Anthony Katzor (2:53)

Thank you, atg. Happy to be back.

Podcast Host (2:55)

Thank you. And happy holidays. This today is officially, what, the last day before we go into the New Year? Last couple of days. So thanks for joining me.

Anthony Katzor (3:04)

This is pretty much the last week before going to the New Year. Yesterday was the first evening of Hanukkah, so those celebrating. Happy Hanukkah. Yeah.

Podcast Host (3:11)

Yeah.

Anthony Katzor (3:11)

We're in the home stretch of the year and what a year it's been. Holy crap.

Podcast Host (3:13)

Yeah. And it's not slowing down, which is the good and bad part. I feel like I was expecting things to start to slow down. For me, it's not. It's actually just getting worse. And so thank you for joining me. I know your schedule's been pretty crazy with the travel.

Anthony Katzor (3:26)

This has been one of the. I've been in this industry now for. Gosh, we were just talking about how old we're getting before we started the podcast, but I've been in this industry now for quarter, a little bit over a quarter century. And I would say this has been the most hectic year in 25 plus years. And working in digital media, definitely it's.

Podcast Host (3:43)

A lot of change. And a lot of change is coming, too, which is actually why I invited you to be on the POD today. I hear a ton of industry leaders, companies, individuals just speaking about agentic, speaking about how it's going to disrupt or change the market. I've had my own personal views of how this will change the industry, but I wanted to ask you in particular, how are you feeling about agentic AI and what's happening in the industry overall as it relates to agentic?

Anthony Katzor (4:12)

I think we're. I think we're really in the early stages of how agents will impact digital media buying and selling or digital media trading. There's definitely a place for it. I think anytime we can introduce greater efficiencies through new sets of tools or protocols, I think that's a place where agentic AI can play a role. But I think we're. I think it's early days. I think there's a lot of. I think there's a lot of talk around agentic and what it could do. It does feel a little Bit like the south park underpants Gnome episode for those of that of you seen it. Step one, you know, steel underpants. Step two, step three, profit. Right. Like there's, there's this big gap as I talk to the industry of like oh, I want to do, I want to do something agentic. I'm like well what do you want to do that's agentic? Well, I want to do an agentic CTV buy. Okay, what is an agentix CTV by? I don't know. I was hoping you tell me. Like, I feel it's a lot, there's a lot of those conversations out there. I do think it introduces a very unique form or advanced form of automation, particularly through the use of natural language. I think we'll gradually see over time I think interfaces to a lot of the systems that we use to buy and sell media, to operate those campaigns. I think the interfaces will gradually evolve. I don't think the traditional UI goes away, but I think certainly a more natural language approach to describing audiences, transacting against audiences, setting up campaigns. I think that is where this will ultimately evolve to over time. I do think it's going to lead to efficiencies in the buying and selling of media. I think you'll see the most immediate impact will really be in the form of generative AI. I think copywriting and creative are already being disrupted. General Motors just ran their first all artificial intelligence ad recently. Nike has run one. Coca Cola is now on their second one around the holidays. So I do think, I think we're going to see AI, generative AI in particular really impact the ecosystem is going to start with creative and copywriting and then I think agentic will certainly bring efficiencies to media discovery, negotiation, planning, buying.

Podcast Host (6:19)

I've heard a lot of different takes. You know, ADCP has been kind of a hot slack channel. A lot of conversations happening there. Brian o' Kelly and a few others are very active. The podcast right before you I had LG Ad Solutions, the former CMO of Comscore and MediaOcean and we talked about it and I actually labeled the whole agentic era as probably the biggest hype of 2025. I've seen a lot more social posts, newsletter write ups, promotions of we are agentic and we're going to be agentic. But I haven't seen anything tangible. And so I think that's the biggest challenge is that when the hype cycle is at its peak, it's when you start to look around and wonder, okay, what's the actual product here? Like, what's the actual benefit to the ecosystem? And that's why I wanted to invite you because you work with so many different companies across the space to help with the programmatic space. You know, we still have some issues that we're trying to resolve and I don't want the shiny lights to distract people. And so I want to hear from you, what are those things that we should still be focusing on outside of Agentic and what does that mean?

Anthony Katzor (7:25)

Yeah, I would agree. I do think the hype cycle is at a fevered pitch at the moment. Again, my earlier statement of talking to holding companies and publishers and ad tech and everyone in between about, oh, I want to do an agentic transaction, I want to do an agentic X. I'm like, okay, can you tell me what that is? And I don't, I've yet to get real clear answers. Now, again, that's not to say that there have not been initiatives going as far back as almost two years ago. Talking to some of the holding company execs and their technology teams and talking to others. There have been agentic approaches or AI solutions either internally or some external facing solutions that are being used to drive greater efficiencies. But those aren't necessarily even built on mcp. I mean, MCP is a very young protocol. It's about a year old. So talking to folks, I'm hearing, you know, they have agentic initiatives internally, but some of that is built on, you know, a lot of that is built on proprietary code. It's built on, you know, it's, or it's built on legacy languages.

Podcast Host (8:27)

Right.

Anthony Katzor (8:27)

They've built an AI agent in C or they've built it in Python or like. So you're, I'm not hearing, I'm not hearing anything about, you know, oh, I built this in at CP or MCP or Agent Agent, just because all those things are so new. Right, right. So I, I do think there is a ton of smoke and not a lot of fire. I think we eventually get there. But it does also feel like there's a lot of. This is definitely a shiny penny moment for the industry. And as an industry we love shiny pennies and that doesn't mean we shouldn't innovate. But you know, there are still a lot of fundamental things that we have to address as an industry. You know, greater supply chain transparency and no protocol. I keep hearing, you know, oh, Agentic is going to solve supply chain transparency Protocols don't solve for misaligned incentives and bad actors. They don't like that's not what protocols address. Right. So I think there's a lot of like hype around what agentic is going to do to the ecosystem while we're still not tackling some of the fundamentals like supply chain transparency, privacy. And I know the industry has a tremendous amount of privacy fatigue and I get it. The rate at which laws have been passed, the rate at which laws have evolved, the complete fragmentation or privacy within the United States is a real challenge to deal with. You now have this new omnibus legislation in the EU perhaps to relax some of the privacy laws, particularly when it comes to AI. But privacy is not going away. The regulators are not, you know, they're not done fining, they're not done, you know, subpoenaing companies like they're still doing it. If anything, I think you're going to see them step that up in 2026. So that's still something we need to keep an eye eye on that ball. One of my concerns with agents is what happens when an agent for company X interacting with an agent for company Y violates a privacy law. Like what? Like what are the. We have to put guardrails around that. And that's some of the interesting stuff we're looking to do with UCP, which LiveRamp donated to us, which is effectively kind of a audience and signaling trading protocol in its simplest form. I'm sure the more technical listeners are like, that's not it at all. But for non technical listeners, I'm taking 100,000 view of what UCP is. We do need to build privacy guardrails around that. We do need to make sure tcf, the Global Privacy Protocol, things like the data deletion request framework are built in to these agents so we don't step on our own feet when it comes to privacy measurement. Still not solved as an ecosystem. And again, I don't know if there's any protocol that will solve the measurement challenges. Again, I think those are, I think measurement challenges really around business incentives. Not necessarily. They can be solved technically. If we would all just agree on a common set of telemetry to create a baseline of a measurement framework for the industry, we could probably solve the measurement challenge. A protocol is not going to do that. So there's just still a lot of foundational work we need to do as an ecosystem before we jump to this new phase of Agentic. And by the way, agentic is not the end all be all. There are still going to be. You know, there's been talk of like, oh, agents are going to replace rtb. Not anytime soon. RTB is incredibly efficient when used properly. The ability to inspect every single media impression and the audience attached to that and determine a fair market value of that between hundreds of buyers and sellers, you know, billions of transactions happening per second. That is an incredibly powerful protocol. So I think RTB isn't going anywhere anytime soon. I think agentic interfaces as RTB become really interesting, but again, you've got to build guardrails around that as well. So some of my concern with agentic, again there is a place for it, I think we can recognize efficiencies in the ecosystem is let's tone down the hype a little bit and let's I think be a little more practical around what we think agents can solve. I also see a lot of agentic solutions kind of pointed at direct media buying and selling. Sure, that's a solve, but when I think of agents, I think agents are really good at high cardinality problems. So things like discrepancy, resolution, why won't my creative work on these three CTV partners? Agents can figure those things out really fast. Where it may take a human hours, days, sometimes weeks to reconcile this like. So are there tools that the industry could also develop to help streamline those parts of digital media trading? Again, not to say that there's not value in direct media buying and selling or audience discovery or audience matching. There is. But I also think there's also some more practical low hanging fruit solutions that agents can bring to the ecosystem.

Podcast Host (12:58)

You know, you mentioned this a couple times, not just now, but even earlier about the Agendic interface. What would that look like in an ideal world? What do you mean by an agentic interface and how would that change the way we operate today?

Anthony Katzor (13:12)

I think, I mean if you start with, I think a simple example of what an agentic interface would be like is think of audience and how audiences are defined, how audiences are discovered, how audiences are matched. Like imagine going from today where you've got all of these predefined audiences, new homeowners, new parents, auto intenders. I think the natural language capabilities of an LLM powering an agentic audience platform probably allows buyers to suss out maybe more refined audiences. It probably allows sellers to curate, there's the C word for everybody to curate, maybe more nuanced audiences for buyers to buy. So I think it's, it's going to be less, I think you're going to see more prompts versus knobs, dials and sliders, the dobs, the knobs, dials and sliders. Aren't going away. I think you're going to see a prompt that will complement those similar to like self driving cars, right. When you get a self driving car, a Tesla, it will. Your hand is still on the wheel occasionally, right? Your hand is still there in the wheel. I think that's going to be the same for the interfaces.

Podcast Host (14:23)

You know, one of the things I talked about two weeks ago in my 2026 predictions, and Terry said that I was too early and I needed to stop, but I'm not listening to him anyway, was the impact of. Yeah, I mean, yeah, you've got your own thing too, Terry. I got my own thing. So one of the things I talked about was the impact of agentic on ad operations and that it would probably impact that department more than any other department in the space. Mostly because, you know, knowing how ad operations work, a lot of it is repetitive tasks, but they're manual repetitive tasks. And those repetitive tasks can be automated in a way. They can be duplicated quickly and you can pivot and change based off of whatever optimizations you need to do. And that it would impact the adopts roles the greatest moving into 2026, but not in a downsizing manner, but rather in a slowdown of hiring manner. Because now instead of allocating two hours every morning to optimize and tweak, you could probably click a button or just confirm that those changes have been applied and that it's performing correctly. Where do you see outside of kind of the UI aspect, where do you see ad operations changing as it relates to agentic? And where do you see maybe other departments utilizing agentic to improve their output?

Anthony Katzor (15:36)

I think that's accurate. I think the, if you look at what's happened from an engineering perspective in some of the larger engineering shops, some of the big tech companies, there's been a tremendous slowdown in hiring of new engineers. There's been a tremendous slowdown in hiring of new data scientists, particularly at a, at a junior level.

Podcast Host (15:52)

Right.

Anthony Katzor (15:52)

So you know how we all came up in our various disciplines in our career, whether it was business development, engineering, product management, you name it, right. You got to start at the bottom, right? And I think what we've seen in the past is that historically every, almost every professional vertical, whether it's digital media, programmatic, whether it's healthcare, whether it's finance, you know, pick a profession, there's always been this kind of pyramid model where you've got a, you've got a wide swath of junior talent that you know, over time, through varying degrees you know, we all go through a lifelong apprenticeship in our professions, move up the top of that pyramid. I think that pyramid is going to become a diamond where you're going to have less opportunities at a junior level. And we're seeing this, you know, we know this for a fact that at big tech companies they're not hiring as many junior engineers. I'm already hearing from some publishers, they've closed a lot of open roles for junior ad operations people because AI or agentic solutions allow you to, it's a tool, it allows you to do more with less. So you don't necessarily have to hire as many ad ops folks in the future. You don't have to hire as many data sci junior data scientists or junior engineers. You know, AI can write your unit tests for you. You know, one of my first jobs as a junior engineer was writing unit tests for the chief architect on the systems we were working on. Like AI can do that really well now. Like Claude is a pretty damn good coding platform or vibe coding platform. Same I think is true for adopts, you know, the junior adopts task, you know, reconciling. We just talked about reconciling discrepancies between campaigns. Like that's typically a junior, like here, go find the needle in the haystack. Like now it's like the agent can.

Ari (17:31)

Go do that again.

Anthony Katzor (17:31)

They're superhumans. I think of agents as superhumans that can take normal human tasks that take hours, days or weeks. They can do it in seconds, minutes or hours. So I think it's going to change the professional paradigm of you. I think we go from a pyramid shape to more of a diamond shape. Less at the bottom, I think more in the middle. I think you're still going to need that good mid level management, but you're gonna, I think you're gonna have to be more hands on than you've been in the past because AI is going to enable, you know, better task management, better project management. AI will be able to track status of projects a lot better than like having a program management. So I think those are going to be the impacts in adopts. I think you're going to see slowdown of hiring. I mean we've, I'm already hearing about it from both agencies and publishers and I just think you'll, they'll, there'll be just less new talent kind of going into the funnel.

Podcast Host (18:21)

Yeah, I spoke to a friend of mine who, who runs operations at a pretty big broadcast company and this topic came up, which is why I posted about it mostly the conversation we had related to, look, you have a fantastic team that does a lot of the, I hate to say it, like nitty gritty work, kind of like the hard work that's at the bottom optimizations. And I told him I think his role will become significantly more important as it relates to partnerships, partnership management, how to extract as much revenue as possible from those partners they work with. And probably a lot less hands on keyboard, but a lot more hands on telephone and relationship building still a very important part of the job, but that his role in particular will probably change to that and he should start building stronger personal relationships with everybody he works with to try to extract and get as much revenue as he can. And he pretty much agreed.

Anthony Katzor (19:13)

Yeah, I totally agree. I think, I think, I think business development, sales, account management, technical account management, I think those are sound roles now and well into the future. That's. You're not going to automate that. Those are people talking to each other, figuring out how do we work better together, how do I drive more revenue? Right. How do I create win wins solutions between publishers and buyers? Like, I think those roles aren't going anywhere. I do think, I do think there's an opportunity for some of those adopt folks to evolve into those roles. Like there's something to be said to know how the engine room works in order to influence it, using an agent properly to drive certain outcomes. So I do think there's going to be some evolution of some of those ad operation roles for sure.

Podcast Host (19:55)

2025 was the year of monopoly. We heard a lot about Google. We had two trials that we had to listen to constantly about. Some made sense to us, some didn't. One of the big challenges with AI is AI browsers and how that could impact the industry in particular. Like is it going to create that power imbalance that we've seen before in the past with Chrome or what publishers should be doing to prepare themselves. What are your thoughts on, I guess, AI browsers and how that could impact the overall advertising industry?

Anthony Katzor (20:28)

I think I have two fundamental thoughts on the AI powered browser. So if you take, take ChatGPT's Atlas One, I think it's just a Trojan horse to continue crawling. But now you're doing it from the browser and from a residential IP address. Right. So I think that's number one, I think number two, I think the wider change with the AI browsers is going to come down to web design and information architecture. What is incredibly powerful. You know, I don't know if you've used Atlas, but what is Incredibly powerful is that you and I could be looking at the same content and Atlas could, could lay it out differently for each of us. So it's, it's, it's almost like personalized curation of content. The content doesn't change, it's just the way it's presented to each of us could be slightly or radically different. I think that that has really interesting implications from a web design perspective. It also, I think, has interesting implications from an advertising perspective. You know, I think page layout, of course, is important to, you know, where do your ads go? How are they presented? You know, how does the consumer experience and see those ads? So I think, you know, we may be heading towards a world where web servers are just publishing MD files and the AI browsers are laying out that content based on how each individual wants to actually see the content represented in their browser. So I think, I think it's going to be a really interesting time to see how information is architected to the consumer.

Podcast Host (22:00)

How do you decide on any of this stuff? Like when you, when you look at all the initiatives that you have that you bring up with your partners, when you bring up with your coworkers and your team, how do you decide what's worth really digging into and pushing forward? And how do you decide, okay, this is still hype. I'm going to put this on the side for now and I'll look at it, you know, next year. How do you make that call?

Anthony Katzor (22:22)

Oh, gosh, we, it's not easy. We have our, we start, we start strategic planning for the next year in June, the prior. So we started 2026 strategic planning in early June. You know, and so a lot of that is we talk to our members, you know, we talk to our working groups, the team convenes, we, we debate amongst one another as to trends that we see shaping the industry. And we also debate, you know, trends that are, you know, no longer a trend. You know, things that we think are in a downturn or things that we also think may actually be, you know, satisfied enough where we don't have to make a ton of additional product investment at this time. And then we run those past, you know, we run those past our board, you know, both in my one on ones. I try me with every board member at least twice a year, one on one, sometimes three times a year. So either through formal board one on ones, and then we also have what we call a tech lab board summit or our entire board convenes. It's not a formal board meeting. So board member, company employees, can come too. So like let's say you are, let's say you're a magnite. They're on the tech lab board. The board seat comes to the meeting. But they can also bring like a pro, a product lead and an engineer or two. And we all engage in a very strategic conversation of where we see the industry going. And that is typically what results in our final, our final roadmap. I mean some of it frankly is, some of it's a little thumb in the air. Like I think this is where the ecosystem's going. So there's a little bit of art to it. You know, we try and be as analytical as possible. What we've done this year, more than we've done in years past, is we've been reaching out to more of the end principles of the transaction. And but what I mean by that are the holding companies and the publishers, right? Ad tech is a facilitator of how buyers and sellers want to transact. So we've done a good job this year of really getting more whole co feedback and more direct publisher feedback than I have say we've done in years prior. Because I think sometimes, sometimes when we roll out our specifications, where the rubber hits the road is an adoption. And that is occasionally due to the fact that we did not get principal buy in. So we didn't get publisher buying, we didn't get holding company.

Podcast Host (24:31)

When we talk about the AI, we also hear air bubble that comes up a lot. Do you think it's going to pop? Do you think it's going to deflate? I know I read some commentary that you said it was just going to deflate. Where do you think that hype cycle is going to level off a little?

Anthony Katzor (24:45)

Look, I think someone wrote a really good article recently. I'm trying to remember who wrote it because I chatted with him after. I'm blanking on his name at the moment. But he said likely your first agentic AI project will fail. And again, that's just part of innovation. You know, Thomas, he didn't invent the light bulb. What he did do is he perfected the filament within the light bulb for the light bulb to stay on longer. And I think there's an analogy here where there's early movers and I think we'll see mixed success in agentic approaches to the ecosystem. But eventually something, some things will start to take hold. Whether that is at cp, whether that is, we're going to be rolling out our ream of agentic initiatives starting in January. We're going to be very aggressive in rolling those out or whether that's something else, who knows. So I think we're going to see subsequent failures and successes, I think before we stabilize around an approach. And I don't know if there's going to be one approach. That's what's really interesting about Agentic. It can take a number of different approaches. It doesn't have to be relegated to a single protocol. It could be multiple protocols. So I think it's going to be an interesting experiment. I think you will see as we go through that, that this period of innovation, which I think is a two to three year period, we are going to have our, you know, our, our, what was it? The, the Gartner, you know, peaks and troughs of the Gartner wave. Right. So I think right now we are at the peak of optimism. We will hit some drops of despair for sure. But I don't think, I don't think Agentic is going away. So that's when I say I don't think it's going to, to necessarily pop. I think we'll see a deflation. Now if you think of the wider AI ecosystem, I think how that all lands out is going to be due more to financial management and investment between a lot of the big players, whether it's OpenAI and Nvidia and Anthropic and Perplexity. I think you may be, you may see some popping bubbles there depending on how those investments play out in the success of those companies.

Podcast Host (26:52)

Tony, thank you. And tomorrow is a new year, so thank you for joining me on the podcast and I appreciate your time and all the insights that you give me every time you're on my podcast. So thank you again and Happy Holidays and Happy New Year.

Anthony Katzor (27:06)

I try have a happy holiday and a happy New Year. I'll see you in 2026.

Podcast Host (27:09)

Definitely speak soon.

AT Tech God (27:12)

Thanks for tuning in to another episode of the AdTech Godpod, a podcast for the people, about the people people. Stay connected with me for more insights, trends and interviews in the realm of ad tech. Don't miss out on the latest updates. So follow me on X Instagram and connect with me on LinkedIn. Don't forget ATG Slack community has insights, networking opportunities and jobs. Keep the conversation going and stay at the forefront of ad tech innovation.