When Publishers Get AI Agents: Andrew Mole on Agentic Trading and the Future of Media

Watch on Youtube

Listen on Spotify

Programmatic advertising automated the transaction. AI may automate the negotiation.

In this episode of Signal & Noise, Brett House and Rio Longacre sit down with Andrew Mole, CEO and Co-Founder of PubX, for a provocative conversation about agentic trading, publisher monetization, and what happens when AI agents begin representing both sides of the advertising market.

Andrew has spent much of his career watching programmatic evolve from a breakthrough in efficiency into an extraordinarily complex ecosystem of DSPs, SSPs, exchanges, data providers, verification platforms, and intermediaries. His argument is simple: much of this infrastructure exists because humans needed interfaces and platforms to manage complexity. AI agents don’t necessarily have the same limitation.

That raises a bigger question: If we designed digital advertising from scratch today, would we build the programmatic ecosystem the same way?

Andrew’s answer is essentially no.

PubX is already experimenting with agent-bought and agent-sold media, with Andrew revealing that the company is currently transacting several thousand dollars per day through these emerging workflows. The volumes are still small, but agentic trading is beginning to move beyond demos and PowerPoints into actual transactions.

The conversation explores two possible futures. In one, AI agents operate on top of today’s DSPs, SSPs, ad servers, and other infrastructure. In the more radical version, buyer and seller agents communicate directly—potentially eliminating significant portions of the traditional programmatic supply chain.

The result could be a shift from automated auctions toward autonomous negotiation, where agents negotiate not only price but audiences, first-party data, measurement, inventory quality, outcomes, and other commercial terms.

For publishers, the implications could be enormous. Instead of simply accepting market prices, intelligent sell-side agents could continuously represent the value of a publisher’s inventory, audiences, data, and commercial interests at a scale no human sales organization could match.

Andrew explains how PubX is approaching this opportunity across more than 3,000 publisher sites and why publisher first-party data becomes dramatically more valuable when machines can discover and activate it at scale. The conversation also gets deep into the plumbing: OpenRTB, Prebid, ad servers, buyer agents, seller agents, and whether the industry actually needs today’s DSP and SSP architecture in an agentic future.

That leads to one of the episode’s biggest questions: Who wins and who loses?

Andrew argues publishers and advertisers have powerful economic incentives to embrace agentic trading, while intermediaries whose primary role is passing transactions through the supply chain could face serious pressure. DSPs, SSPs, and agencies won’t necessarily disappear—but their roles may need to change dramatically.

The discussion also explores the future of agencies, verification and brand safety, publisher sales organizations, and whether AI could actually strengthen the premium open web by allowing publishers to capture more of the value they create.

And this may not be a distant future. Andrew expects meaningful adoption of agentic trading to begin in 2027, with a potentially significant share of media transactions shifting in this direction over the following few years.

In this episode, we explore: agentic trading vs. traditional programmatic; buyer and seller AI agents; publisher first-party data; the future of DSPs and SSPs; OpenRTB and Prebid; agency disruption; autonomous negotiation; publisher monetization; brand safety and verification; and the future of the open web.

The big idea: Programmatic automated execution. Agentic trading could automate judgment. And once buyers and sellers are represented by intelligent agents capable of negotiating directly, the architecture—and economics—of digital advertising could look very different.

#SignalAndNoise #AgenticTrading #AgenticAI #AIAgents #AdTech #ProgrammaticAdvertising #DigitalAdvertising #PublisherMonetization #Publishers #OpenWeb #MediaBuying #FirstPartyData #PubX #FutureOfMedia #FutureOfAdvertising

Read the full transcript below.

---

Andrew pubX (00:00.512)

Yeah.

Brett (00:01.65)

Hey, this is Brett House, Signal and Noise. Welcome back, everybody. I've joined with my co-host Rio Longanker. And today's guest, Andrew Mole. I didn't ask if it was Mole, but I think it's Mole. there's no accent on the E. Am I right there? This the CEO and co founder of PubEx. It's a Friday episode. I've gotta be a little goofy, can't help it.

Rio Longacre (00:08.124)

Everyone?

Andrew pubX (00:17.256)

You are, yeah, your intuition is correct. The mole is fine as well.

Brett (00:28.464)

So yeah, so you're the CEO and co-founder of PubEx. You founded it with Nick Finneck and Lou Carpino, who we've not met. and it's a company that's helping publishers maximize the value of their inventor inventory through a modular AI stack. I took your corrections via email in advance of the show. Don't call it an agentic marketplace. And we'll get into what that means going forward for the audience's purposes.

So you've been building PubEx for what more than six years, right? Six and a half years?

Andrew pubX (00:59.534)

Yeah, but yeah 2020 we founded it. So yeah, I should also I should clarify Alex Rose and our CTO is also a co-founder lest he resign as a result

Brett (01:02.11)

Twenty twenty? Yeah.

Brett (01:06.202)

and and and and Alex Josen who we were talking about before, who's a diehard mountain biker and skier. apparently you guys are having quite an event

Andrew pubX (01:11.857)

the very same.

Rio Longacre (01:12.606)

Who we we spent time with again and we have a a a picture of him, I think after a couple of drinks, holding up a bottle of wine.

Brett (01:18.918)

Yeah, yeah. I you Rio's like, Leave the wine behind, you're not gonna drink that. And it was like a full bottle of like expensive white wine and I'm like, sure. It sat in our refrigerator for the rest of Can and nobody drank it. So so you're from London town, I think originally, you don't live there now, but I had to say this is one of our our favorites. We talked about this in one of our our past episodes. It's our our Rio and I are big fans of cast conditioned hand pulled bitter ale in you know, classic style.

Andrew pubX (01:19.79)

It's duh.

Andrew pubX (01:29.614)

Hahaha

Andrew pubX (01:33.697)

Exactly.

Andrew pubX (01:43.192)

WAH-WAH!

Yeah, it's served tepid room temperature. It's slightly quite... Yeah.

Rio Longacre (01:46.556)

Huge fans, yeah.

Brett (01:49.351)

temp and room temperature flat, right? I mean just, you know, I wanna I wanna be there in a dirty in a dirty old town, in a dirty old pub, and

Rio Longacre (01:54.6)

Bitters are great, I'm a huge fan.

Rio Longacre (01:59.645)

In a pub that's older than this country. And I think the the the the walls behind you, you were telling us before, are actually older than the United States of America.

Andrew pubX (02:06.872)

They actually are, and congratulations on 250 years, but yes, the barn is older, so.

Brett (02:07.059)

Yeah.

Brett (02:11.686)

Yeah, that so my my joke for Friday is the difference between yogurt and America. Yogurt has culture. so anyways. that's that I heard that from a Welshman a while back, but but yeah, lo we both love London as well. you know, I've been there multiple times. you know, and and a lot of a lot of what I was influenced by certainly came from that town. Yeah. Yeah.

Andrew pubX (02:18.478)

You

Rio Longacre (02:18.984)

Kids have culture, come on house.

Rio Longacre (02:32.51)

Food's gotten better actually the last twenty five, thirty years. So big improvement.

Andrew pubX (02:35.658)

enormously. Yeah, yeah, it really is.

Brett (02:36.868)

Enormously. Yeah. And and the joke there is the reason why it's gotten better is 'cause the English aren't cooking it.

Rio Longacre (02:42.728)

Okay.

Andrew pubX (02:42.766)

Yeah, Well, unfortunately, we are again now following Brexit, so it'll probably start to tail off, but there was this Halcyon period in the middle. Yeah. Yeah, Curries have remained excellent, yeah.

Brett (02:48.568)

It's like let y bring in the immigrant populations and let them do the cooking. Yeah, I've had I've had yeah, ch curries, I mean I had a I had a Jamaican guy as a cab driver in London that told me his entire recipe during like a twenty minute drive of how to make the perfect home curry. And so very interesting. And I've had really good like Taiwanese food, all sorts of different genres within London.

Rio Longacre (02:52.18)

Well the curries are very good, so that yeah.

Rio Longacre (03:13.022)

Great Persian food, yeah, the good but still the fish and chips. Like I'm love fish and chips. So and that it does n very few things are more English than that, right?

Brett (03:18.589)

Yeah.

Andrew pubX (03:18.738)

Classic fish and chips. Yeah, and we stole it from the Portuguese, so that's how English it is.

Brett (03:21.971)

Yeah.

Yeah. And and very few things more English than football. but what the hell happened with you guys in the World Cup? I meant my God. you did pretty well, but you did the best that you've done in sixty years, but I meant come on, Ted Tim Tuchel. you know that that I just knew you guys were gonna lose that game. The the the beginning in the sixty sixtieth minute mark.

Rio Longacre (03:32.436)

Well it's a tradition now, right?

Andrew pubX (03:34.069)

Yeah, yeah.

Rio Longacre (03:46.942)

Well Brett, I think everyone everyone in in the country did too, unfortunately.

Brett (03:49.671)

Yeah, the sixtieth minute mark with suddenly they had ten defensemen, right? And he subbed in a couple of fullbacks. I was like, no, they're gonna have to take this onslaught for the next like thirty, forty minutes of this game. Of this match.

Andrew pubX (03:54.85)

Yeah.

Andrew pubX (04:04.29)

Well, that's it. It's preempting that last minute, that famous last minute break, but then not being able to do anything about it if it happens. So this is, yeah, it will go down as a tactical misdecision, I think. But still a great performance. So, you know, it's iterative.

Brett (04:16.252)

Yeah. Yeah, and who do you blame? Do you blame the old the old guard and the old model of English football, or do you blame Tuchel and his coaching tactics? There's kind of a big debate around that. But on to the topic of the show. so you spent years in the intersection of programmatic advertising, publisher revenue, machine learning at companies like Platform Three Sixty, which is a a native ad tech company for brand amplification, Pub Galaxy, a publisher monetization platform, and now you're

kind of pioneering a future where autonomous AI agents just don't optimize advertis advertising. They actually negotiate on behalf of publishers, which we're calling agentic trading in this space, evaluating inventory, packaging first party data, creating more intelligent systems for really open web advertising. And one of my favorite quotes about you when it comes to AI and some of the thought leadership that you've had around that was you said this isn't the AI age and it's not the data age, it's the agento scene.

I thought that was the coolest quote I've ever read. I was like, it's like, you know, so so t tell you know, welcome to the show. great to have you and would love to hear all your perspectives about all these things I just mentioned. And did I did I miss anything about your background that is worth talking about?

Rio Longacre (05:15.025)

That's cool. Yeah.

Andrew pubX (05:20.494)

Thank you.

Andrew pubX (05:28.958)

No, no, think that's it. Thank you so much for having me. I'm really excited to be here. I think my background is neatly sort of mapped over the dawn and then the evolution of programmatic. So it's been able to see for sure the major benefits of this, but then also maybe the drawbacks. And I think the drawbacks have started to compound at a time when...

technology in the macro ecosystem, if we for one second look outside of our wonderful industry, has maybe given the opportunity to fix some of those drawbacks. So my career has been very luckily timed to have enough of a foot in that first camp and hopefully enough of a foot in our second camp to really make a difference. But no, it's a great summary. So yeah, I think that's exactly right.

Brett (06:09.041)

Yep.

Rio Longacre (06:14.685)

Well Andrew, we're thrilled to have you on here and this conversation, you know, we w we wanted to have it. I mean, I think initially we s ran into you guys through Rachel Klausner. Shout out to her for I think beginning this, but we wanted to have you guys on here because we've talked a lot about publishers and the problems publishers have been having with monetization, which have become more acute with the

Drop in referral traffic. And we love the fact that your companies really had this focus on that, which we think is very important. But we wanted it to be even like a broader conversation than that. You think about the last 20-so years, you mentioned you've been involved in programmatic advertising from the beginning. It's automated everything, right? Through auctions, exchanges, DSPs, SSPs, you know, alphabet soup of different acronyms, right? And while we've automated this execution,

Brett (06:58.803)

Yeah.

Rio Longacre (07:02.897)

I don't know, I I know I'm not the only one who's who thinks this. To a certain extent, I think judgment behind media buying and selling has in theory remains human, but it's I think there's it's it's there's been a breakage between the planning and the actual execution, which is really a lot a lot of the actual like not simple, like where the rubber meets the road is actually decided by an auction, right? So our thesis here is that AI is potentially going to change a lot of this, right? And we've been talking a ton about.

quote unquote agentic trading, which I don't really think is a thing yet, but we are believers that media some media will be bought and sold by with buy buy side agents and sell side agents negotiating and and making deals that way. So I we it's a ziffitsy, but but for s

Brett (07:43.038)

Well it's a thing, but in its infancy, right? Right. We had an episode with Patrick O'Leary from Booster and their Vox Media partnership which kind of announced that that they were doing something in real you know, in the real world. Yeah.

Andrew pubX (07:45.622)

Mmm, yeah.

Rio Longacre (07:55.668)

Well they ran a campaign. Yeah, they actually bought and sold media and transacted it that way. Now, will it just be reservable media that goes that way? Will it just be direct sold deals? will eat into RT will will open RTB actually be impacted? Will that be will that be more remnant? W no one really knows, but a lot of things are happening, and we know you're kind of at the the center of this with some of the if we're in the agento scene epic r era right now,

We know you're experimenting with some of those things right now, so we're really excited to talk to you about this and how you see that evolving over time. So I guess that was one of the the main themes you wanted to jump into. But like kicking off the questions here, like, do you believe we're entering a world where AI is fundamentally changing how media is planned, bought, sold, optimized, et cetera? D do you feel that?

Andrew pubX (08:41.229)

I completely do. Yeah. We are seeing this firsthand and you've referenced booster. There's a number of other people in the space. We are transacting media. I should have got brand clearance before coming on this and I could have said it, but we're doing a few thousand dollars a day, which is agent bought and agent sold. And in that we're the counterparty, we're the selling agent. this isn't, people are not going to retire on this, but it's real money, right? And I think one of the problems with the hype is everyone's doing, every time someone puts

Rio Longacre (08:59.667)

Okay. Very cool.

Brett (09:04.327)

Yeah.

Andrew pubX (09:09.901)

$50 into one of these things they put out a press release before they've even run it and this creates the impression that it's happening really quickly and early Which is amazing until the press releases say the same thing for six months because they haven't actually done the underlying thing So we've taken a view of kind of trying to build it more slowly Responsibly and actually deliver good technology because as you see in every change that happens It's publishers that end up suffering so programmatic came out to help publishers, but it really hurt them it

Brett (09:16.912)

At scale. Yeah.

Brett (09:35.069)

Yeah.

Rio Longacre (09:36.571)

That's the rule of the internet, right? Is that publis publishers automatically get screwed, every changed.

Brett (09:39.408)

It's like how how do we how do we continue? Yeah, yeah, the middlemen are gonna be removed, but then who's gonna be grabbing that share? And oftentimes it it it works against the publishers. Yeah, a hundred percent.

Andrew pubX (09:48.438)

Yeah, completely. So we're trying to do this in a way that works with the publishers and for all market participants. shouldn't punish someone against the other because ultimately that would be an efficient market that we can strive for. But to the first question, Rio, I think the agenda scene is beyond advertising and that there's no way you could look at, and we're all in a bubble, I completely get it. This is this privileged bubble. But if you look at usage of Claude, OpenAI, the number of people subscribing to these platforms.

and using them every day, the world is becoming agentic, right? Like use it to help prepare the podcast, to do these routine things that enable you to operate more at scale and to be more performant in whatever the thing you're doing is. And I don't think our industry, which is wonderful, massive, scaled, brilliant, but quite inefficient or can be quite inefficient. I think it's ripe for disruption through agents.

Brett (10:21.246)

Yeah.

Brett (10:28.124)

Yeah.

Brett (10:42.054)

Yeah, and and that's we're certainly hearing a lot of that. And you've argued that the programmatic stack has only really existed because humans needed interfaces between systems, right? They needed UIs. There's a lot of talk about headless and the death of cesspo apocalypse and all that sort of stuff. And and agents kind of remove the reason for these these interfaces, these UIs to exist, right? And so we're hearing a lot of that. Can you can you s what are your thoughts on that?

Andrew pubX (11:06.497)

I mean, I think that SaaS, and you know, we're a venture backed business, a lot of our investors are in a of SaaS companies as well. But I think that SaaS exists because humans can't hold everything they need to in their head at the same time and execute it simultaneously. So it's a scale issue, right? We have a CRM that we pay for because we can't remember the 1000 customers we have that we want to talk to on a daily basis. So they're there because of that. But an AI can remember them and it can...

in a bespoke way for each of them because it isn't tired, it doesn't have the human limitations. And you migrate that into programmatic. And if you imagine like a programmatic trader who can only manage 20 deals on a campaign at any time because they're looking at performance, they're going to run a report, they're comparing it to the next one, they're remembering what the last one said, they're exporting it to Excel, they're putting it into a pivot table. All of this takes time. So they couldn't have a thousand deals running simultaneously, but an agent can.

Brett (11:54.43)

Hmm.

Brett (12:01.296)

Yeah. Takes time and is ripe ripe with potential human error.

Andrew pubX (12:04.959)

It's copy paste. It's all of the things that go wrong. It's, I took the wrong deal out of the email chain. So I put the wrong deal on the wrong campaign. Or one of the first things I did in media, and I mean like first month working in media, was spend $5,000 on a campaign for Ford in Brazil that was meant to be targeting Brazilians in North America. And this was just misread an I.O. 20 years ago, ran it in the wrong country, really pleased with the results. Client was devastated with the results.

Brett (12:28.158)

How how how did you report that? You were pleased with the results.

Rio Longacre (12:29.01)

Crushed it in Brazil, right? Yeah.

Andrew pubX (12:34.667)

Yeah, and I was...

Yeah, exactly. Yeah, they all bought a Ford, though. But then I was like, my God, I'm going to lose my job. And I just went to my boss and I was like, hey, look, this is what I've done. And they were like, things happen. And it's like, things do happen. People do make mistakes. But they don't have to. And the agents can do this at scale. As long as there are guardrails and governance in place to make them work, they can do this at an almost infinite scale. And this is where we start to see value accrete, because it's not just one of the best 20 deals for this campaign that a human thinks.

Rio Longacre (12:50.45)

Shoulder shrug, I love that. Yeah.

Andrew pubX (13:06.869)

It's an agent testing those deals every hour, minute, second to make sure they are the best. Rolling new ones in, moving spend to the ones that most performant today, that hour, over the weekend, whatever it is, and then optimizing in real time. So we're not there yet.

Brett (13:20.38)

Yeah, that capa that capacity for test and learn, which, you know, is a data science, data analytics play of like, you know, you're running campaign, you do these smalls, is just exponentially greater than any human team could possibly do, which probably increases your chances that you're gonna get it right and your optimization work streams are gonna be faster, more efficient, right? As long as there's guardrails, that's a that's a that test and learn ability is incredible.

Rio Longacre (13:45.565)

Well, Andrew, let me ask you a question though, really about about agentic and how you're defining it. 'Cause I I mean I can pause I mean, looking at let's say digital media buying and I'm you know, outside of Walled Gardens, right? 'Cause Wall Gardens are gonna have their own version of like agentification and what that looks like and you know, meta's gonna look different for Google, but they're gonna be to, I think, do some very interesting things. But let's say open web, p paid media. I think they're it

Andrew pubX (13:45.775)

Exactly.

Rio Longacre (14:08.4)

I've heard two definitions of a agentic, right? One would be we're going to build agenc workflows on top of existing systems. We're gonna probably get rid of a lot of the time spent in UI by people and maybe get rid of a lot of things people are doing and use these agentic systems that are gonna hop like these agents will hop across s existing tools, existing DSPs, SSPs, exchanges, what have you.

And automate a lot of the things. That's one that's one version. And I've seen and and maybe ad tech gets a little rerouted in some of that. I've seen some like publishers going right to like, let's say, through like a PubMatic, right? Which is is more on supply side and and and maybe cutting out a DSP. For I mean you get a little bit of savings, right? In terms of like some of the know the the money that comes out of working media by the

But by ad tech, but then there's a more radical version. And this would be you actually have buy site agents and sell site agents. Let's say the sell site agent has access to the OMS, right? We talked about this booster with with Patrick. Brett, remember this? But then you have to have some kind of connection to GAM in order to actually like get the ads served and get the orders in, right? And then you have to have

Brett (15:05.714)

Yep.

Rio Longacre (15:15.544)

some kind of communication and negotiation then to a to a demand side agent which would which would be representing the advertiser or the agency yes who would be ostensibly be representing the advertiser, right? So and that in theory is very radical because is completely rewiring everything, right? It's completely changing it and the amount of money in working media it gets impacted very dramatically and you know, what what happens to agencies, but I I mean it's w like

Thoughts on those two versions and what do you think's the future? Or is it both?

Andrew pubX (15:47.278)

Yeah, absolutely. And I think if you look at this from a protocol level, there's space for both. So I think RTB and RTP kind of as a comparison. My view, and the thought experiment I used to justify this rather than it just being an either explicit or at least oblique attack on people in the same space, which is it's just too self-serving. So, but the thought experiment is

Brett (16:12.083)

Yeah.

Andrew pubX (16:14.099)

If you had the level of sophistication in agentic technology that you have now 25 years ago, would the programmatic ecosystem have been invented in the way that it was? And I think the answer is no. And we built a DSP. Our previous business that we sold was a DSP. We really understand the buy side technology well. We were partners for all of the SSPs you've named and a bunch of others besides, right? So we've been in that ecosystem for a long time. And even when we were running it,

we saw that there are flaws in it in terms of, know, talking like 10, 12 years ago, ditching bids because we didn't have capacity, QPS limitations even then. Whereas I don't think that would have been created if you could say to an agent, hey, I've got this much inventory in real time and it's plugged into the buy side that have this much demand in real time and you can sell it that way agentically. And this is what I mean about that. If you imagine that trader, if that trade is an agent, you can have a a deal ID in inverted commas.

of one impression with one bit of data. And every time it appears, you want to get in front of that because it's exactly the right audience to book your test drive for Fiat or whatever it is. But you can run it because you've got the capacity to run a deal with one. Obviously, a human trader could never do that. So we are massive advocates of the skip the current ad tech infrastructure. Now, if I'm Michael or Rajiv, I completely understand why you're an advocate of build on top of it, because that's what you've got.

and that makes sense as well. And that's also not necessarily invalid because we're talking about moving a trillion dollars from where it's pointed currently to a new direction and this will be transitional. But rather than being controversial, I don't think we would have invented any of those platforms if we had agents 25 years ago.

Rio Longacre (17:55.761)

Yeah. Well well you think about it, eighty percent of of bids are duplicate. I mean like like this clearly like this is this Rube Rube Goldberg machine we call programmatic advertising. No one would have built it this way if we had planned it out, right? Eighty percent of bids, right? Of and that's why

Andrew pubX (18:05.046)

Yeah.

Brett (18:05.98)

Yeah. Eighty percent are duplicate. That's really interesting, yeah. And then and then the the the the the dearth of working

media dollars, right? What is it? Forty one cents of every dollar is being earned by the publisher and the rest is going to the to the middleman, right? Which which back to the original sort of kind of the the theme that we've been talking about is that publishers are getting shafted in this mechanism. And they're the source of all the content and and and information that we all consume. They're the reason why there is advertising. It's un it's unbelievable how they've gotten sort of disintermediated out of this

Rio Longacre (18:17.586)

It was the publishers, yeah.

Andrew pubX (18:17.72)

Yeah.

Andrew pubX (18:41.55)

100 % these are the originators of quality and audience and experience in the marketplace, but the other person who's getting shafted Fairly up till now no longer fairly is the advertiser because if they're putting a dollar in and they're getting 40 cents of Value out of it that is an incredible tax on productivity right from a purely economic if you think of MasterCard They take 40 basis points for the same transaction right the trade desk takes whatever percent

Brett (19:01.715)

Yeah.

Andrew pubX (19:08.974)

SSP takes whatever percent, intermediaries take the... So that's all been cut out. And that was great because a human can't interface with the bid stream. The trade desk was great. They built a great business. They had great people from the start. That was necessary. But our view is not that that was wrong, but that it's no longer necessary. And when you take that out...

Rio Longacre (19:09.81)

Twenty percent, yeah.

Brett (19:24.764)

Yeah. I love how you're referencing that the past tense. That's that's interesting.

Andrew pubX (19:28.162)

Well, that's either unintentional or I've self indoctrinate the Kool-Aid wiping from my lips. But yeah, but that's right. Like it's that's not predictive spend. So if you're an advertiser, you're not really spending a dollar to get in front of these people. You're spending 40 cents. And then you're looking at the results and going, I don't think I've got a dollar's value here. And that's because you haven't. So if you take those intermediaries out, what we see as the North Star for agentic is does the CFO commit more money to ad spend?

because they're actually seeing really great return, now it's actually going into ad spend rather than the platforms to facilitate it.

Brett (20:03.228)

Yeah. And and yeah, so we the describe to the audience like where public PubEx sits. And the way I understand it is you guys sit in the from an architecture and tech perspective, you sit in the in the header wrapper, is that right? Right?

Andrew pubX (20:14.68)

So we've had technology in the past that sat in the head of wrapper. that's, so that's how we've got this footprint of publishers. So the text on thousands of websites and historically we've been trusted to price open market auctions for a bunch of Exactly.

Brett (20:22.823)

Yeah.

Brett (20:26.524)

So so it collects logs and inventory data from all of those publishers, this publisher network. How how big was that publisher network or how big is it?

Andrew pubX (20:33.292)

So there's about just over 3,000 sites using it, premium and then scaled. So scaled monetization businesses and premium publishers.

Brett (20:36.518)

Got it.

Brett (20:42.084)

So it's focused on and then the sales the the agents are actually negotiating against that that that traffic that you use.

Andrew pubX (20:47.106)

So that has been the base of the product. So that was like our pricing product that was in market. And then what we did over the last year was over exactly over the top of that developed the agent technology. But we had this really privileged view of one being trusted to price inventory in the open market anyway in real time, which is quite unusual, two really great partnerships with our publishers who see us as a tech partner. So we've been able to sit down for the last year with these people and say,

Rio Longacre (20:50.801)

That's been your core product up to date.

Andrew pubX (21:12.92)

how can we co-design the future of this? How can you be part of this? What would it need to do for you to want to do it? And like, yes, you might be able to build your own sales agent, but what we want to do is to build the best sales agent because we've got, for example, cross-market insight. And that doesn't mean you shouldn't use your own as well. And if your sales team is selling into it, then more power to them. That's great. But we can say, well, we're on thousands of sites. We can give you insight into, you know, if you increase your price by 12%, you wouldn't actually see a reduction in demand because you're below the market.

or you're slightly over the market. So if you take a small haircut, you're going to see a huge volume increase and also give tooling around audience management and things like that. So we were able to turn that position into a publisher facing agentic proposition.

Rio Longacre (21:51.803)

So you

Rio Longacre (21:56.434)

But you you're really clear in your in the information you sent us that you're not creating a marketplace. That's correct?

Andrew pubX (22:02.797)

It is so what we have is that I'll clarify the nuance but we have an end-to-end solution in that we built technology for the sales side we built publisher technology for agentic and Then we built buy side technology for agentic because we wanted to bring spend to our publishers So rather than go to an agency who doesn't have a buyer agent and just go, okay Well, whatever we can give them our buyer agent, but we are we're a pub tech business. We want to flow money to our publishers

but we've built technology to enable that. But the reason we say...

Rio Longacre (22:33.627)

So you have the buy side agents and the sell side agents and then as well as that kind of control plane in the middle that where like the governance and identity and all those things get and many of the rules get interpreted.

Andrew pubX (22:41.517)

Exactly. And the governance sits on both sides. So you're encoding the intentions and the parameters on both sides. So any negotiation that happens. And right now, this is all very early, right? So this is not, you know, 17 layers of negotiation back and forth, like 12 angry men type stuff. This is kind of, well, what would you sell this for? Okay, that's within my threshold for buying. Okay, I'll buy it. But we, so it's end to end in that sense, but it's not a marketplace because we're not.

We're not inviting people to participate within that and giving them the underlying infrastructure. We've created the tooling that sits along the value chain. Whereas if you look at like Scope3, who are a partner of ours, we're a founding member of Agentic Advertising org alongside Brian and Kelly and team. I think what they've done is fantastic. They are creating a marketplace. They've got the shop window for sales agents, for buyer agents. They're the underlying financial technology, I think is where they're going with this.

Rio Longacre (23:34.319)

They want to run the rails for the actual transactions and manage the float for all that, right? And actually act as the market maker, right? Is my understanding.

Andrew pubX (23:38.637)

Exactly. I think that's right. Yeah. And I think that's a major opportunity. And it's also it's a gap in the industry due to the insane opayment cycles between the buy side and the sell side historically.

Rio Longacre (23:51.656)

yeah, between brands put in net ninety, net a hundred to actually pay anyone, platforms wanting to be paid net thirty, right? And then all of the the different payment terms. I mean, yeah, I mean that's a big part of what SSPs do today and I just manage a lot of that for the publishers. I can you know, so someone needs in if if we if we do rewire this, like someone needs to manage all that. So I guess scope three has basically said that's what they want to do. It is needed, I agree.

Andrew pubX (24:13.197)

I think so, and I think it's really valid, and I think it's important. One of the things that think Scope 3 and Brian in particular are doing are having been there and done this, seeing what were the major challenges that happened last time and trying to not build them into the underlying infrastructure this time, which is good. So when you look at AdCP, kind of, the releases that come out are very sort of iterative and very rapid in those iterations because it's fixing these core problems as they come up, rather than going, we'll deal with that someday.

And as we saw in programmatic, that meant literally we will never deal with this. This will be a problem for the rest of time. And I think they're doing a great job of highlighting what those things are and then preventing them from becoming core to the technology.

Brett (24:55.614)

Do you do you envision AI agents negotiating it it seems like you're sort of leaning down this path, in engag negotiating more than just price, right? Things like audience quality, first party data, outcomes, measurements.

Andrew pubX (25:06.367)

Yeah, 100%. Yeah, yeah, I mean, they will because of the scale. So if they can't do it, what we envisage is nothing defaulting back to a human minus approvals as required. So you might say, I don't care, go for it. You know, I trust my governance agent. Go and do what you want. I think more realistically for this transitional period, could be months, could be years, I think people will want to be in the loop from an approvals perspective. But

Rio Longacre (25:07.845)

have to, right?

Brett (25:19.144)

Yeah.

Andrew pubX (25:32.852)

And one of the things we do now is we surface the creative to the publisher so they can approve it rather than the buyer just transacting automatically. And that is that's an embedded human. That's someone checking that. But I do think that this will happen in the future. And some of the smartest work we're doing is around embeddings audiences in the vector space and saying, well, where can we find this audience elsewhere? What does it look like? What's the cosine similarity between these audiences? How could we use that to extend audience if the

the budget exceeds the realistic. To your point earlier about 80 % of duplicates. I don't think we know how big the actual internet market is right now, but if there's 80 % duplication, it's not as big as we think it is. So saying where is that real audience? Because we're taking forecasting data from real ad servers. We're not just saying, I've got as much as your budget is. So we're coming back with real numbers here. But all of that really smart stuff can be done over the top of that. And then agents can go, okay, well look, we've got more budget than you want.

and the decisioning agent can go, yeah, but I can get you 80 % of the same audience because we also happen to know that whilst you're targeting teenage boys, moms will buy this for their kids. So that's still a good audience to get in front of. And you get all of this like vector space decisioning in real time. So it's not just the price.

Brett (26:47.58)

Yeah, it's it's it is the next best action of like this is who you you could reach this and it's not necessarily limited to the core audience that you set up as parameters to start the process, right? It could be ideas as to and and how are they how is that d how is the agent being trained on on being able to make those inferences and the and and kind of identify those patterns that say, Hey, there's audience extension play, you know, that that is adjacent to your core target.

Andrew pubX (26:56.301)

Exactly.

Brett (27:15.698)

that you might not have thought of like, you know, moms buying stuff for their kids or, you know, whatever it might be.

Andrew pubX (27:20.299)

Yeah, so I think the semantic version of that, so the thing that sits in the LLM type interface for the ultimate buyer is saying, hey, you've got a million dollars for this, we've gone to every publisher's sales agent and we think we can accurately spend $700,000. Would you like to just spend 700,000? Or do you want to use 300,000 to target adjacent audiences? I will check in every second of every day to make sure they're performing for the outcomes. But what we might find,

is that we've got a net new audience for you. And because one of the things we've built is this audience data spine in the middle, we can do that for the publishers and the advertiser data, and then we can push that back to the advertiser. So maybe you just spent 300k developing a new audience that you can target next time because it performed really well. And you would never be able to do that with a person.

Brett (27:51.71)

Yeah.

Brett (28:05.843)

Yeah.

Rio Longacre (28:06.416)

So you have your data spine, what are you are is it is it is your proprietary data spine you're using to match?

Andrew pubX (28:11.725)

So, yeah, the technology is internal. We've partnered with an infrastructure company, which I must shout out because they're excellent, which is Joe Lux Data Links, also based in New York.

Brett (28:21.38)

It w we we were we were that that story that we had about the bottle of wine, the photo with your s with your chief T ch CTO right was with Joe Lux. yeah, that was that was a serendipitous run in that we that we ran into and then we ended up for a great dinner. So that's great.

Rio Longacre (28:21.626)

Joe's a friend of the pod. We we we you know Joe really well, yeah.

Andrew pubX (28:26.549)

Yes, this was unbearable. Yeah, yeah, yeah.

Rio Longacre (28:31.576)

Okay, very cool.

Andrew pubX (28:35.127)

So very much friends of PubX, great guys. So we're building over the top of that. we have some proprietary data as well because we have the performance data of the campaigns and we have the tag on page in the ad server. But what we're trying to do is to match the first party data of the publisher, which is super valuable and really often underutilized. It sits in the ad server. Absolutely crazy. It's so rich.

Brett (28:55.632)

Yeah, and we talk about that a lot. Yeah, yeah, yeah. It's and and how how well and how do you do that cost effectively without you know, just sucking in a huge amount of data which is gonna drive up your your compute cost and things like that.

Rio Longacre (28:55.65)

It is mostly barely used at all. Yeah. That's a problem.

Andrew pubX (29:10.259)

So we will read the taxonomy within the ad server so we can understand what's there. One of the things on this is labeling, which I think is going to be a really interesting sub genre for agent it going forwards, which is, you know, we're reading this DMPs data. What is audience one, two, three, four, ABCD versus sports moms or whatever. But we'll ingest the taxonomy and then we'll activate it within a campaign. So we only need to know what's there.

it's not reading on the long level. And there's a bunch of conversation about this this week on LinkedIn where it's like, how much do you trust the audience? And I totally agree. think that's a super valid concern. I think if you're optimizing for outcomes at n equals infinite or almost infinite, like simultaneous scale, you trust the audience by seeing what's performing and the labeling is almost academic. I think people are still thinking about this.

Brett (29:58.525)

Yeah. Yeah. And when you say tw when you say trust the audience, just to to double click on that, do you mean trust the audience segmentation schema and the data behind the audience? That that that i is it actually reflective of

Rio Longacre (30:09.794)

Or even trust trust that the i the ideas of the person that you you're you're confident is the person you think you're targeting, right?

Andrew pubX (30:14.443)

Yeah, exactly right. You've been told it's this, but is it this? Like, how do you know that? I...

Brett (30:16.435)

Yeah.

Brett (30:19.846)

Yeah. And we and we know that historically and programmatic that's been largely it's almost like a flip of a coin. yeah, I'm I I might be part of thirty five I think the average person traditionally has been part of thirty five, forty different segments and it's not and it and a lot of it was based on sort of cookie farming and traditionally, right? Like

Rio Longacre (30:25.424)

Yeah, it's not that accurate. Yeah.

Andrew pubX (30:27.169)

Yeah.

Rio Longacre (30:37.008)

Yeah, Brett, weren't you a soccer mom or something like that and one of them you found out? Yeah. I mean

Brett (30:38.886)

Yeah Yeah, I was like a single female living in New York City, according to Exolate. and I in and so this was Exolate I went into I was part of Excelite that went over to the Nielsen acquisition and I just was I was that was my first view into my god, they've got me completely wrong. They've got me completely they have no idea who I am.

Andrew pubX (30:39.693)

Yeah.

Rio Longacre (30:53.07)

And I was just using cookie data, right? Yeah.

Andrew pubX (30:55.565)

Yeah. We used to buy that data, so we probably targeted you in some, you know. No wonder you didn't click.

Brett (31:01.118)

You didn't know you didn't know I was like an urban influencer female young in in hot New York City.

Andrew pubX (31:06.655)

Yeah. But that's it. That's it, right? It's like, how can we trust that the publishers label this properly? And it's like, well, how can we trust that a third party that gets paid, whatever happens, is labeled it properly, right? So I think the thing is what we can do is we can rebuild this from the ground up. yes, it should be labeled properly. Yes, this should be represented. But if you get away from the concept of a human bias, and you guys mentioned planning earlier, it's something I'm keen to talk about.

Brett (31:14.962)

Yeah.

Andrew pubX (31:36.171)

It's planning is typically done within experience. So it's been so pressured and crushed on the buy side of the agency side that it's certainly not an art anymore. And it's only a rushed science. But if you can give people the space to do a great job, then it's there. But the machine can just go and do that. It can just go and find the audience and it can go, well, I showed it to this person, they converted. Let's find more people like that in the vector space. Okay, now I'm building an audience for you.

And this goes from being soccer moms as declared by, I don't know, Ziff Davis in their ad server to these are the people that booked a test drive on your campaign for this car that, and then here's all the data we have about them, where they live, who they are, et cetera. So you can create these net audiences.

Brett (32:09.852)

Yeah.

Rio Longacre (32:18.638)

Andrew, how are you getting the data from the from the publishers typically? As would you do integration directly with their with their s with their I mean, some of them still have a DMP. They might be running like a I don't know, like a permitive or you know, they could you know some some of them might have like an an optable or they might have a C RM or like wha what like how are you getting the the the audience information, the identity, first party data into your system? How's that work?

Andrew pubX (32:43.16)

So currently we're mostly syncing this with the audiences they hold in GAM and I think we've got a couple of like DMP type integrations. But we will just continue to build those. So it's very much for us led by the publishers. And I'm really at pains to reiterate, this is very early. So this isn't, we're not 10,000 campaigns in how have we done it? But on each, yeah, yeah, it's over a few campaigns. So for the publishers running those campaigns,

Rio Longacre (32:48.058)

Okay. Okay, makes sense. Yep.

Brett (33:04.606)

This what you say a few thousand dollars, I think you said.

Andrew pubX (33:12.11)

We're up and running with them. We've got, we've synced in, I think they're mostly synced in GAM or maybe a DMP.

Brett (33:17.682)

So so does that fundamentally give the advertiser, the the buy side an advantage of just higher fidelity data? You talked about labeling and taxonomization. You would assume because you've got a much

richer source of publisher data that you just have a much greater degree of fidelity around these audiences, which are more likely to be accurate because it's this historical data that you've captured over people that might be if they're premium audiences, they might be authenticated. They're they're repeat visitors. You have tons of information about them over time, right? So I'm assuming that all adds to quality and fidelity when it comes to Europe. So you don't end up targeting me as a as a single female in New York City. Right?

Andrew pubX (33:57.933)

Yeah, or maybe we do if whatever campaign is resonating, but yeah. You've authenticated that.

Brett (34:01.451)

if I'm likely to perform for that cohort. yeah, so so you're saying the logic is is drivi is built around outcomes, but you are you are leveraging a much larger source of data than the traditional DSP would would leverage for building out and targeting.

Andrew pubX (34:16.844)

Yeah, we're leveraging it, we're taking it straight from the publisher. So this is, so you're right. It's about like cardinality, high fidelity. Publishers are quite good at maintaining this. They collect it properly. They put it in there because it's, really is a huge store of value. But then as you guys have said, it's under monetized because can the salesperson stand in an agency, you know, with this massive list and go soccer moms, Idaho? No. Auto intended dads, Delaware? Yeah.

Brett (34:31.687)

Yeah.

Andrew pubX (34:45.772)

And like just so it doesn't work like that. If you open a DSP and you go to buy someone's audience, you're scrolling through a list of 70,000 audiences provided by loads of great companies, but it's beyond the human scale to do it. But we're taking it straight out of the publishers environment. And then one of the, can do it in two ways. So if you're a buyer agent, you could say to our sales agent, I want to target this audience on this specific publisher or even on this website of this publisher.

Brett (34:47.837)

Yeah.

Andrew pubX (35:11.886)

And I just want to buy directly and you're basically it's like the publisher has its own MCP its hours But they go to that and the buy is kind of fairly transactional or you could say to us I want to target moms in Idaho and then we can go to all the publishers and say well Where do we have moms in Idaho? What's a great environment for moms in Idaho? and then we could blend the the inventory in the data into segments varying sizes and then we can return that and say

Brett (35:20.454)

Yeah.

Andrew pubX (35:39.522)

Well, this is people who are in a mom's segment from this publisher and they live in Idaho because we've got third party verification on a mom's website or made for moms, know, whatever. And so, and so this is 98 % match. You can scroll down to 60 % match, which is like same data, different website, less mommy or same mommy website, less mommy data. And we're combining into segments.

Brett (35:48.86)

Yeah, yeah.

Brett (35:59.401)

So Yeah, so if the advertiser's doing this, how much sovereignty do they have over that over all that pattern building and and sort of IP that's being built around their their targeting criteria, right? Which could probably it's gonna improve over time and it's gonna get smarter over time. How much of that is sovereign to them and pr you know and and protected for them so that it's not shared with like their direct competitor to be like, P and G's doing this. You know, it's it's obviously not a direct data steal, but it's a pattern.

Right. That's

Andrew pubX (36:29.569)

Yeah, that's a really great, I haven't been asked that, that's a really great question. think the, but yeah, think that's a great question. So I guess one way in which it's sovereign would be if we were blending that advertiser's first party data into the segment, because they would only have access to that data. So that's a tertiary layer of targeting. And then I guess the other way is that on a fundamental level, it isn't.

Brett (36:35.41)

Those are any drink conversation Rhea, we which people will hear about. But but but go on.

Brett (36:50.141)

Yeah.

Andrew pubX (36:58.241)

but only in the same way that it currently isn't. So if you were going for a Ziff Davis Lenovo Intendor audience that they've captured from IGN.com, you wouldn't have sovereignty over that buying it now, and nor would any other agency that wanted to buy it. But we would be making it available for you. So it would be sovereign in the sense of if we're adding your data into it, that would only exist as in your data being the advertiser data. That would only exist in that instance, that audience.

Brett (37:01.277)

Yeah.

Brett (37:08.541)

Yeah.

Andrew pubX (37:27.233)

and then you would have access to that within the portal that we give you and we can sync it back to your data store or whatever it is. Not the IDs themselves, but the concept of the audience. But yeah, it's a great question.

Brett (37:33.544)

Yeah.

Brett (37:39.164)

Yeah. 'Cause it 'cause it

Rio Longacre (37:39.513)

So Andrew, I I have a question about this is about the

About what me what types of deals this will apply to. And what I mean by this is so I can see like agent-to-agent transactions. I think anything that's direct sold, especially if it's a reservation, right, where it's like the CPM is defined by the publisher, type of media buy. I mean, what the limiting factor has always been it's a pain in the butt to do these things in negotiation, creating the assertion order, getting it done, getting it set up, right? So I I can see identifying that would be it'd be great. And I would expect more media would probably be transacted that way because you would take the

all of the potential things that slow it down, you remove them, right? And you can automate it, right? So in theory, more more media get transacted that way, which I think makes a lot of sense. But looking at if it's going to expand, like like it's say if it's going beyond that, how do you or can you actually have agent to agent transactions without kind of recreating an SSP, right? So the thing like someone needs to set the floor price price, right? Someone needs to someone needs to communicate to the ad server to GAM and get that information there.

Right. Someone needs to communicate back to receive the bid. Someone needs to send it back and then decide on the winner for the auction. Then someone needs to then, you know, work with the ad server to get the ad ad inserted. So it's like you're kind of that's what an SSPs were designed for was to manage this process, really do it at scale and do it super fast. So do you do you see this type of do you see this applying to that? Or is it is it TBD whether it does or not? Like like thoughts on it?

Andrew pubX (39:12.041)

Yeah, lots of thoughts. It's the ongoing recurrent internal conversation. think the, so firstly, I agree with the first statement that more spend will migrate to this kind of more direct buying because programmatic existed because this wasn't possible. Now it is possible. I advertisers love it. It's guaranteed. You know what you're going to get. Price is fixed, et cetera. So I think we'll see programmatic spend moving here. But this is to the point I made really early on about the protocols.

Brett (39:16.806)

Mm.

Andrew pubX (39:41.858)

been existent going forward. So you could absolutely, we went down this route originally, but then we thought that the direct buying route was probably more interesting, certainly better for publishers. And also it gives us a lot of the integrations we need to do the second part, which is you could still run what is effectively a bidder on ORTB and you could integrate with pre-bid to affect the same kind of like auction type mechanics should you have them. And then you can reconcile within pre-bid and then use price priority in the ad server.

to deliver an ad that affects the programmatic supply chain transaction, but without using any of the existing infrastructure. And obviously, every publisher in the world has pre-bid or a version of pre-bid header on their site. And so as the buyer agent, you could pump this out as an ORTB request that interfaces with the publisher's pre-bid instance and do all of the reconciliation within there using something like the universal creative and then push that back into the ad server.

So you can still skip those chains. I mean, the technology currently would be called a bidder. It's like the heart, the beating heart of the DSP, but without the rest of it. But you don't have to do it in a DSP. An agent can call that using RTV. And this is what the IAB is pushing, I think. Absolutely. Yeah, yeah. So you...

Brett (40:49.629)

Yeah. Yeah.

And the bid the bidder's open source. This is within pre the pre bidder open source framework. Yeah. Yeah.

Rio Longacre (40:58.991)

So you'd have to strip down versions of both DSPs and SSPs to do some of the functions, but only the basic functions that you need in order to actually facilitate what you just described.

Brett (41:06.941)

Yeah. And they're all open source, which makes it available to anybody that's that's running these agentix solutions.

Andrew pubX (41:06.957)

Exactly, that's exactly right. So this is this really interesting part where we've got these like, you know, almost like the dawn of the PC where we've got all these like garage engineers can really make a difference and make a run at this because the technology is open source. And that's why we think it was so important that it sits within PreBit. And we're long-term members of the PreBit community. We think what they've done is fantastic. And I think they've got a big part to play in the future of this.

Brett (41:33.566)

Yeah.

Andrew pubX (41:33.999)

And it's the key is distinguishing between the function of a platform and then the name brand platforms that do that function currently, because you can still facilitate the thing, but you don't have to be an existing business that does that. So you can still get around that. And I think for people that want immediate scale, they don't have a target price, but they just want it to be like as low as possible. You can do this through this ORTB system.

still way more effective than paying the fees of a DSP and an SSP because as you say, it's an open source platform.

Brett (42:02.801)

And ORTV's open real time bidding? Is that what is what is the O part of the the O R T V? That was not meant to stump you. I like O

Andrew pubX (42:08.674)

Great question. It's only openRTB, yeah.

Rio Longacre (42:12.815)

Yeah, no, well it's open RT open R T B but but it's only a yeah it's open RTB, but it's only a but I think that's only open RTB is what, like ten, fifteen percent, I believe, of transactions. It's not a huge number, but but yeah, I mean it's that is interesting. So I didn't thought of it that way. So in theory this could be applicable to all of the auction based transactions as well, not not just kind of the direct sold like invent in like reservable media, right?

Brett (42:15.527)

Yeah, open yeah, open RTV, that's what I thought.

Andrew pubX (42:35.382)

Yeah, we think so. And this is at the point where I think one of the things you guys say about is not just kind of, I don't think it's pointless, but pointless futurism or, you know, it's not just pontification, Performative futurism, right. So this...

Brett (42:46.259)

Perform no performative futurism. Performative futurism where you're like you're presc pr prognosticating about some future that Yeah, and it's like it it's like in theory, yes, but it's like divorced from reality.

Rio Longacre (42:48.494)

Perform it ever.

Rio Longacre (42:52.641)

Most futurists are so obnoxious and performative, so yeah, I think that

Andrew pubX (42:56.226)

Well, yeah.

I completely agree and we only have an idea on this stuff, right? So that's a technical path you might be able to do to do this. And you could kind of do that now. But what we don't know is where these changes are gonna go in, you know, even six, nine, 12 months, right? But I think the efficiencies brought by the technology, particularly the economic efficiencies are inarguable. You can't say that by taking these things out, this isn't better.

Brett (43:24.051)

Yeah.

Yeah. As long as that as long as that margin's going back to, you know, 'cause there's somebody's gonna come in to t to take to to want to grab some of that margin that's being freed up because it's not being handled by the DSPs and SSPs anymore, as long as that's going back to the publishers so that we continue to have a content universe that we all know and trust and love, right? And is your is your network of publishers, you said it's it's it's almost entirely premium publishers. 'Cause I sort of had a question around you talked about audience extension. Hey, we're gonna target this audience or like audience.

audiences beyond let's say current you know your current targeting parameters, right? Where does that happen? Where does that audience extension happen? Is it going outside of the premium the gates of your premium publisher ecosystem? and is that raise some concerns, meaning you might be going into just crap inventory that's in the open web that's not very trustworthy

Andrew pubX (44:18.934)

Yeah, so obviously if we did have that, I wouldn't say, yeah, it's going through the crap, but no, it is. So we're basically working through, and I'm going to be meeting with them in a week, we're working through the, like the Jones list as our North Star on this. it's premium, that there is more than enough inventory if it's handled properly on these premium publishers. think looking at as if Davis of this world type person alone, the stats are staggering.

Rio Longacre (44:34.476)

Yeah.

Andrew pubX (44:47.02)

in terms of unique users reach in premium territories. So for us, it's very much about that. And what we've, and again, in this kind of doing it responsibly way, we're bringing sort of small batches on in every couple of weeks or so, getting another five, 10 premium people live kicking on the system, seeing it so they can see a bit of spend as well. So we're trying to make this real and then working through the kinks and saying, well, okay, well, what doesn't it do that you would need it to do now you're seeing it in real life?

Brett (44:48.735)

Yeah.

Andrew pubX (45:16.462)

How can we make this better for you? And iterating this in real time, rather than just smashing it to a load of publishers in one go. But for us, it's about preserving and enabling to thrive the premium open web. And that's not premium in a snobbish way. That's not about views or biases or anything like that. But it's about people who produce quality content in good environments.

Brett (45:24.05)

Yeah.

Rio Longacre (45:38.707)

Not recipe websites, in other words, yeah.

Andrew pubX (45:40.494)

Some recipe websites, but the ones where it's only a mere reference to a trip to Catalonia.

Brett (45:44.512)

And when you say pre let's define that let's define premium open open web because you know and I i I d I don't think this is g necessarily the definition of of premium, but it does it need to be authenticated? Does it need to be subscription based gated audiences, right? Or is it a combination of both authenticated audiences and non-authenticated audiences for for those publishers that don't have gates?

Andrew pubX (46:09.452)

Yeah, so I think it can be a combination. I think we can handle a lot of that in governance. And I think that there's been a long-term debate, and I know the trade desks kind of top 100 kicked this off again a couple of years ago. And I think a lot of that seemed really to be driven by political bias more than anything else. it would be an environment in which someone would want to advertise where there is an audience that can be triggered agentically.

Brett (46:22.206)

Yeah.

Andrew pubX (46:37.868)

beyond any human bias. when we're, of course, we're not devoid of this. So when we're onboarding these publishers, we...

Rio Longacre (46:42.818)

Well, probably not the long tail of like of of websites, but you know, this would be the premium places where people are spending, let's say, probably ninety nine percent of their time.

Andrew pubX (46:46.476)

Yeah. Yeah.

Exactly right, yeah exactly and typically known good environments not like horribly jammed with advertising, you good load times, good user experiences and part of that can happen because we're driving more revenue per ad, we're driving high yield for publishers by taking out these fees so they don't need to stick all of these ads on the site, they can get the same amount of revenue for not doing so. But yeah.

Brett (46:50.942)

Yeah.

Brett (47:11.742)

Yeah.

Rio Longacre (47:12.14)

Andrew, I I got a question about the role of agencies in this, 'cause let's say agencies have the they have the scaled demand, right? You know, especially, you know, the ones that you know, the hold codes that are principal based buying businesses, but e you know, even big indies, right? They're managing billions of dollars of ad spend, right? They've got lots of advertisers, many of them who just give them budgets and they say, Listen, buy for us. And r today they're in DSPs, but at some point in the future, maybe they'll be doing this agentically. Do you see like who do you see driving the like

From the buy side, will the agents be run by agencies? Will they be run by brands? is it up in the air? Thoughts on that?

Andrew pubX (47:48.441)

Yeah, I think it's really interesting. And from the different agency conversations we're having, it's different within each agency. So some have seen where kind of like program senior, very senior programmatic or addressable type people have said, okay, well, I want to lean into this because I see this as giving me scale that I've never been able to get, know, at scale kind of play. In some cases, there's an AI SAR who's been appointed to look across everything. So, you know, we might be speaking to them, but they might also be looking at a platform that

reconcile data between different DSPs currently agentically or you know whatever it is but each agency has sort of peculiar problems internal to their culture or their makeup that they I think they perceive agentic will be able to solve which is different per agency exactly right exactly

Brett (48:35.039)

Or or or their business model. Yeah. Yeah, their revenue model and where and where where do they where do they extract either margin or g or how do they grow, right? In in is this

Rio Longacre (48:42.486)

Yeah, 'cause g yeah, 'cause that's been a conversation it's going back a year. Does does AI potentially risk the FTE model? Which I think's been under pressure for a while anyway. But generally speaking, I within the media business, there's been pressure on certainly margins in media. And so the th I think the initial belief in agency world was crap, this could be a problem if we can't justify billing for FTEs. But then, actually maybe this is good because we can automate things and have fewer people, but then

Andrew pubX (48:45.326)

Yeah.

Rio Longacre (49:11.798)

Like in theory if ad tick gets cut out, that unlocks a lot of money, right?

Andrew pubX (49:15.566)

Well, it does, but I think you're right that it's an internal struggle. And I think actually it's more difficult for the holdcos or any publicly traded business versus a high-scaled private indie where they can make decisions like this because it's not going to impact the next quarter. But this is a major transformation. This isn't just like when programmatic came out. This is like when the internet came out and people going, okay, there's this whole new way I can make money, but they already make money through these platforms. So you've got this almost this innovators dilemma.

Brett (49:26.654)

Yeah.

Brett (49:36.958)

Yeah.

Andrew pubX (49:45.326)

Well, if I'm shifting from this and even if I miss two weeks of spend through this platform By moving over here to make these savings when I report my quarterly's I'm gonna go yeah, we are down but we're gonna be up in the future and then

Rio Longacre (49:55.224)

Could be disastrous, yeah.

Brett (49:56.404)

Yeah, yeah, it's gonna it's gonna have a short term pain, right? And that is that is the the definition of the innovators dilemma where it's like, hey, you're so beholden to your existing revenue streams and your existing clients that you're less likely and you're listening to them, you're less likely to go and take a leap into some innovative direction that may be a loss leader for a little while. That's why you have you know, you spin these things off. There's all these kind of fancy ways that you can do this. But but 'cause companies, you know, they they've got numbers to report on when they're public especially.

Rio Longacre (50:21.494)

Well but you wonder it'd be a loss leader, I mean, if okay, if you're an agency then you're

You might change the payment cycles, right? Because instead of instead of relying on, let's say, SSPs to pay publishers whenever whatever cycle they pay them, you're then then maybe y you might have to pay them directly if you're just having agents negotiate directly with agents. So that you know so but then but then again if you're paying out less to ad tech overall, the margin should go up quite a bit. But then will the brands come back and say, Well, we know you're spending less in ad tech, now we want more out of working media. It's gonna you're right. I mean, this is a profoundly

Struct this is a profound structural change.

Andrew pubX (50:59.821)

Yeah, and it came in overnight. This is the key thing, right? It's not it hasn't been evolving over the last 20 years Yeah, I think so I think meaningful radar is the last year we've been we've

Brett (50:59.966)

Yeah.

Brett (51:03.645)

Yeah. This is this is like in the last couple of years or last year even, right?

Andrew pubX (51:13.643)

We've been in holdcos a lot over the last year and a real lot in the last six months. And so it's clearly being taken seriously. We want to present ourselves as, we want liquidity for our publishers, but for them this is, well, this is fundamental to our business model. We operate in a certain way. We've got 8,000 people working here, we've got 100,000 people working here. This is a major change that we'd have to do.

Brett (51:35.261)

You guys are fighting over this bag of cash, aren't you? Right? It's it's there's this push and pull between the buy and the sell side.

Rio Longacre (51:39.513)

Well, but I but I would think if agen agencies who don't embrace this and make this part of their business, I'm not saying it's gonna replace the existing business right away, but it could replace it quicker than we think, or maybe it takes a long time and the two systems run in parallel, which is probably what's gonna happen. But if they're not embracing this and experimenting, figure out how to make it work and figuring out, okay, if we're gonna need to pay publishers directly, great, or maybe scope three handles that and we figure out some way to manage that. I think if the if agencies aren't experimenting with this

And getting comfortable with this, they could be in serious trouble. Be my that'd be my guess.

Brett (52:09.851)

And and dire direct to publisher relationships, both economic relationships, right, and and technical relationships in terms of how the advertising's being delivered. To me that's that's the connection that really needs to happen. Yeah.

Andrew pubX (52:10.101)

I agree.

Rio Longacre (52:19.308)

And they've avoided those. Like historically the the big agencies have avoided those. Let's face it.

Andrew pubX (52:23.563)

Yeah. And I think, I think the thing is, it's also quite difficult because I think one of the things that we've seen this a little bit with some experiments that have then come back to us where I think if you're a hold co, you perceive that if you're just taking money to a publisher, they'll just go, cool. Yeah. I'll do whatever you want. But, but actually they've, they've got businesses to run. They've got practices. They've got roadmaps that you need to get on and things like that. And so we've spent six years working with these publishers to truly help them. Exactly.

Brett (52:49.897)

Yeah. For integration and implementation and they're trying to manage a content business that might be multi channel, multi platform, right? That's complex in and of itself, not to mention the monetization.

Rio Longacre (52:58.284)

And they've been doing it with their businesses shrinking. The publishing businesses, fifty percent of the size it was twenty-five years ago, it shrunk by fifty percent when you look at overall revenue. It's been it's been an actually devastating collapse. Many people not in this industry don't realize that.

Andrew pubX (53:01.248)

Exactly.

Brett (53:02.195)

Fewer people.

Andrew pubX (53:09.949)

Yeah, come-

No, think you still see newsstands, you still go on websites, but you don't realize the economics that drive it. But one of the things when we started Pavex all these years ago, pitching our first investors was most of the world doesn't understand how internet advertising works full stop. They've never really considered that the newspaper they used to pay for is now free somehow. So there isn't that sympathy for this in general. And now it'll probably swing back and change. But yeah, I think the thing is you have to be really sympathetic to the fact that these across the supply chain, these

Brett (53:30.484)

Yeah.

Andrew pubX (53:42.825)

major organisations that have built businesses upon this infrastructure and you can't be too cavalier. We are all in on agentic, we think it's going to happen, but you're talking about real businesses, real people, there is a huge impact.

Brett (53:56.382)

Yeah, if you're too cavalier, you're gonna get blacklisted and and just not be able to participate in one way or the other, right? There's 'cause people are gonna protect their interests and their revenue streams, right? For especially the publicly traded companies because th that, you know? Right?

Andrew pubX (54:00.131)

Yeah. Yeah.

Rio Longacre (54:09.805)

Well, but looking at publishers, I mean I think historically they've not been great at selling the any any digital inventory, right? I mean I these are organizations that I and Brett used to work again at, right? These are organizations that were great at selling space ads and d different like not like traditional products, right? And suddenly they had to sell digital stuff and some some are better than others, but generally they said, let's just you let's just put some tags on our website and we'll call it digital and we'll just ha you know ha you have

Brett (54:35.047)

Maybe some rich media ads, right? You know.

Rio Longacre (54:35.167)

Open RTB, right? And then next thing you know, it's under a dollar for CPMs and 60 cents and dollars being siphoned off, right? sometimes more. So it hasn't worked out well for them. I don't think they've been very good at figuring out how to sell, how to translate the expertise in traditional and like into into digital. And they've, but if this happens, this could be a huge unlock for them. But what happened I mean, in theory, like what happens to their sales teams? Do they even need sales teams, or maybe they need smaller or different looking sales teams?

Brett (54:42.025)

Yeah.

Andrew pubX (55:04.308)

Yeah, so it's another really pertinent question to all of this. And the thing that we believe is you can start to do value add stuff. So in our previous business, we also had a content marketplace. We used to work with Condé really closely. And we would say we would love a, we need an advertorial on GQ, let us know the price for some traffic driving ads, cetera. And then, it's 100K. And then they go, come in for a meeting. And you'd go sit there and be like.

Now you could spend 100K or for $5 million you could sponsor a Range Rover at London Fashion Week. And we'd be like, right, well we only need the advertorial, we're a digital company, but thank you. But this is what sales teams can do now. Yeah, but this is what sales teams can do now is like they've got all this amazing data, they can see who's spending and where. And rather than just doing RFPs that they send back to the agency every time and the agency goes, the New York Times, wow, it's a newspaper.

Rio Longacre (55:44.445)

Thank yeah. Thanks, but no thanks.

Andrew pubX (56:00.271)

Well, they know what it is. They know the audience. They didn't need to do an RFP for that. They can do this higher value add stuff. They can go to the added us CMO and say, hey, I see you're spending a lot on this section. Why don't we do something cool with you? Why don't we even sell, you know, to quote Tom Triscari and his tech.

Rio Longacre (56:12.941)

Yeah, maybe the RFP pro I mean, it's ridiculo I mean, that's so much time in these sales teams is spent responding to these RFPs. In fact, a l I many publishers are creating systems just to ingest these and speed up a process, which I think would be an improvement. But you're right, maybe in the future you just don't need many RFPs.

Andrew pubX (56:19.128)

Yeah.

Andrew pubX (56:26.798)

Well, I think so, and they can focus on value add activity, right? How can we actually do something meaningful? What's a cool execution that our brand can do? And this is one of the things that's been lost in programmatic, which I think is the saddest, but because the inventory is being commoditized.

Brett (56:37.555)

Yeah. That's what that's what we did in Gannette. That's largely what my team did in Gannette. We built custom

oftentimes rich media driven implementations, which again to your point, Rio, was sort of taking like, you know, the the the the belly wrap, all these kind of interesting things that you could do in print in the print ecosystem of like, you know, a full wrap around the newspaper or a belly band or or some sort of insert, right? They took that model and sort of applied it to digital and said we've got all these interactive things we can do with this and then you would customize an entire program around a brand's requirements and needs. And so there there would be some, you know, obviously editorial discretion.

Andrew pubX (56:43.374)

Yeah.

Brett (57:13.171)

And there's still a Chinese wall between sort of journalistic integrity and the advertising needs. But there was some push and play in that department. And that was kind of that's what they were trying to do to differentiate their their content, which with USA Today and stuff was getting kind of commoditized, right? Because of all the sources of news and that type of news wasn't exactly investigative journalism, so it was sort of losing out to UGC and a lot of the more easily created digital content. So yeah, that was that was it.

Andrew pubX (57:30.232)

Yeah.

Brett (57:43.081)

How do we package this up in a digital way that's super attractive to brands? So for certain certainly custom, very custom.

Andrew pubX (57:47.03)

Yeah, it becomes the creative solutions team rather than just the respond to RFP team.

Brett (57:52.372)

Yeah. Yeah.

Hundred percent. So li so speaking of a agentic and buy side, you know, agents negotiating directly with sell site agents, how do you think about some of the and I think this is more of a obviously a buy-side, you know, brand and and agency consideration, things like ver you know, ad verification, brand safety, fraud prevention. There's obviously players like IAS and Double Verify that play in that space to to ensure that ads are being delivered in safe environments, etc.

et cetera, et cetera. how do how do you think about that in this entire yeah, in this entire e a gentic world?

Rio Longacre (58:28.589)

Yeah, how's that work here? Yeah.

Andrew pubX (58:32.814)

Yeah, I think probably the answer in the short term is.

exactly the same as it works now. And we're speaking to all of those companies about encoding them using their measurement, like the same for like IVT and things like that. And I think that's really because of the too many miracles scenario where you're already asking for a huge change in behaviors. And if you go, hey, can you buy through us in a new way? And don't worry, we're going to check that it's going properly. This is a huge ask. So I think those companies will have a place in this. And I think they'll participate agentically. Maybe you can call them through

Brett (58:44.532)

Yeah.

Brett (58:57.887)

Yeah.

Andrew pubX (59:05.84)

signals agents, maybe as a buyer you can go, hey, how am I performing for this? And they'll be able to respond in the same terminal and tell you your stats. maybe you could have an agent to agent threshold where if something is below a certain level of brand safety, the IAS agent tells the PubX sales agent to stop selling it. And you can imagine it agentifying in that way, although this is almost futurism. But I think currently they have a role

Brett (59:26.941)

Yeah.

Brett (59:32.244)

Yeah.

Andrew pubX (59:35.8)

to play. Now of course it's a controversial role because if you're people like Jonah at Mobian you're saying well these guys are blocking amazing content because it's got one bad keyword in it and I think there'll be a next raft of technology.

Rio Longacre (59:45.361)

That that is happening. About forty percent of like news publisher content is being currently being blocked and cannot be monetized.

Andrew pubX (59:50.991)

Yeah, like, Taylor Swift bombed at this gig. Okay, no advertising, but a million people read it because it's about Taylor Swift. So I think they have to get better. But in this new in the modeling that's enabled by this, I think they also probably can get better at that. And I think there's a real opportunity for a lot more nuance to go into the way that they operate or for somebody new to enter the market and to do that. But I don't I think the critical thing here is this isn't about saying

going right back to the start of programmatic where someone would just email you going, you've got a million views. Can I my check, please? We should do the best practice stuff to make sure that verification exists, but we can probably improve the things that are still lacking, particularly those that punish publishers.

Brett (01:00:32.701)

Yeah, and you're not gonna be doing that. That's gonna be the the job of those providers adapting and and adjusting their own agents to be able to to to do a lot of this diligence within e whether it's with it you know, by communicating. I mean I guess it would be through C P or Ad C P directly with the agents that are doing the buying and the selling, right?

Andrew pubX (01:00:50.698)

Yeah, exactly. And I think all of these companies, or certainly the ones we're speaking to, and probably just every company, is developing some version of an MCP over its underlying product offering. So I think that they see this.

Brett (01:01:00.979)

Yeah. And that and how futurism is that? Or futuristic is that? Is it is it is it happening today in tests like with with the double fur? We gotta ask Ligorsky this.

Rio Longacre (01:01:08.141)

Well yeah, pretty you remember we talked to we talked to s we talked to Xperia and they built C P servers, they said no one's using them. They said they built them, they deployed them, like n they they they expect there will be usage in the future. It's not like it's not like that they think it was a waste of time, but they they like the market hasn't caught enough to the technology was their point. So love your thoughts on that.

Brett (01:01:19.187)

Yeah.

Andrew pubX (01:01:25.55)

Well, I think that's right. And I think this is again the problem with all of the hype around it. So you read AdExchanger and then you rush to your product team and you go drop everything. We've got to be agentic. But then you need to bring demand. And this is really, we see this as our job as much as it's all of the other participants, the scope threes of this world is getting demand flowing so we can validate the ecosystem. And then we can work out who does what within it almost. It's not that important, but if we can all lift agentic up.

Rio Longacre (01:01:34.901)

Right.

Brett (01:01:35.357)

Ha ha.

Brett (01:01:47.273)

Yeah.

Andrew pubX (01:01:53.847)

preach the benefits and get spend for anything that's useful. I'm not surprised to hear you say that Rio that it's not being used, but I do think it will be used. And I think there is an advantage in being ahead of the industry on that, but there isn't a kind of real centralization around evangelism here. There's a lot of people talking to a lot of different people.

Brett (01:02:10.899)

Yeah. Yeah, let's hope we don't pollute the word agentic, which like orchestration is becoming it's starting to carry some baggage, right? You know, it like there's a bunch of these words where like Yeah.

Andrew pubX (01:02:18.378)

Yeah.

Rio Longacre (01:02:23.533)

But it can mean anything, right? I mean, or nothing or or something important. I mean, and and I think that's the problem here is being b beaten to death. But the question though about like setting up an ingenic trade. So do you start with the publisher and say, Okay, we've got a publisher, let's get them set up, let's get their inventory so it's structured in a way where it could be ingested by agents and then, you know, let's build like make sure that we can communicate this with let's figure out what types of deals we're gonna tr like is it maybe just direct sold.

Andrew pubX (01:02:25.549)

Yeah.

Rio Longacre (01:02:52.279)

preservable media first. and the and then and then you go to an and then you have to find a buyer to actually test it out. How how does this work? Do you go to an agency and say, hey, can you figure some clients? Like how are you actually like I know it's not a ha big you said it's three thousand dollars, I think a month, three thousand pounds a month. Like that's like how do you execute these deals? Can you maybe walk us through that a little bit? Because I think that's where lot of people are stuck. The theory sounds great, but okay. Publishers want it. Seems like

Brands want it, but how do you like do which side do you start from?

Andrew pubX (01:03:24.354)

Yeah, it's a really good question. And this is the closest we can get to saying we're building a marketplace. We're building both sides of an ecosystem. So I found that out after Rachel emailed you guys. And I was like, hey, don't say that. And she's like, well, you should speak to your own website. Yeah, by the time we go to press, it won't exist anymore.

Brett (01:03:32.499)

The word marketplace does appear on your website quite a few times, by the way. So Yeah, you're gonna have to be like

Rio Longacre (01:03:34.519)

No.

Brett (01:03:42.407)

Yeah, go to the go to the publisher section of your website and be like, We need to take that off

Andrew pubX (01:03:51.407)

But yeah, it's 3000 a day at the moment as well, which is still not a major amount, it's, so we're always building supply first and we have great relationships with publishers. So we're not going in cold and we've basically just set expectations around what's likely to happen. So we haven't said to them, if you go live now, you're going to get a million dollars next week kind of thing. And we were asked by a very premium global news publisher this week.

to give them forecasts. And we just said, we're not going to be able to do that right now. As soon as we can do that. Yeah, when we can, we will. We are doing everything possible to get you as much money as possible. But right now, whatever we give you is gonna be wrong. And they really, they respect this. know, you're treating people with respect. we do the sales side first, we get it on boarded. We run a couple of tests just to make sure it's up and flowing. And then we can connect in. As I say, we work with Scope3 who are transacting campaigns as a buyer agent currently.

Rio Longacre (01:04:22.092)

You refuse to give him a forecast, basically. Yeah. That's smart.

Andrew pubX (01:04:47.148)

And we are exactly as you guys have said, speaking to a lot of agencies, lots of Indies who are doing this, we're bringing demand through from that side because we want to prove it can work. That can even be as a managed service where we show them what we're doing in the agent. We don't want them to break from what they're doing, but we can say, you know, if you've got some test budget, let us know. We can show you the benefits. You look at the improved economics. We'll do the analytics with you. And we just want to get spend through. So we're not trying to make money from this right now.

We just want to get it flowing and see that we can create this ecosystem. But it's always sell side first because we never want, firstly, if you've got demand and nowhere to put it, that's a problem. But secondly, we don't want the publishers to feel like we're coming to them with this sort of ransom where it's like, well, now we've got money. If you don't do it today, you're going to miss out because we really want to be a partner. This is about building for the long run. So it's always supply first.

Brett (01:05:37.821)

How yeah, how are you guys thinking about pricing? I mean, 'cause algorithmic pricing has certainly drawn some regulatory scrutiny in financial services and other places, right? And has anyone in the ecosystem really thought about that yet?

Andrew pubX (01:05:45.197)

Yeah.

Andrew pubX (01:05:50.855)

I think probably the most addressable model in the short term is just a percent of spend, but just a relatively low percent of spend. My partner works in financial services and for them it's basis points, right? So I don't think we're going to get there, but you know, massive cut from any one of the fees that's currently leveraged between the advertiser and the publisher and only charging one fee still makes an enormous amount of revenue go back. I think it's very low.

Brett (01:06:02.708)

Yeah.

Rio Longacre (01:06:13.958)

Where do you think it might be? Can you estimate?

Andrew pubX (01:06:18.69)

double digits is the kind of ballpark that we're talking. yeah. Exactly. all of our modeling is on the 10, 15 kind of spectrum. And then there's obviously, well, it's not obviously, but we can also do, we can scale data for publishers. So we could take their first party data, but we could distribute it on other publishers' sites, as long as they opt in for this. So then there's potential data brokering.

Rio Longacre (01:06:21.936)

a huge difference. Yeah. Instead of like fifty, sixty percent, you know, if you brought it down to like fifteen, I mean twenty. Yeah.

Andrew pubX (01:06:47.574)

revenue over and above the inventory revenue. So even if they don't get their inventory in a campaign, they get a data fee for that as well. And then I think once we're engaged, there's so many new opportunities to say, well, this new way of agentic buying has come out that we can help them facilitate that we hadn't even thought of that has this type of revenue model, because there's so much more liquidity to play with. You can improve the models between the buyer and the seller, right?

Again, I think a percent of spend is going to be the short term thing. think the publisher is going to get so much more of that spend, which is critical. And then we can iterate this and do it in a way that's going to be the best for them to work with.

Rio Longacre (01:07:27.03)

Who benefits ultimately and who gets hosed in this new model, do you think? Or is it too early to say?

Brett (01:07:32.095)

D somebody di by definition get hosed in this

Andrew pubX (01:07:33.102)

No. Yeah, there's a hosing. There's a hosing effort. Yeah, I think the incumbent platforms get hosed. I think the trade desk get hosed. I don't think they'll go, I mean, people still, there was still something like a million people paying dial-up six years ago, right? So things don't go over the cliff immediately, but I imagine that the trade desk is maybe resized to be the same size as a kind of like a basis or like this kind of, you know.

a variant type DSP type platform. I think the current, no, they're not and they're happy to go very loudly and then loudly in a completely different direction until it resonates. But I think the SSPs, again, I don't think all of them will go, I think Rajiv is doing some fantastic work in general for raising the profile of Agentic and I think that's great and we're totally aligned on that.

Brett (01:08:04.817)

Not gonna go quietly, I'll tell you that.

Brett (01:08:13.811)

Yeah.

Rio Longacre (01:08:26.038)

Yes.

Andrew pubX (01:08:29.945)

but I don't think there'll be as many. Anything that's just a base reseller is dead. Yeah.

Rio Longacre (01:08:32.896)

Yeah, the SSPs that are just like like grab grabbing bids and passing them on, right, and and just marking them up are basically just just resellers. I mean all of the all of those go away. There's no way they survive.

Andrew pubX (01:08:43.757)

Yeah, there's a small cohort of like the high value resellers where that's the only mechanism to get data in. If they're doing that, that's great. People will still do that. That will still perform. But yeah, I think that space consolidates massively. think they're also in a rush to move either really far down to be very close to the publishers rather than just an SSP or far up to be close to the buyers. Exactly. So I think that that will change, but I don't...

Rio Longacre (01:09:05.159)

they have to. Yeah. If they don't have direct publisher relationships, they're in trouble.

Andrew pubX (01:09:13.667)

I think that the people in it can move around. think there'll be a huge growth in agentic businesses and people within this space. But yeah, think the DSPs and the SSPs have had an amazing run.

Rio Longacre (01:09:25.356)

Well, I think the DSP is the only way for them to really probably thrive in this new environment is if they get more involved. And I think almost Brett, they have to become like focus on MMM, right? And the planning and allocation and move further upstream, right? And and you know, where the decisioning is gonna start happening. I think if they do that, th they could do pretty well, right? But I think that the ones

Andrew pubX (01:09:37.295)

Mm.

Brett (01:09:39.348)

Yeah.

Brett (01:09:45.393)

Is it i you th y do you think like does like chalice AI come to mind? I mean with containerization Yeah, where you're bringing in more data within your within your containers which can help with decisioning in a way that that you just ordinarily can't.

Andrew pubX (01:09:49.647)

That's what I was going to say. Yeah. Now if a DSP could

Andrew pubX (01:09:58.414)

Yeah, so if you, but if you look at the smart work in this space right now, like Bedrock, Index, and people like this, this is really interesting. containerization used to be called on-prem, right? And when we used to have bare metal servers, we used to sit them in the same room, so we had minimum latency. yeah, you've gotta be on-prem. You can't get this data out of the building. But it's still really smart, and Bedrock are doing some really cool stuff. And so are Index Exchange. And so I think that...

Brett (01:10:14.611)

Yeah. On prem. That was a demand that clients make. It has to be on prem.

Rio Longacre (01:10:14.868)

Yeah.

Andrew pubX (01:10:27.725)

I don't think they will go away as an SSP. think they'll change what an SSP is, but that's putting both of those systems next to each other, which is really smart. I think the problem with, for example, the trade desk is they're not actually connected to value minus the utility they provide as a platform. So if that utility has gone away, so is all of their value versus an SSP still has a relationship with a publisher, which is a driver of value. And Chalice, I agree with you that

Brett (01:10:32.136)

Yeah.

Brett (01:10:53.129)

Yeah.

Andrew pubX (01:10:56.429)

the DSPs should move upstream to do the MMM stuff. But if they could do that, Chalice wouldn't exist. So Chalice is the smart stuff sitting over the Chalice will move upstream and do the MMM stuff and work with with holdcos and things like that. But if the DSPs could do that themselves, Chalice wouldn't exist as a business now. The fact is that they're relatively dumb platforms in the true sense of modeling. So I think you'll see the Chalices that have spent years learning.

Brett (01:11:15.796)

Yeah.

Andrew pubX (01:11:24.727)

on all this data from the DSPs can just jump one stage up and go, hey, hold co, you we can do this thing for you now. And I think sadly that will, you know, spell the, that'll be the death knell of the Demonside platforms.

Rio Longacre (01:11:25.558)

Learning how to model, yeah.

Brett (01:11:37.309)

Yep, yep, it's the innovator's dilemma again. Right.

Rio Longacre (01:11:39.744)

Yes, Adam, if you're listening, I mean I I don't know what your product roadmap is, but you know, some good ideas for it for sure.

Andrew pubX (01:11:45.033)

Yeah, be happy to share them.

Brett (01:11:45.408)

Yeah, we've had we've been out about a couple of times. So should we should we go to quick hits? This has been great. It's been great and I know we're all we all have well you're ending your day. We're slightly beginning our day. are you gonna go for a little ride? And a countryside ride, right?

Rio Longacre (01:11:48.906)

He's great. Let's do it.

Andrew pubX (01:11:52.131)

Yeah, go for it. Yeah, thank you.

Rio Longacre (01:11:58.156)

Yeah, we're standing between you and your hap in happy hour. Or or a bitter, yeah.

Andrew pubX (01:12:02.37)

Yeah, a warm flat pine in a 500 year old boiling building.

Brett (01:12:07.039)

Yeah, exactly. Exactly. Gonna hand pump that ale. Alright. So so what's one thing publishers still don't fully appreciate or understand around AI and everything we've been

Andrew pubX (01:12:11.552)

Yeah.

Andrew pubX (01:12:23.148)

I think probably how difficult it is to build and the amount of effort that's required to build it. Because I think there's still that mentality of, you know, this is maybe something we can crack out ourselves. And I think it seems simple, particularly when it's effective. But it, and they absolutely could do this themselves, but it requires an enormous investment in the right skills. It's not just putting a little, you know, half of somebody's resource on making this thing happen. So it's a huge undertaking.

Brett (01:12:47.049)

Yeah.

Rio Longacre (01:12:51.167)

What's one word or one phrase you would use to describe the the ad tech stack in five years?

Andrew pubX (01:12:59.16)

well, how to avoid the one word that's incredibly obvious. I do think I would use the word agentic, but I'll think more about, would say improved.

Brett (01:13:11.175)

So so what do you think the timeline of this is? You ch we're talking about sort of futurism, future versus around all of this, around agentic trading and in buyer seller agents actually doing you know, a large share of the total trading. what wha how I mean what are we looking at from a timeline perspective?

Andrew pubX (01:13:28.226)

Yeah, think for a large share, think 2026 has been this sort of anto room for this. I think you'll start to see real uptake in 2027. I think for a large share, it's probably a couple of years out. But I think the pressure coming from the advertisers as this filters in, it's very different to normal. This isn't CTV coming out to eat linear. This is, I'm the CMO of Nike and I use Claw to do my PowerPoints.

Brett (01:13:42.206)

Yeah.

Brett (01:13:51.913)

Yeah.

Andrew pubX (01:13:56.451)

why is my agency not using a version of this to do my buying for me? So I think the macro element of this accelerates there. But every single planning meeting we go to, I'm like, so this will take two weeks. And the guys are like, no, it'll take two months. So I'm always very optimistic. So I think maybe one to two years is a reasonable timeframe to get enough things that need to exist built and enough old things that are still in, contractual lockups, for example, to start to fade away.

Brett (01:14:00.425)

Yeah.

Rio Longacre (01:14:23.957)

Do publishers do better or do they get hosed again with this transition?

Andrew pubX (01:14:29.762)

I think they can only do better. think they're, yeah, I can't see. Yeah, it really is. The main problems that they suffer from are capacity where they interface with the buy side. So we've mentioned the great sales teams they have, and so many of them do, and we work with so many of them. But there's only so much that you can do on the buy side when they're also reducing the people you can speak to. When you don't have that capacity issue,

Rio Longacre (01:14:32.839)

It's that bad, right?

Brett (01:14:35.749)

It's only it's only up from here. Right.

Andrew pubX (01:14:57.302)

suddenly all of their amazing audiences are available everywhere. If they use our sales agent, our Japanese country lead can sell them to Dentsu in Tokyo and they've never been to Japan, right? So they become available everywhere. They can really only benefit from this. The economic incentives are just too strong for them not to be able to win.

Brett (01:15:16.913)

Yeah. And by by extension of that is what are your feelings about the open web? There's a lot of proc prognosticators out of there out there saying that, you know, the the open webs just doesn't have a future, that it's gonna be eaten by all the walled gardens and things are becoming increasingly more walled. what what do you think of that future?

Andrew pubX (01:15:36.695)

Yeah, no, well, I've never seen a newspaper or listened to a radio. I think that you, I think that there will be a decline and then a plateau, right? So I think that every, every new thing begets this hystericism about the previous thing. I don't think the open web is going to go away, right? I do think it will, numbers will decline. I do think, search through LLMs will, will eat a lot of this. And I think there's particular categories that get

Brett (01:15:51.935)

That's true.

Brett (01:16:02.186)

Yeah.

Andrew pubX (01:16:05.155)

that do get hosed, right? Like if you wanna know what to do this weekend in London, you're just gonna ask Claude, right? I landed in JFK the other day and I was like, I've got four spare hours, what should I do? And it's like, hey, go check out this thing. And I would have gone to Time Out, I would have read the New York Times. That will suffer.

Rio Longacre (01:16:11.571)

Yep. That's what I do already.

Brett (01:16:21.207)

I miss Time Out. I miss the old Time Out magazine, you know, and you'd flip through what's happening in London, what's happening in New York City. Yeah, and I'd go all the way to like the live events and and I just missed that experience.

Andrew pubX (01:16:23.629)

Yeah, it used to be a thing, the physical copy of it. Yeah, fall down the pages and be like, I'm gonna go see that. Yeah.

Rio Longacre (01:16:32.469)

Yeah, but Brent, you wouldn't have time, like you would like y you missed that experience, but to think about it now, like just the the thought of like having a subscription to it, getting it or finding it, like it's just who has time anymore, right? So I I think it's this is so much better. I mean I I I like a couple of weeks ago my my a couple of weeks ago my wife said to me, She's okay, we have the night off. Our daughter's like gonna be with a friend. So I suddenly had to plan a date night. So I just said, plan it out and it was it was great, right?

Brett (01:16:43.636)

Yeah.

Andrew pubX (01:16:43.715)

Well, exactly.

Brett (01:16:47.024)

Instant access.

Brett (01:16:59.316)

Ha ha.

Rio Longacre (01:17:00.647)

It suggested three restaurants. I ended up picking one of them from reservation. I mean I could have asked it to try to call, but I didn't try I don't I don't trust the

Brett (01:17:05.917)

And then Chris is like d Chris is like, did AI choose this restaurant, Rio? This sucks. Is this is is this an AI slop restaurant? Like why would you choose this? It wasn't me, it wasn't me, it was ChatGPT.

Rio Longacre (01:17:10.411)

No, she's like, You actually planned No, she said you actually planned. Blame it. Yeah, no. It it ended up being she thought that I actually took the time to plan it, which is I never have time to do. So it was good. So yeah, th these tools are great.

Andrew pubX (01:17:14.529)

Yeah, yeah, yeah.

Brett (01:17:25.415)

Yeah. Yeah. Well we we I think Andrew and I both agree that we wish we were with you this weekend to go into the mountains, Keystone and see the Bluegrass Festival. That sounds like a ton of fun. I mean you're gonna have a blast, right? I now I need to go and find something to plan for my weekend because I didn't have much

Andrew pubX (01:17:33.059)

Yeah.

Rio Longacre (01:17:37.131)

It's gonna be super fun, yeah.

Rio Longacre (01:17:41.941)

Well maybe next year ha next year house it's it's a the first weekend of August pretty much every year is the bluegrass festival out here. Yeah. It's a good time.

Brett (01:17:48.477)

Yep, yep. And Andrew, just so we'll end with this a bit. Rio and I go to jazz fest every year in New Orleans. so that's a big thing. So we t we tend to do these like these escapades to musical you know festivals and things and and it's a ton. It's right after possible. Yeah, so we do Yeah. Yep. It is just if you're if you're a lover of music and you're a lover of food

Rio Longacre (01:18:03.563)

In fact it's right after possible, which is in Miami, is like it's Jazz Fest, so I have to basically tell my wife I'm gonna be gone for a week every year if possible and then it it's really fun. The Jazz Fest is great.

Andrew pubX (01:18:06.402)

really? Yeah!

Andrew pubX (01:18:14.014)

It's been a long-term ambition.

Brett (01:18:17.759)

'Cause the food is phenomenal. You just it's just like you feel like you're just I don't know, sucking the bone marrow of life, I think is might be the way to put it. 'cause it's just it's so rich. Like you're just like the amount of experience you have in like a twelve hour span there.

Andrew pubX (01:18:18.99)

Yeah.

Brett (01:18:32.073)

is just insane. Like you've seen so much and tasted so much and and you know talked to so many people. It's really fun. So we recommend everybody check out New Orleans Jazz Fest or your Bluegrass Festival in your local region or whatever you do for a for for fun. But thanks Andrew, this has been b a blast. I'm hoping maybe we can we can hang out in the city, maybe even ski together sometime. but definitely looking forward to meeting you in person. And for everybody that made it this far, visit us at Dev Dev Dev

Rio Longacre (01:18:49.419)

It's been great.

Andrew pubX (01:18:50.563)

Yeah.

Brett (01:19:01.79)

ai and we're on all the platforms Spotify Apple Podcasts TikTok etc that you get your content and we will see you next time thanks

Andrew pubX (01:19:12.655)

Amazing. Thanks so much.

Rio Longacre (01:19:13.611)

Thank you.