[RECORDED OCTOBER 2025]

Brendan Norman: 0:00

I think one of the first agent uh protocols that will launch and it's already started to test are these sales agents. And like they're automated systems that kind of go in and try to figure out bid strategy. They try to figure out how do we get the best campaign at the best value. So you you basically will be setting, you know, some different parameters, and then over time I get a learn and kind of fine-tune it. But effectively it's like having this very powerful uh system that kind of automates a lot of these tasks for you that previously were very manual.Kevin Kerner: 0:30

Hey everyone, this is Kevin Kerner, your host of Tech Marketing Rewired. We have spent the last five years obsessing over the death of the cookie, but while the industry was doing a bit of panicking about tracking, AI has been working in the background to rewrite the rules of targeting. Today I'm excited to sit down with Brendan Norman, the founder of Classify. Brendan spent years at Meta helping build the Facebook audience network, so he has seen the machine from the inside out. Gives him really unique perspective. Now he's building what he calls contextual 2.0, where we aren't just matching keywords anymore. We're talking we're talking about AI that reads a web page like a human does, understanding sentiment, tone, and nuance to place ads with really surgical precision. But um building the tech's only really half the battle. As we find out, Brendan puts it and Brendan puts it, the ad tech industry is like New York City plumbing, old, leaky pipes that are hard to replace. Uh, we discuss uh how to innovate in a legacy infrastructure, the rise of the eugenic web, and why the future of advertising isn't about following the user, but understanding the moment they are in. But before we get started, I wanted to mention this this episode is sponsored by my company, Mighty and Trude. Look, too much of B2B marketing is just random acts of management. Mighty and True is an agency that replaces that chaos with systems. We help B2B tech companies build revenue engines using proving playbooks and our own innovative technology, not guesswork. If you're ready to stop improvising and start scaling, visit us at Mightyandtrue.com. So I'm excited about this one. Let's get to it. This is Tech Marketing Rewired. Welcome back to Tech Marketing Rewired, and I'd like to welcome Brendan to the show. Brendan, great to have you on. Thanks for having me, Kevin. Yeah, I'm really excited to have you here. Uh maybe to get started, I think um uh you could give us a little background on yourself and maybe a little bit about Classify, and then we'll get into it.

Brendan Norman: 2:23

Yeah. Short and quick backstory. I got really lucky over a decade ago and joined uh a small supply-side platform called LiveRail. And that was my entrance into the ad tech world. And we very quickly got acquired by Facebook, and I got to spend the next five years building out Facebook Audience Network, which was taking the power of Facebook advertising and allowing that to extend off of Facebook into third-party mobile apps. So think Pandora and Spotify, Words with Friends, basically anything in the mobile app ecosystem, you know, we were helping to monetize using Facebook ads. And I got a really cool seat, you know, in helping to shape a lot of how the best advertising engine that's ever been built works, in addition to understanding what publishers really need and also what advertisers are looking for. And that's kind of shaped the rest of the course of how I've thought about advertising technology to begin with. And then fast forward to classify. After I'd left Facebook, I was at Unity for a while, helping to build their ad business, did some consulting in the go-to-market space for some other startups. And I was building an ad to a platform in a totally different space and just for myself that I thought was interesting. And I tried to market that platform and couldn't find the right content to run ads for it. It led me down the journey of kind of asking, why doesn't this exist off of Facebook? And I found out that nobody had really cracked the code for understanding at a very granular level how do you scan a web page and how do you read that blob of content in a way like a human would, where we are just naturally really good at understanding, you know, the category of the webpage might be sports, but there might be a lot of nuance to the specific ways that they're talking about it, you know, the specific entities, the people that they talk about, the products, you know, certain parts of it talking positively and negatively. And to do that is very complex. It hadn't really been good at understanding contextual analysis and sentiment analysis and some of these other terms for content, you know, for a long time, just because ad tech is huge and there are billions and billions of web pages and hundreds of billions of transactions that happen daily. To do that at scale takes a lot of processing power. Um, but thankfully, you know, we're kind of at the the point where the technology is finally caught up to be able to support. And you asked about Classify. Happened to meet some other folks who were working on kind of similar things who were interested in the same problem. And we came together and just decided to build a company around it to initially solve the marketing problem that I was facing. And then we realized pretty quickly I think other folks would benefit from this too if we were able to help them find the right web pages to run ads against. And so we we got started, we built a full product and we launched it uh back in June, right before Cannes.

Kevin Kerner: 5:21

That's awesome. Remind me, what what was the product that you were trying to market? You told me before, I can't remember what it was.

Brendan Norman: 5:27

It was called beta, and beta is like a mountaineering and ski touring thing that literally means just local rock climbing, also. It means like local information. What's the beta on this route? That's cool. It was a platform to connect backcountry ski touring guides, rock climbing guides, mountaineering guides to people who wanted to meet them and learn more about them and improve their skills. And also hired these guys to take them out from all around the world. And if anybody's listening who cares about that space, please go build the thing. I would be happy to share all the research that I've done, all the contacts that I made. Um, I don't have time to focus on it right now, but I I hope it exists at some point in the future because it's a cool thing.

Kevin Kerner: 6:08

Well, what fertile ground for this for the product that you're building now? Because that's gotta be a hyper niche audience. Yeah, it's a super, super high. You can't go to outside magazine as much, maybe. There are probably some, and maybe you can, but there's gotta be some really interesting targeting that needs to happen to hit that niche because it's so specific.

Brendan Norman: 6:28

It's a great example because previously some of the existing tools would categorize all of outside magazine in that space, and then try to run a guy an ad against mountain biking or kayaking. And those are really cool sports, but they're not as specific as like backcountry ski touring. Yeah. And the technology that we built would literally scan all of outside com's uh content and then classify it into ways that make it really easy to understand. You know, if you want to run an ad for kayaking, these are the specific URLs. Or if you want to run an ad for backcountry ski touring, these are the specific URLs. And the cool thing is that applies to anything pharmaceuticals, you know, PNG, like any any type of product or service that exists. We've built this infrastructure to be able to quickly find that content across the open web.

Kevin Kerner: 7:16

Yeah, and it I we a lot of our customers are in the tech space. And but it's interesting what's happening in B2B and tech right now is that a lot of B2B and tech companies, let's say you're a cybersecurity company, we might want to contextually advertise to you based on your own personal interests, in addition to the fact that you're into cybersecurity. And so there can be both consumer targeting, but to a business person and knowing what your business people might be interested in from a consumer perspective, or what you know, what the TV they watch, what magazines are where they're at online, can be used to target a very specific business message where they show up personally. So the target targeting world's getting much more interesting and complex in B2B these days. I was gonna ask you, so the probably the most the biggest conversation over the last few years, other than AI, one of the big conversations around targeting has been the cookie. And it seems like that's such an old term now because uh AI's kind of like sucked all the oxygen out of the room. Can you explain the problem of targeting now? I think there's talk of the cookie going away. I think Google has slowed that down some, but what's the what is the problem that faces advertisers in terms of targeting today?

Brendan Norman: 8:35

So it's a really good question to unpack in the sense that historically, you know, the ad tech ecosystem and the large players have been really good at figuring out how to solve a problem and then create a new problem at the same time and then deprecate their solution by having another problem. Yeah, 100%. So the cookie was originally kind of like the IDFA Google or Apple's identifier for mobile. These things were originally created to kind of have obfuscated identity tracking, you know, technology that kind of follows somebody around. And it was supposed to be privacy safe, but the more you get into it, the more you realize that like any one of these technologies, people figure out how to exploit it pretty quickly. And you know, the general narrative around like privacy is an ongoing discussion, I think, in the industry. So to your point about cookie deprecation, you know, Google's threatened third-party cookie deprecation, you know, so many times, and it's it's kind of an old conversation topic. I think every time that that happens, you know, the contextual vendors, like folks that are in a similar space, get a lot of traction and interest. And then it's kind of fairly died down because that targeting has been fairly limited to just high-level categorization and keyword targeting. And the industry goes back to more audience-based targeting and retargeting. One of the best things that Facebook does really well is it combines a lot of these different types of signals into one algorithm. And, you know, when you're running an ad, and this is again like my former employer, so I'm not trying to drive people. I know it works really well. Social channels, you know, meta-targeting is phenomenal. And it's pretty common that they command, you know, a pretty decent percentage of media spend because it does work very well. And it works well because they're not only targeting you, but they're targeting, including a lot of signal around, you know, kind of what you've done around the internet, what your interests are, in addition to the context of what you're looking at at that moment. And that's been kind of the one of the biggest missing pieces for marketing in the open web, I think, is providing that layer of knowing. Like I'm reading this very specific web page, you know, back to your point about financial or crypto. You know, if I'm reading about an Ethereum page, I should be able to run an ad if I'm a crypto exchange that wants to target ads for Ethereum to know this person also is reading this specific piece of content, and then to tailor that ad to that very specific article, that's just starting to come into existence right now in the ad tech uh ecosystem across the open web. So it's a it's a pretty exciting time to be in this space because the level of sophistication and the targeting, you know, and these different channels of contextual and audience are really getting better very quickly. And then there's some new technology that's allowing them to kind of work together more seamlessly that we're just kind of pioneering right now in real time.

Kevin Kerner: 11:28

Yeah, I would love to hear about that. So let's say it's contextual 1.0 is more general about let's say outdoor magazine. And we know that's you know, an outdoor magazine. It has certain types of personas that go to it. You can refine the type of people that are in that one category. But what's what does contextual 2.0 look like? And I imagine it involves some AI magic, right?

Brendan Norman: 11:50

Definitely does. And I might even argue that we're kind of in the contextual 3.0 right now. But the contextual 2.0, going back to one, you know, a lot of metadata and kind of manual tagging about a page. You know, outside.com puts a whole bunch of different keywords into the metadata for the page. Two is having more automated machine learning systems that you know are able to kind of read a page and then classify it into a bucket that says this is outdoor activities or this is outdoor sports with some basic keyword extraction, three or four keywords that they find. And then three is really reading the entire content of the page and attributing hundreds and hundreds of different vectors per page around saying, you know, it's talking about this specific ski brand, it's talking positively and negatively about it, it's talking about this specific uh you know skier positively and negatively, in addition to all the different categories, in addition to all the different keywords. So what it's doing is it's kind of creating a much deeper understanding page by page. And because of how that uh works on the back end through vector databases, it's much easier to just make it searchable. So you can kind of organize this content in a way that allows you to kind of search and then pinpoint, you know, using natural language. You know, I want to drive sales for black country uh for black crow skis, you know, to backcountry ski touring enthusiasts for people in Tahom. It's very easy to now go find exactly which what web pages exist across the web that you should be running that ad campaign against.

Kevin Kerner: 13:31

Wow, that's wild. And is the is it possible that the AI can read both the front end of the database in addition to anything that might be added on the back end metadata that you know might be schema data, that type of things, does all that factor into it too? Does images, do images factor into it as well?

Brendan Norman: 13:48

It all factors in. And um, you know, the intention is making sure that you know the right ad shows up against the right piece of content. And knowing that, you know, there are for the same general direction of travel and content, there are really good pages that are high quality. And there's a lot of pages that are made for advertising, there's a lot of spammy pages, yeah. A lot of pages with ads that are kind of below the fold or pop on top of other ads. So to your point, also understanding the quality of ads, there's a lot of other technical signals that are happening underneath just the actual blurb of text. You kind of need to take all these things into consideration when you're thinking about classification and also when you're thinking about targeting. That's amazing.

Kevin Kerner: 14:34

So I had a Dr. Augustine Phu on the podcast, I don't know, a few months ago, and he's the uh sort of ad spam guy. He's got really good research to prove that a lot of what's happening in ads is going to is is our bot pages and bots themselves. And then uh does this have an impact on if you could classify a page, could you classify a page as something that's that's low quality that you wouldn't want to advertise on and it would keep you away from? That's gotta be, from what he would say, that's gotta be a significant portion of the web. And it is. So in some ways, this new um contextual analysis would be able to fuel that it's just another level of a page that's safe, something that's of value, right? Just gives you a higher level of confidence that it's of value. Yeah, that's really wild. Yeah. Now, the quest this all sounds great. And I know this is what is being built inside the classify engine. One of the, I think the first time we talked, that the one of the biggest questions in my mind, and how do you solve for it seems it's it's difficult. It's like, how do you get this amazing vector database of all this great information? How do you get it into the places where people actually the plumbing to actually make it target to be to be targeted? How do you solve for that problem?

Brendan Norman: 16:26

It's really good timing for that question because you know, advertising technology is is fairly complex and it's it's very complex by design because it's a big legacy too. There's so much legacy in it. There's a lot of legacy in there. So, you know, it's kind of like you know, the the plumbing in New York City is old, right? It's like it's very old. These pipes have been around a long time, and I guess the analogy is the city has to go in and constantly like fix things, right? Fix old iron pipes, cobalt pipes, whatever they are, cobalt pipes. And everywhere there's kind of leaks going on, and people, you know, people tolerate it. Like it's fine, it works, it works fine. But if you can build a new something, brand new something from scratch right now using modern technology, it is just by default going to be much more efficient. It's going to have a lot less waste. You know, you can measure and track it more. And we're part of a group called uh the um advertising content protocol. Sorry, can we pause? Yeah, yeah. Sorry. It's been a long day. This is like I think the third podcast I've done today.

Kevin Kerner: 17:27

Oh, you should be warmed up by now. I should be, but I'm also tired. Yeah, but you're losing it.

Brendan Norman: 17:31

I'm sorry. Um I'll go back to that. Okay. We're part of a working group called the Ad Context Protocol. And it's basically built on top of uh Anthropic's model context protocol, MCP. But AdCP is a new layer that's effectively going to be redesigning all these pipes that allow these different systems to talk to each other. Data providers, signal providers, creative providers, selling agents. Everybody who's built these really interesting back-end systems will now be able to kind of deploy those systems uh in a much more standardized, secure, safe way, um, where they'll be able to integrate it directly into an agency's intelligence platform or a supply-side platform, a demand side platform. And the protocol is cool because it's standardized. And instead of having to design a custom API endpoint for every single connection, um, there's basically one. So it's like, you know, you call the same phone number over and over again. And the goal and the vision of this whole movement right now is to basically standardize how this data gets transferred back and forth and to make sure that everything is going to be a lot more efficient, a lot more secure, privacy-safe, trafficable, measurable, um, and design a system that's really built, you know, with the advertiser and then the publisher in mind.

Kevin Kerner: 18:51

Who's in the what type of companies are in the can you name some of the companies in the consortium when we know some of that? Definitely it's cool.

Brendan Norman: 19:00

Some big companies. Um, Brian O'Kelly, it was his original idea. So he's now at scope three. But he was one of the founding members of Write Media and also AppNexus. Um arguably the godfather of programmatic advertising and then prebid, created this, you know, giant ecosystem to begin with and is now kind of solving some of the inefficiencies that got created along the way over the last decade or so. So Scope 3, PubMatic, who's a large supply-side platform. Yeah. Yeah. Um, there's a couple of big agencies that are in this space, and then a couple other data providers like us, Optical, Swivel.

Kevin Kerner: 19:38

Wow. That's so cool because um, your use case is this vector database of contextualized web pages. Someone else might have an audience sort of network. There's so many ways you can combine this data in interesting ways, but getting it in getting it to meet. That's what I love so about MCP. It's so easy to use because it takes the difficulty of getting into data back into an LLM, reading the data directly. It's just so easy. There's so many use cases for it. It's ridiculous. And in terms of like a before and after example of buying an ad in this new world or using this ad context protocol, can you give us simple, some simple use cases of how it might work in the future state?

Brendan Norman: 20:20

Very much. And it's a lot about agentic experience. So agents talking to other agents. And what that means is, for example, data provider who's doing creative or contextual data wants to talk directly to an agent that's living in an agency's intelligence platform. Those two agents, they can kind of set in some basic logic and have those two agents go out and kind of negotiate on a price. Or they can figure out what's the best ad creative to deliver for this specific campaign. Or they can figure out what's the best audience to use to target against for this specific campaign without necessarily having to have a human involved and making a lot of these manual decisions. So part of that is training these agents, making sure that they're starting to work together in a good way. And I think there will be a lot of human interaction that for the same way that when you go to Chat GPT or you go to Perplexity or Cloud, sometimes it doesn't give you back, or a lot of the time it doesn't give you back the best answer in that specific prompt. Then you kind of have to fine-tune and revise and go back and forth. We'll continue to see a lot of that. And you know, if I'm somebody who's in ad operations today or who's in campaign management, you know, who's building these campaigns, working in an agency, there's definitely some concern around, you know, like, wow, is this my job at risk? And I I think it's the opposite. I think a lot of this new technology just enables you to go back to being a lot more strategic. You know, you're kind of thinking about it in a way that allows you to just get a lot more done a lot faster. Yeah. Um, so that's the goal. I mean, we're figuring a lot of this stuff out too as we're building it. Like, what do these policies look like?

Kevin Kerner: 21:58

So cool. So cool. I have a question about another use case. I don't know if it I'm thinking about this the right way, but you've let's say you've got Classify with these contextual data sets in it. You've got another, you've got some other tools. Is there a P or will there is there a other middleware software or LLM where I can chat with both of those things at the same time? Will there be a um or is there in the vision here the ability to have a chat screen that digs into these things using um ad context protocol?

Brendan Norman: 22:32

Very much. Um and you're already starting to see a lot of that in the native and/or kind of homegrown ad agency intelligence platforms, and some of the large supply side platforms already have a chat agents that are connecting back and forth with these back-end systems. So I think what you'll what you'll pretty quickly see is that start to trickle down outside of the larger folks who have been doing it for a while or just starting to do it now. I think you'll start to see that a lot more accessible across um like all these different platforms.

Kevin Kerner: 23:02

Wow. In addition to just interrogating interrogating the data and coming up with a strategy, is it feasible that you could also purchase or buy or bid directly using these systems?

Brendan Norman: 23:15

Big time. I think one of the first agent uh protocols that will launch and it's already started to test are these sales agents. And like they're automated systems that kind of go in and try to figure out bid strategy, they try to figure out how do we get the best campaign at the best value. So you you basically will be setting some different parameters, and then over time I get all learn and kind of fine-tune it. But effectively, it's like having this very powerful uh system that kind of automates a lot of these tasks for you that previously were very manual.

Kevin Kerner: 23:49

Wow. That's up. That was really cool. Is that most of that um innovation on the on that side? Um is most of the innovation in innovation that's happening in the actual um strategy side, like purchasing stuff, is that being driven by the agency platforms primarily, or is there some leader that's actually the interface that you know type into to access all these other systems? Or who's the who is the what is the entity driving that the fastest?

Brendan Norman: 24:19

Just curious. It's kind of all the above, right? Like if you're somebody like Classify that wants to make sure that the technology that we've built is accessible everywhere, it's deployable everywhere, folks can use it. We're on one side and we're helping to set a lot of these protocols and policies and just evangelize the thing that we've built. And then if you're on the other side of that equation, the buy side in an agency, you know, obviously like you want to be more efficient and effective at the tools that you have. And going back and forth between a whole bunch of different platforms using a whole bunch of different um services and dashboards is a lot of work. So they're working to streamline their own workloads quickly. So I would say kind of everybody is invested in this uh the success of helping to make sure that these new agentic experiences just work really well. And they're actually providing value, but you know, they're adhering to the things that you're setting up, you're asking them to do.

Kevin Kerner: 25:14

Yeah, super cool. Yeah, and I guess it depends on your lens. If you're an agency, you're gonna want to do some certain things. If you're you know a demand platform, you're gonna want to do other things. So that makes total sense. Uh let's talk a little bit about your go-to-market because I'm really uh and by the way, I was gonna mention people, if you want to get a little preview how at least the classify stuff works, you should go to is it classify tryclassify.com. Try classify. Yeah, yeah, it's pretty cool. You you can actually you have on your main homepage the ability to like type in a query and it will it will list out the very specific pages that are targeted for whatever your query is. It's really cool. Um it seems like you're you're pretty early in the launch of some of this stuff, but what is let's talk about your go-to-market a bit. How's that been going? Like what's what is it's as a company that's in the AI space, there's a lot of competition. What's working, what's not working in the go to market for you guys? A big part of it is is is this.

Brendan Norman: 26:08

It's chatting with folks to kind of help evangelize you know, this next version of contextual. And we have effectively two products at the moment. One of them is on the classification side. So we're working with the publishers and supply side platforms to just help give them better insights into what's going on inside their content at large scale. Hence the name classify. And then the second piece to that is the demand side. So it's working with agencies, brands, and kind of anybody that's on the buy side, helping to make sure that when they're running a campaign, it's targeting against the very specific pages that they should be running an ad against. So go-to-market strategy is kind of, you know, part of it is just chatting about it in the industry, part of it's um, you know, being a lot of the events, and also part of it's just leveraging network and doing a lot of direct sales. But with that go-to-market kind of flywheel that we've been generating some really cool publicity and kind of interest, it's spurred a lot of inbound uh requests. So it's been fun to like respond to these folks and and help educate the market. And we've got a lot of cool clients who have just come to us and they've heard about what we're doing and they want to start testing and then they see the results. So, you know, at the end of the day, like the product is very helpful to making sure that an ad campaign shows up against the best content.

Kevin Kerner: 27:27

Yeah.

Brendan Norman: 27:28

And when you're increasing somebody's, you know, whether it's their their click-through rate or another metric, when you're doing that, you know, two to three times what it was previously doing before they added this in, the results have been really positive across the board. And that kind of flywheel in terms of just advertiser value creation has been really strong.

Kevin Kerner: 27:46

Yeah, I would guess too, you probably gotta you're probably working with relatively sophisticated types of companies on both sides of those product equations. Because if you're if you're um like this isn't something I don't think that like a small advertiser would be so interested in, would it? It's more you're right now you're working with fairly sophisticated buyers of this type of solution because you have to integrate it into what you're actually your own solution, right?

Brendan Norman: 28:12

You definitely do. I mean, fast forward to the future, I think it'll be really cool to give the same level of targeting sophistication to anybody, right? And if you go back to Facebook, right? You could run an ad campaign for $20 a month. You you can't do that in the trade desk. You have to spend a lot of money in the trade desk to get access to these tools. And you know, I think the ad tech ecosystem ha hasn't really courted or figured out a way to support the rest of the SMB channels that want to spend less money or they have less money to spend. And I think the big miss there is that when you're really helping to deliver value, you're helping those businesses scale. And if you're able to work with somebody that wants to spend 50 bucks a month, you know, 100 bucks a month, and that's actually delivering value, that's just gonna scale over time. So what's kind of cool about these new pipes, this new ad CP ecosystem is I would also envision that there will be folks who are able to help some of the SMB's like action on running campaigns at a much smaller level or at a much more modest kind of entry-level ad campaign, but consciously optimistic that um that this whole ecosystem will change over the next five years and make it a lot more democratized and kind of focus on how do we find the rest of these folks who want to run ad campaigns but just haven't been able to.

Kevin Kerner: 29:34

Yeah, I totally agree. I mean, with this type of technology, there'll be a bunch of innovation on the small advertiser side. There has to be. Like if you could give that, if it's if it's that easy to access the things like what you're building, there'll be innovative companies that go, oh, I want to do that, but only for small advertisers. Like you want to spend 20 bucks a month or 50 bucks a month or whatever on and I can do that. It's just it's really cool. There'd be a great mission to have to take the trade desk level. complexity and boil it down for some small mom and pop. Super cool.

Brendan Norman: 30:06

I think on Google and then on Fed Facebook or Meta, you know, each one has something like roughly 15 million unique advertisers. Yeah. A lot of those campaigns are are are very modest campaigns and then the rest of the open web, the programmatic web through DSPs, you know, it's a much smaller list that are all spending a lot more money, but there's still a lot of other advertisers you can help support if whoever builds those tools and that allows the smaller advertisers to connect into these pipes. Or maybe they'll just redesign new pipes entirely. We'll see what happens.

Kevin Kerner: 30:40

Yeah, wow. So last question uh and then I'll get into AI roulette here, but what's the what do you think a marketer should be doing today to prepare for all this shift that's going on?

Brendan Norman: 30:50

I think it should it's staying up to date, you know, with all these different changes that are happening. Because ad tech was kind of boring for quite a while. You know, and it's been kind of the same you know remarketing, you know, it's kind of like in the ski industry. A lot of times they will literally release the exact same ski with a new top sheet or a new graphic on top. You know, they I video they call it but they call it a new graphic technology. NGT. It doesn't really change anything about the performance but it looks cool so you want to get them the same kind of thing has been happening in ad tech for quite a long time. And we're all of a sudden just reintroducing like this brand new way that all this stuff can connect. So the whole ecosystem is changing pretty rapidly and following along the conversation, getting involved in the conversation directly I think is the best way to help influence you know what gets built. But every day I mean I learn about really cool new companies that are doing everything like native creative generation or new podcasting technologies. And there's so many cool innovators in the space right now that will now have access to kind of play in the bigger uh the bigger play box.

Kevin Kerner: 32:01

Yeah it's crazy. You know that the in the B2B and tech, this a similar trending technology for probably 10 years now is intent data. The intent data space has been very static. People talk about a new intent source, but it's all kind of the same stuff. So there are there's these technologies that will eventually like like what what you're building at Classify where someone will rethink it in a completely different way and AI will really help. The intent space in B2B has been AI based for a while but it's still this there's no one's really come out with anything new in intent that I know about for I don't know at least five to ten years now. So it's super interesting. Okay let me um let me ask you I'm going to I do this thing called AI roulette on the show and I'm gonna load I've loaded your profile and a little bit about what we're gonna talk about today into perplexity and it's gonna give me the best question of the day which isn't hard when I'm asking the questions. So let me put in here and I'll click enter okay here we go. That's a little bit long one okay um okay Brendan the ad industry loves to stay just to say contextual targeting is the ethical alternative to surveillance marketing but AI now reads tone, sentiment and even emotion better than most humans. If algorithms can psychoanalyze articles, videos, and our reactions in real time, aren't we just building a smarter kind of surveillance? One that manipulates context instead of people at what point does privacy safe advertising become privacy theater? Huh so it's maybe I think it's basically trying to say well contextual should be pri it's more private but is it really private?

Brendan Norman: 33:44

I mean I I would I would really push back against that AI generated answer.

Kevin Kerner: 33:48

Maybe there's some ulterior motor from the AI that doesn't I don't know this is the first this is the this is I put it I put the extra word spicy into this one so maybe it could uh maybe it could a hacked it a little so what do you think a little bit I think the short answer is I mean back to privacy like what does privacy mean?

Brendan Norman: 34:07

It means that it's not looking at Kevin and it's not understanding you know what you're actually browsing, what you're clicking on, all those things. Like those are private things that you do. Yeah. The fact that a certain ad has the ability to show up against you know a specific type of content I think is completely disconnected from any concept of privacy at all. Yeah. You know content adjacency you know is a totally separated thing. You know and it's actually in a lot of ways like we we did a lot of research on this um a while back but it can be not only better for the advertiser and drive a better outcome but it can actually be a better user experience. So if you are reading an ad about football and you see something that's totally irrelevant, you may or may not engage with that because you've been already browsing, looking at that, been following you around using some retargeting. But more likely than not, if you see an ad about something that's like really related to the article that you're reading or the podcast that you're listening to or the video that you're watching, you are much more likely to engage with that because you're already in the mindset. So there's also some psychology you know around how relevant it is and those are other considerations to make. But the AI generated question I I think is I kind of kind of missed the mark.

Kevin Kerner: 35:23

Yeah it definitely did. It also is sort of inferring that the AI may cook the books in some way where it's you're just trying to um you're you're trying to uh create you're trying to create a con a context that's not there. So it's it's kind of manipulating context which the AI doesn't really isn't going to want to do like when it's when it looks at your page it's gonna want to accurately accurately describe that describe that page as accuracy as possible. So it's kind of a it's it's kind of oh go ahead.

Brendan Norman: 35:55

I'm just gonna say what is interesting is that I joined on the IAB Tech Labs working group for content monetization for AI. And basically what that is is right now we are figuring out, we're chatting about and there's a lot of cool folks in this group from like the big AI companies, from the big social companies, from publishers, from advertisers, from other folks like us in the tech space. And there's a really interesting discussion around how do we help content creators, publishers that are putting out interesting content protect more of that from you know we all saw the the recent lawsuit that just happened I think it was Perplexity and suing uh or Reddit suing perplexity for scraping without you know giving them any monetary value. And it's a really important question to be asking through all of this right now. You know, it's how do we build the right guardrails in place that help publishers protect the content that they're creating in whatever format that that that shows up as. But it's a really important consideration because one, there's going to be a lot of AI slop that exists. We're already seeing it with Sora and ChatGPT and some other generative AI tools. And at the same time I I think that that's going to drive towards really high quality content. So anybody who's an expert in in making you know a really nice video or writing an article you know great writers, great musicians, like whatever it is, all the same doesn't necessarily diminish their quality. I think it elevates that the same kind of a thing happened when Instagram came out I had a lot of friends who were professional photographers who were terrified that they were going to get buried under everybody now having an iPhone in their pocket and being able to post it on Instagram.

Kevin Kerner: 37:38

Yeah.

Brendan Norman: 37:38

But it kind of did the opposite it elevated the really good ones and it kind of retrained us to look at okay there's a lot of pictures now if I'm gonna spend time I only have 24 hours in a day I want to spend most of that time looking at higher quality stuff. And I think the same thing will happen moving forward with content creators with publishers we're just gonna have to figure out kind of different channels and how to how to navigate it better and how to structure it in ways that the economics are more favorable towards the content creators themselves. But I'm I don't know I continue to be cautiously optimistic.

Kevin Kerner: 38:15

Yeah me too well I love what you guys are building because it if anything it the ability to classify good you know what type of content is on a page what's good and what's not good what might be safe and not safe it's just it's just a really smart and needed technology. So I applaud you guys for what you're doing and I'm a fan and uh I'll keep watching what's happening. I do uh hope that everyone that's listening to this goes to try classify.com and checks out the just the simple little targeting that you have on the homepage it's really cool. Okay it immediately gets across what you guys are trying to do. It's really super cool and a very easy way to see it. So Brendan I want I want people to be able to reach out to you and ask questions. If they want to get a hold of you what's the best way to stay in touch with you guys?

Brendan Norman: 38:55

Yeah either shoot me an email or connect on LinkedIn and and shoot me a message.

Kevin Kerner: 39:00

Cool. That's great. And I'll put your uh details in the podcast uh description so I really appreciate it Brendan I know you've been traveling a ton so I really appreciate you uh joining me here and I wish you the best of best of luck. Thank you Kevin. Okay talk see you and see ya

‍