backlash to, to de and I. That's happening. So earlier this week, McDonald's made some changes kind of in the same line as what we've seen with now Walmart and a handful of other companies. McDonald's came out with a press release after an internal email came out where they are retiring and I'm quoting them. Retiring, setting aspirational rep representation goals. Excuse me. And instead focusing on continuing to embed inclusion pract into the business. Now who's to say what that means? They're also changing the name of how they refer to that diversity team as the Global Inclusion Team, which there seems to be this rebranding of DE&I or bending the knee to conservative backlash, depending on who you ask.

[Tim Peterson]

Oh yeah. No, it's the latter. It's absolutely the latter. These companies don't want to have happened to them what happened to Bud Light. They don't want to get called out for D and I commitments. They're not committed to D. And that has become very clear and a very big trend over the past year. And honestly it pisses me off.

[Kamika McCoy]

There's the, the, the walk back that's happening. Right. I think a lot of this is going to be determined if these are as sincere as they're making it seem if the ad dollars follow. Because if the goal is to embed DE and I into your everyday business practices, that should be more money flowing into that space. Right. If we see that happen, I think will be the determining factor. So new fronts, upfronts and kind of how those dollars are invested will. Will really say a lot about what the intentions are for this.

[Tim Peterson]

Yeah. And if you know to what extent you like Disney. I don't remember if this was 2020 or 2021, but made a very direct effort to push dollars towards like it's multicultural programming, if I'm not mistaken. Like they like actually like were very explicit in requiring commitments around that as part of like the broader upfront commitments. And you had had in the past the group Ms. Of the world having these commitments. Similarly, of we're going to spend X percent on, I believe it was minority owned media companies. That's going to be very interesting to stay on top of and whether these companies continue with that, whether they meet the goals that they had set or they just let that. Sarah Guadaleoni, our senior media reporter, has been tracking the diversity reports that media companies put out. Not all of them are putting those out as regularly as they used to. And it also seems like some of the progress has slowed a bit. I think there still is progress being made, but not to the same degree. There's also other factors at play. Like the media business is really tough. A lot of publishers are going through layoffs. Vox Media just had layoffs again this week. You mentioned Publess is having the layoffs. So hopefully it's something where we continue to see the progress like you're saying. But given how things have been going, given how this week has been going.

[Aaron Luber]

I'm not going to hold my breath.

[Kamika McCoy]

Yeah, yeah, I'm with you on that one.

[Tim Peterson]

But sorry, let's actually, I'm just really in it right now.

[Kamika McCoy]

It's just we're feeling Tim's wrath on the podcast today.

[Tim Peterson]

A little bit. A little bit.

[Kamika McCoy]

Let us also double back to Disney. They made some headlines this week, which I'm sure is gonna, you know, add to your plate of reporting. But the streaming service, Venue Sports, I think we've closed the loop there. Talk to me about what's happening.

[Tim Peterson]

Yeah, if, if anything, it may simplify my reporting going forward. So what was it last January? Yeah, it was like around this time last year. Disney, Fox and Warner Brothers Discovery say, hey, we're going to launch a streaming service altogether. We have all our individual streaming services, but we're going to create this one specifically around sports programming, live sports, because 2024 was a big year for live sports in streaming, as we've talked about and you had talked about on the podcast before. And I almost said fubu, Fubo, which is a streaming pay TV service, took issue with that because I think it wasn't going to be able to sell Venue Sports or kind of like co sell Venue Sports as part of its subscriptions. And so it filed a lawsuit to block the launch of Venue for anti. You know, I think it was specifically over anti competitive concerns. Judge said, yeah, Fuba, you got a point. We're going to not let Venue launch. And so that didn't happen in September as it was planned. Make, you know, come Monday of This year, Disney says, hey, Happy New Year, everyone. So, no, the venue lawsuit, like that's.

[Aaron Luber]

Been settled because actually, we're just going.

[Tim Peterson]

To buy Fubo, so. So we're going to take a 70% of ownership of Fubo, and this clears the path for venue if we would like it to. Which is kind of a wild thing to do, if you think about it, because it's like Fubo says, oh, this thing shouldn't launch because of anti competition. And then you have one of the companies just buy Fubo to effectively remove competition anyways that gets announced. Disney says, we're also going to combine, or kind of combine Fubo and Hulu's live TV service, but we're also going to continue to operate them separately, which is just. I don't. That was something where I was just like, all right, let me focus on ces, because that doesn't make any sense to me. I can't deal with this right now. Fast forward to Friday at the end of the week, and Disney, Fox and Warner Brothers Discovery say, actually, no, we're going to just scrap venue altogether. We're not going to launch it. So never mind. Everyone, I know this has been a big topic over the past year. Sorry, just, you know, forget about it. Let's move on with our lives.

[Kamika McCoy]

I love a whoopsies after. After careful consideration. Press release.

[Tim Peterson]

Yeah, I mean, good for. Good for Fubo, because, like, they won out. They got an exit out of this whole thing. Whereas for the rest of us, feels like it's been a whole waste of time. Although it never actually launched. So not a huge waste of time, I guess. Just something we were able to talk about and laugh about and, I don't know, maybe, you know, venue is the new Quibi in some respect.

[Kamika McCoy]

Oh, man, R.I.P. that's a case study that should never stop being studied.

[Aaron Luber]

Yeah.

[Tim Peterson]

Yeah. And now we can add venue to the list.

[Kamika McCoy]

Yeah. If nothing else, the start of 2025 has been a busy one, but we've got an interesting guest this week. As you mentioned, you recorded it on the floor at cs. So talk to me one more time about who we're going to have talking to us soon.

[Tim Peterson]

Yeah. Jonathan Nelson, who is the CEO of Omnicom Digital. And so we were able to talk a lot about. Well, about a lot of things from CES specifically. But then also last month, Omnicom and IPG announced this big merger, Omnicom acquiring ipg. So we also get into that, how that is kind of coming together. The conversations that he's having around that acquisition. So he was actually like fairly forthcoming and being willing to talk about the merger, which is, you know, helpful to have. But then we also talked all about what was probably the biggest topic of conversation, two of the biggest topic of conversations, at least I had during CES. And anyone who's checked our daily CES briefing on digital.com will be able to, you know, find out more there. But Jonathan, I go deep in both topics. One being the how agency compensation models are going to need to change in the AI era with agencies get paid on a billable hours basis. But if you're having AI complete a lot of that work and AI is doing it a lot quicker, billable hours can shrink significantly. So there needs to be a new compensation model. So Jonathan offers his thoughts on the compensation model he thinks agencies should move towards and kind of the complications in doing that. I won't spill too much, people will be able to hear it in the interview. And then the other one being this era of AI agents or agentic AI, which is the new vocabulary I've had to add to my or the new term I've had to add my vocabulary. This idea that as eventually you're going to be able to just have ChatGPT or whatever book your travel for you. You'll just say hey chatgpt, I'm looking to go to Europe later this month. You know the kinds of places I like this to stay, you know, what seats I prefer, what time I like to fly out, so just handle the booking of my entire itinerary. And ChatGPT is the one who would then me making the decision on like are you flying Delta? Are you staying in a hostel or a five star hotel? All down the list. And so what marketers are having to think about now is how do we start effectively marketing to these AI agents. What's the search engine optimization for this agentic AI era? And so again the CES briefing will cover a lot of that conversation. But with Jonathan we also dig into how they're thinking about that, which conveniently also ties into the Omnicom IPG merger, as does the conversation around compensation models. So it's a pretty. I thought it was a really interesting conversation. I hope the listeners do too.

[Kamika McCoy]

Love that, Love that. I haven't said juicy scoop at all during this intro, so I'll use it now. I'll let you guys get to the juicy scoop.

[Tim Peterson]

Thanks Kamika.

[Aaron Luber]

Jonathan Nelson, welcome to the podcast.

[Jonathan Nelson]

Thanks for having me.

[Aaron Luber]

Thanks for doing this live at ces.

[Jonathan Nelson]

Live early in the morning at CES on Day four. Let's see how this goes. Yeah.

[Aaron Luber]

Long week and I imagine especially like for you being at Omnicom, coming off the IPG announcement, I imagine that's dominated a lot of conversations you're having this week.

[Jonathan Nelson]

Yeah, IPG is a hot topic. AI is a hot topic. And I mean, we just have a lot of usual meetings and CES is fun because we know a lot of. You know, I've been at this for. It's my 32nd year, so I know a lot of people here.

[Aaron Luber]

Okay, so you go. I mean, you go back to. What was it called before Computex?

[Jonathan Nelson]

Well, it was ces. It's always been ces. In fact, my. I can't figure out this is either my 40th or 41st anniversary of CES. I used to go when I was a kid because it was all stereos and TVs in Chicago. But. Yeah, it's been a long time. Yeah, a lot different now because.

[Aaron Luber]

Have you even made it over to the. I feel like no one makes it over the show floor anymore.

[Jonathan Nelson]

We're going to do the show floor, right. As soon as we're done with this. Yeah.

[Aaron Luber]

Anything in particular you're looking to check out over there?

[Jonathan Nelson]

I think we. When you see it over such a long period of time, you can sort of benchmark how it's evolving, you know. And of course we're looking at TVs and self driving cars. You know what's interesting? Human drones.

[Aaron Luber]

Wait, human drones?

[Jonathan Nelson]

Yeah. Drone taxis are a big thing.

[Aaron Luber]

Flying over New Jersey.

[Jonathan Nelson]

Flying over New Jersey. Actually, they have the air quarters already set up. They're saying it's going to happen next year. But you watch the evolution of technology and the proliferation of things like AI infusing into everything.

[Tim Peterson]

Okay.

[Aaron Luber]

And I mean, I think one of the things I struggle with, with the CEO show floor and beyond, just how massive it is. But how much of the things over there feel so far in the future that it's hard to understand, like what's tangible now. So when you're over there. Cause I imagine you may be over there with clients as well and trying to, you know, help them decipher how to make sense of what it is that they're seeing over there. Because you go over there, you've been, as you mentioned, coming here for a long time. So there's the tech side of you that's just interested in these things. But how do you look at any of that from kind of the omnicom lens?

[Jonathan Nelson]

Well, I'd like to think this is where Experience matters. After looking at the same thing over and over and over for many, many years. One, you've watched evolution too. You've developed a point of view about what's real, what's not real. Three, I find myself saying, okay, that's gonna get released right away. That one, maybe three, four years out and that will never happen.

[Tim Peterson]

Okay, got it?

[Jonathan Nelson]

Yeah.

[Aaron Luber]

Because I was with raja Minar from MasterCard yesterday. He was over the show floor. He's like you, where he is very interested in the tech. Has been coming to see us for a long time. The thing that he said stuck out to him was, I think it was at lg they have like a digital out of home activation where it's billboards with face tracking built in and the ad kind of dynamically changes, which is interesting. Obviously there's all kinds of privacy considerations when it comes to face tracking in the public like that. Because the face tracking can happen before there's even consent necessarily given.

[Jonathan Nelson]

Yeah.

[Aaron Luber]

But then it's also the kind of thing of well, how feasible is that really? Because on the one hand it feasible feels like, okay, that's something that could roll out tomorrow. On the other hand, cycling through all the different creative, having all the different creative doing the kind of targeting, accuracy, measurement side of things, it feels like, oh, but that could still be five years out.

[Jonathan Nelson]

I mean other than the face tracking source, the tech to do the rest of that's here today. I mean all the targeting, the creative development, the automation, that's exactly the kind of stuff we're doing and actually the kind of stuff we're going to do even more with ipg.

[Aaron Luber]

In what ways?

[Jonathan Nelson]

Well, if you think about the IPG acquisition, we will have a broader platform to do things. We will have the broadest data set on the buy side anywhere in the world and more expertise, more clients. I think it sets us up really well for the future as well as we'll have a much broader, much better cash flow. Okay, so larger base, not only of capital to put into the platforms, but also taking the platforms and putting those, deploying them across a broader surface area of companies and clients.

[Tim Peterson]

Right.

[Aaron Luber]

Plus you'll have all the. Because what is it the cost savings that you all are estimating? 750.

[Jonathan Nelson]

Yeah, it's really large. I've heard there's been a couple different numbers, but yeah, it's in the hundreds of millions.

[Aaron Luber]

A lot of money to play around with.

[Jonathan Nelson]

It's a lot. Yeah. I don't think, I mean that's not cost savings. That just goes Straight into whatever. I think what we're looking at is operating capital. So if you think about our business, say we made a dollar in operating income. IPG is about 2 thirds our size. So instead of having a dollar, we now have $1.66 in operating income. And so that in turn gets reinvested back into technology or dividends or servicing debt or whatever. But quite a bit's going to go into technology. Right. Data. Yeah, all these things that everybody's talking about that's, you know, we have a broader platform to invest in. Right.

[Aaron Luber]

Because I mean, as much as there's the cost of human employees and infrastructure buildings and all that AI requires data, that requires a lot of money in terms of the support, processing, keeping servers running, all of that.

[Jonathan Nelson]

Yeah. The compute costs are very high. And one of many benefits of the IPG deal is we're sitting on the largest data set to begin with. And this is only accretive on top.

[Aaron Luber]

Of that, how long will it take to bring the data sets together?

[Jonathan Nelson]

So some of the data sets are going to come together almost immediately and then over as we close, it'll get more tightly bound in.

[Tim Peterson]

Okay, got it.

[Aaron Luber]

Yeah. Because I mean, that gets really interesting when it comes to, well, all of this. We're talking about the financial side of it. I've been having a lot of conversations this week with agencies, with people like Raja about agency compensation models and how those need to change in the AI era because it's so many of the, so much of the time it's billable hours. Everyone's working off of. How do you do billable hours when some of the work is being offloaded to.

[Jonathan Nelson]

Yeah, that's very insightful. Right now we get paid for, you know, full time employees, hourly rates, you know, what boils down to an hourly rate over time. Yeah, there's huge amounts of investment in R and D, in compute, in data, all of those things. It's only fair that we get compensated. And I think one of the great ways that we're looking at this is with outcomes. And people have maybe not seen the connective tissue between what we did exactly this time last year, which was really implement the acquisition of Flywheel, which is the largest buy side commerce business, largest buy side commerce data set we did in January of 2024 and then we bookended it with the IPG acquisition announcement in December of 2024. Connect those two things together. We are now sitting on what is the largest and arguably best the gold standard of audience data. In Axiom, we have Omni which is our, our media orchestration platform, sitting inside of Omni, sitting on tens of petabytes of behavioral inventory pricing data. We're sitting on arguably some of the most creative businesses in the world. So we're moving into a content world with AI and then connecting it to outcomes, connecting it to commerce, connecting it to Flywheel Commerce Cloud, which has tens of petabytes of data itself. So to us, this is check, check, check, Audience activation outcomes. We have the best data set across all of it, which will inform the AI.

[Aaron Luber]

Okay, I mean, that's really convenient because one thing Raja mentioned is he wants to move to an outcome based compensation model. I think he said in the next 12, 18 months. MasterCard is a client of McCann, so that's going to be something you're inheriting there with outcomes. So I was also talking with Amy Lanzi from Digitoss about this about kind of compensation models, and one thing she was saying is, yes, we should move to outcome based compensation. The challenge there is not all clients are set up to be able to have the measurements that are needed for outcomes. And as an agency, one of the risk is the bet that you're taking that the outcome is reflective of the work that the agency did, as opposed to a retailer just not keeping a product in stock or a salesperson not doing their job or calling in sick.

[Tim Peterson]

Or something like that.

[Aaron Luber]

How do you think about that?

[Jonathan Nelson]

Yeah, that's a fair criticism. And it's something that we need to work out and it needs to, I think we do this on a client by client basis, figure out kind of what the desired outcome is, figure out how we can drive towards that result, and then have a open and honest conversation about how we can be compensated for that result. Yeah.

[Aaron Luber]

How much of the conversations you're having this week with clients are touching on this topic?

[Jonathan Nelson]

To be honest, I've been pretty heads down, but the clients, and I don't touch clients as much. I do touch some. It has come up in every single conversation. Clients get it. Yeah, yeah, clients get like you guys are spending, you know, truth is, we've spent a billion dollars over 15 years on Omni. And they can see it, they use it, or they know that we use it every day. It's what makes everything happen. It orchestrates it all together and yet it's bundled in with the people and they know that that is backed by nine figures of data that we pay for. This isn't cheap and it's not getting cheaper. Right. AI, why is OpenAI raising hundreds of billions of dollars? It's because of the compute, the electricity. We got to have a different conversation.

[Aaron Luber]

Right. Well, and plus, it's not like there's going to be a one LLM rule like model for every company that you're just going to be a GPT shop or just a cloud shop.

[Jonathan Nelson]

No, in fact, we have something we're developing called Omni AI that uses many, many, I mean all the major foundational models and many slightly more obscure models that we orchestrate together. And then we're putting that on every employees desktop and Omnicom right now.

[Tim Peterson]

Oh wow.

[Jonathan Nelson]

And that's live. I could show that to you.

[Aaron Luber]

And so I mean, what's the training that needs to be involved there for the average employee to know? Okay, do I want to use GPT for this or do I want to use Gemini for.

[Tim Peterson]

Whatever.

[Jonathan Nelson]

We guide them, we give them tutorials, we have office hours, we have written documentation. People learn in different ways. And so our model has always been to create very, very open systems like Omni. This, this is, I said as an extension, it's Omni AI. And so we give everybody most of the models that we think are deemed fit for production. And in the back office we are testing, I don't know, probably 15 to 20 models at any given time. You know, maybe 10 of them are on the desktop.

[Aaron Luber]

Okay. But even then, I mean, that's still because I have a hard time figuring out what the different models are best suited for, especially with the updates that are coming out.

[Jonathan Nelson]

Yeah, it's changing at lightspeed, but you'll find, I mean, if you play around, for instance with the GPT models, the text models, you know, Google's a little bit more analytical. The Gemini models, the Anthropic is a little bit more creative in its writing. I would say that the Nova model, which is great from Amazon just came out, is somewhere in between, as is ChatGPT. I mean that's just four. And Llama is kind of right in there as well.

[Tim Peterson]

Right.

[Jonathan Nelson]

So I think you see that if you work with it. And so what we're really trying to get at is, look, we're not. This isn't a sermon from the mount. Here's Omnicom's point of view. You should use this model. We're saying, hey, figure it out. If you're doing technical writing, if you're doing pharmaceutical or healthcare, you're gonna want something that's like a little bit more constrained, a little drier. If you're doing more creative copy, you might want something that's a little bit More florid, a little bit more narrative. Lyrical.

[Aaron Luber]

Yeah. It feels like at some point there's going to need to be some version of media mix modeling for these large language models.

[Jonathan Nelson]

And you might find large language models actually orchestrate that. So also something to, you know, something in the labs.

[Aaron Luber]

And then someone starts complaining about, oh, ChatGPT is grading its own homework. And we're just kind of back to the conversation.

[Jonathan Nelson]

Easy to get them to grade each other's homework.

[Tim Peterson]

Fair enough.

[Jonathan Nelson]

Yeah. It's a very, very interesting moment in time. I mean, on the tech side, we're like kids in candy store. I mean, here we are sitting on this massive data set, like it's coming together across, as I mentioned, audience activation, outcomes. It has that purpose which is driving towards outcomes remuneration. But we also have these incredible creative tools that nobody could have dreamed of. Five years ago, in fact, we were talking about cs. Yep. You know, five years ago, none of this existed. Right. We were playing with AI, we were playing with neural nets, we were playing with reinforcement learning. But, you know, transformer models were a. It was a paper.

[Aaron Luber]

Right.

[Jonathan Nelson]

Or maybe if you were really in tune, you may have known about a couple companies in the Bay Area playing with this, but you didn't see it.

[Aaron Luber]

On the floor here, let alone hear about it at some C space where all the.

[Jonathan Nelson]

Where now everybody's talking about it. Yeah, absolutely.

[Aaron Luber]

Well, and one thing everyone's talking about this year is this agentic era of AI. The idea of AI agents, I think it was. Aaron Luber from Google was on a panel the other day Talking about Gemini 2.0 being released in December, and that really being their play towards this agentic era of AI. For even for me, who's been talking about agentic all week, AI agents all week. I'm still kind of struggling a bit to understand what that means beyond what feels like the stereotypical example of, well, then you can have an AI agent just kind of book your travel for you, set up your itinerary, book your flights, all of that thing. Could you kind of break down as simple as possible? What exactly are we talking about when we talk about the agentic era of AI?

[Jonathan Nelson]

Well, first of all, there's not hard and fast definitions for this or AGI or singularity or any of these things that people are talking about. But when it comes to agents, it's think of it as like an assistant. You can teach it, it has memory, it knows what you like. So if you're, you know, Apple's making a big deal about this Right now. So over time you're interacting with your phone, your computer, whatever. Say you're booking plane tickets. It knows that I like to fly in business class aisle seats. I don't like to take red eyes. You know, I can, I like to order a certain kind of meal. Right. Here's my frequent flyer number. Book the ticket. So instead of doing each of those manually, like we all do whenever we're trying to do travel, the agent will just remember what you like and it will do it for you. Right.

[Aaron Luber]

And how imminent of a concern, maybe is the wrong word, consideration. But like clients, to what extent should they be thinking about getting these AI agents up and running on their end? But then also the idea, and I think it was Amy Lanzi from Digitoss who made this point yesterday too, of needing to brands needing to have content out there. So these AI agents went and someone asked like, hey, what's the best airline? I should be on a given brand. It surfaces basically like the new version of SEO.

[Jonathan Nelson]

Yeah, that, that's the absolutely correct analogy. Which is SEO was always essentially marketing to an algorithm at Google to try to get your content well placed. Right. Essentially we were marketing to a version of a robot. It wasn't an agent or, you know, AI agent. Well, I guess there were some parts of AI in there, but. But the idea was you weren't really talking to a person directly, you were talking to an intermediary that was a computer who in turn would talk to a consumer. I think you're going to see the same idea times a trillion.

[Aaron Luber]

Yeah. Because it feels like that's like 4D, 5D SEO. There are all these other dimensions to it. It can't be that just taking SEO practices as they were in the search era and applying them.

[Jonathan Nelson]

Yeah. It. The world's not getting less complex.

[Aaron Luber]

Understatement of the week.

[Jonathan Nelson]

Yeah, exactly. And I think. And having teams of people that understand the world and where it is, how to get this done, where it's going, is going to keep us in business for a good long while here. Right?

[Aaron Luber]

Yeah. Because I mean, at what point do we. Do you all have to start thinking about clients actually advertising to the AI agents because they're effectively the proxies.

[Jonathan Nelson]

When do we start? Like right now. I mean, we've started like.

[Aaron Luber]

Yeah, but like running like actually paid media for that as opposed to just SEO, which has always been an organic thing.

[Jonathan Nelson]

Well, we're trying to figure, I mean, look at what's going on. Like perplexity. Look at Google's AI assist. Right, right. It's already happening.

[Tim Peterson]

Okay, so that.

[Jonathan Nelson]

So, yeah. How do you do it? Well, it's a lot of trial and error and say. And by the way, the algorithms are all changing, the model's changing, like everything's changing, but it's already happening in search. So this is only going to proliferate.

[Aaron Luber]

And so is that something where like, search budgets are having to be carved out for this?

[Jonathan Nelson]

No, it's just part of the craft.

[Tim Peterson]

Okay.

[Jonathan Nelson]

It's what we do.

[Tim Peterson]

Okay.

[Jonathan Nelson]

It's the service we offer.

[Aaron Luber]

And then, I mean, I feel like it then brings us back to the data, having the data to be able to figure out how to be talking to these AIs.

[Jonathan Nelson]

Yeah. I think it very much comes back to a closed loop data feedback loop where if you can understand audience activation and connect in outcomes, you've got the full picture there. Now can you figure out whether it worked? Did the outcome get achieved or not? If it did, bingo. Reinforce the inputs. If it didn't, adjust the inputs. That's very simplistic. But yeah, that's. And people haven't seen this. But if you think about what we've been trying to do at Omnicom, 15 years ago, everybody thought we were crazy when we started Omni. Well, started Analect, which evolved into Omni, and we slowly assembled the largest activation data set in the world and nobody even knew it. And then January of last year, we bought Flywheel, the largest transaction data set, and ipg, we bought the largest audience data set. Okay. Now suddenly everybody's going, oh, whoa, got it right.

[Tim Peterson]

Okay.

[Aaron Luber]

With like these data. So I think it was Brian Lesser from Group M earlier this week. There's the Axios story he's talking about. Oh, you know, we had been focusing on like a single data set to rule them all owned and operated data. We're going to move to this federated model. What do you make of that?

[Jonathan Nelson]

There is no single data set to rule them all. It's all changing and it's all. I mean, while we have been sitting here, we've probably bought 10 million impressions programmatically, like during this podcast. And think about all the data that just came off of all that we have, all that. This is not static. This is not a single data set. It's not like it's used by one company. It's it by nature, as I described before, open systems, federated.

[Tim Peterson]

Okay.

[Aaron Luber]

And it feels like the conversation around data, the cs, is a little bit different. It seems somewhat reflective of kind of a change in the industry where the past few Years, past decade plus, everyone's been talking about data, but past few years it feels like everyone's been focusing on getting their own data houses in order. There have been a couple people by and sell side who've made the point to me that they're having more conversations around the execution of data. How an agency or a brand can be working with a media company to connect their data, to get past kind of the theoretical conversations that have been in the past to the actually practical applications now.

[Jonathan Nelson]

Yeah, thank God. Data is a complex thing. I mean, I did the first banner back in 1994, same time did the first data analysis tool called Get Stats. Took a data analyst company public in 96 or founded 96 public 98. Did, did, did omni with a great team of people. Been at this a while. Right. Data is complex. Most people don't understand it. You got to start somewhere. And so you start by, you know, is it first party or pii? Well, they're kind of the same thing. It's taken the industry a while, but I think, yeah, we need to move to, okay, now what does it do and how do we get at it? How recent is it? What's the fidelity of it? You know, match rates, all of those things coming together. Yeah, it's not. Data for data's sake is expensive and takes up a huge amount of storage. Data for application sake is what we're actually getting at and what we've always. That's my whole career. That's what we've been getting at for the whole time.

[Aaron Luber]

So what changed that? It feels like this flip has been switched.

[Jonathan Nelson]

I think people are waking up to the power of what's going on. And the good news is we've been at it for decades and seems set.

[Aaron Luber]

To continue doing so.

[Tim Peterson]

Just.

[Jonathan Nelson]

It just gets bigger and bigger and bigger. And, you know, so it was a good thesis back 30 years ago for me. I mean, I got lucky, I guess, but. But I'm glad that people are figuring it out. And particularly as AI comes in here and you can see the amount of. Compute the number. Remember we used to talk about parameters, building parameter models. That's data. Right. People are seeing without a massive input, you don't get any output. And so I think that's kicked everybody into a different level of discussion. And we're like, bring it on. What do you think we've been doing? We've been doing this the whole time. Time. Like, I'm glad you want to join the conversation. Right? Yeah.

[Aaron Luber]

And it'll be interesting to hear what the conversation is a year from now.

[Jonathan Nelson]

But I think it's going to be a lot about agents. And then there's another technology that's emerging that I'd point out that's really, really interesting, which is this chain of thought, chain of reason. I haven't heard about this 1 gpt01 gpt03, some of the Gemini models. I mean, you're going to see all the major players do this. And what it is, is it breaks problems down into subsets and can do much more complex reason and math and all of these things. And you will see there's quite a bit going on right now. But I think Q2, Q3, you're going to really start to see these things creep in. They're still complex, expensive and hallucinate a lot. They'll calm down just the same way diffusion models and chat models have calmed down over 20, 24. You're going to see these chain of reason models calm down in 25 and get quite useful. Now, whether we can afford them is another.

[Aaron Luber]

I think I already talked about how it's losing money on GPT Chat GPT Pro, which is $200 a month.

[Jonathan Nelson]

Yeah, well, they're just trying to grab market share, which makes sense in a, in a market that's exploding like this one. But, but I think what we're. What you can kind of glimpse the future with this stuff because it comes out. It's just really complex, expensive, the interface is bad, like all of this. But I think if you think it's been interesting so far, you just haven't seen anything yet. Yeah.

[Tim Peterson]

So we'll have to see in a year. Well, Jonathan, really appreciate you coming on the show.

[Jonathan Nelson]

Thanks for having me on, as always.

[Tim Peterson]

Absolutely.

[Aaron Luber]

Thanks for listening to this episode of the JJ Podcast. If you enjoyed it, please leave us a rating and a review on Apple Podcasts, Spotify or wherever you're listening. Get more from Digiday with our daily newsletter sent out each weekday morning. Visit digiday.comnewsletters to sign up.