How Scope3 Is Reducing Digital Ads’ Carbon Footprint

We operate in a really complex global business like advertising is massive and global and there's a lot of integration complexity. So we work with a ton of different partners and a lot of those integrations while people may publish an API they don't always work and there's so many weird special cases about every one of them that a lot of our people time is actually spent clarifying things and working through these weird gotchas. And so if you look at how many Slack channels we're in with our partners like so much of the conversation is us negotiating. Hi, my name is Dmitri Bonichi and I'm a content creator, agency owner, and AI enthusiast. You're listening to the AI agents podcast brought to you by Jot Form and featuring our very own CEO and founder, Idkin Tank. This is the show where artificial intelligence meets innovation, productivity, and

the tools shaping the future of work. Enjoy the show. Hey there, and welcome to another episode of the AI Agents Podcast. In this episode, we're interviewing Brian Okelly from Scope 3. Hi, Brian. How are you doing? I'm doing great. Thanks for having me. >> Yeah, I really appreciate you being on. I'm I'm excited to chat about uh media and AI. You know, I don't I don't think we've really had an episode like that with a guest. So, if you could please give us a brief introduction to who you are, um how you got to where you're at and and what you guys are doing over there at Scope 3. >> Sure. So about 20 years ago, I happened to invent this kind of crazy idea that we should have a real-time auction for every ad impression on the internet. So in your browser, you know,

in the blink of an eye, ads would just be auctioned to the highest bidder. And this idea which was called the ad exchange and is now called programmatic advertising uh ended up being on every single web page in the world on Netflix on you know streaming TV on billboards everywhere you go there's these real-time ad auctions happening and it's all because of this idea I had 20 years ago and uh over the last 20 years I I sold a couple of companies uh, you know, have had the chance to really be deeply embedded in the digital advertising world. You know, I've testified in front of Congress. I've been to Davos. I've, you know, had to be part of an antitrust suit against Google. I' I've been all over the place. Yeah. And yet, I've never been more excited or kind of in awe of something

as I've been over the past, let's say, 12 months at the evolution of what AI can do. And it's pretty obviously going to completely reimagine what we all think of as the internet. Like that concept may be obsolete at this point because so much of our lives is digital. And if you think back to what I invented 20 years ago, really web pages and websites, which were the only thing that was the internet then maybe gone, you know, like like when you see that you go to Google and you search for something and they just tell you the answer, why do you need to go to a website? I'm not saying it all completely goes away. But that means that advertising, this $120 billion a year, you know, sort of programmatic advertising industry has to evolve and change. And even things like like commerce, like

there's a fight going on between Amazon and Google right now where Amazon's pulling all of its ads off Google. What they just realized is that if you're searching on Gemini and talking to your agent like, "Hey, you know, what should I buy for, I don't know, a microphone?" Guess what? You don't have to click through to Amazon. Google can just take that whole transaction and finish it without you ever visiting the store. What does that mean for advertising? You know, is is it just mean that Google gets a affiliate fee? Most of Amazon's profits come from advertising. People don't know that. But if no one goes to Amazon, well, how do they make money off advertising? What's the point of a sponsored link on Amazon if no one ever goes there? So I believe that the advertising industry and and media at large is just

completely disrupting itself. And I don't think anybody including me knows what it means. But I do think that this idea of leveraging AI agents or or thinking about agents at the core of how we as humans interact with this digital world, whatever we call the post internet thing. um we're going to learn how to use these agents to do so much of what we currently are doing um through these legacy systems like the ones I built. So, it's a long way of saying uh scope 3 is sort of spiraling into this new idea of what if agents were at the heart of advertising and how do we build for a future we don't fully understand? >> H so yeah, there's a lot of different things that I'm I've noticed recently that are um that feel like they're changing, you know, with the advent of AI.

And I'm curious to you first and foremost, like at scope 3, what are you um actively uh doing and seeing that's uh maybe in the last uh few months? Because obviously there's there's there's a lot of things that are changing in uh the space. I feel like once a month I find out that there's something that I was behind on even though I host a podcast about this. >> Well, I'll give you an example of something that we've we've built. So for the history of the web at least, there's been this idea that it would be really cool if an advertiser could, you know, look at every single web page and figure out which ones would make sense to advertise on. So, let's say that you're Nike and you've got a brand new, I don't know, marathon shoe and you're like, "Huh, like I want

to go advertise on every page about shoes, but not every shoe, only like running shoes, but I also want to advertise to like hardcore athletes who might be interested in running a marathon or maybe some of those like weekend warriors who like, you know, someday aspire to run a marathon. In theory, you could do that 20 years ago, but it was very hard. only Google really had the ability to crawl the internet and understand all the content. But today, if you think about what all these big AI companies are doing, they're crawling the whole web constantly and pulling this information into train these these AI models, these large language models. I thought to myself, well, if they've already been trained on the whole internet, couldn't we just ask them and say, is this page good for Nike? Is that page good for Nike? And so

we build a product that basically does that that you write a prompt that describes your product and your target customer. You tell the story of who you're trying to reach as a brand. And then we basically go out, we copy the content of the of the web page. We put that prompt on top and say, "Does this content match the prompt?" And what we found is it's amazingly good. like like far better than anything we've ever seen before at finding not just the obvious places that Nike should advertise that shoe, but some of the less obvious ones where, as an example, with that prompt I mentioned, it's finding pages about people who are very determined and resilient and who are thinking about recovery. And because those are things that you do when you're training for a marathon, you better be dedicated and resilient. And your

body's going to hurt a lot. And so it's finding all these really interesting connection points. And so the power of this for a brand is that you can literally type in just like you would to chat GPT and say this is the kind of thing I'm trying to do. And it will find it everywhere. And we're about to roll this out for podcasts. So for instance, let's say I'm I don't know OpenAI launching an agent and I'm like where could I find people interested in agents? What if I could just type in and say here are the kind of people I'm looking for and and here's my AI sound, you know, you know, and it goes out and it's gonna find this episode of this podcast because it's like that is the people that you need to go reach. That's the technology we built and

it's not like anything we've ever seen before. And that's just one use case for how agents make the entire concept of AI change dramatically. That's very interesting. Yeah. I mean, so I part of my background prior to getting into the content space was more in the um paid media side of things. So like uh paid search and social and whatnot. So I find I find it's interesting you're kind of um I don't know if you are you familiar with that space much like Google ads, Facebook ads, whatnot. Okay. Yeah. It almost seems like you're kind of remember do you I remember early on the concept of uh paid ads in in the sense of what Google was doing was interesting like and as the keyword types managed to change like I know in the last I want to say three years they added some additions

to where broad match keywords started to be able to take into uh context like page um um word syntax versus like what the search term was in order for it to like uh align there. So, it's it's very interesting to hear this cuz it's all kind of going in this interesting huge shift in advertising uh like paid advertising and like now what it seems like with with media, you know, being able to define those things. So, I guess um my next question to you would be how do you feel like there's going to be more changes uh moving forward? um and how what are some of the big examples of where you feel like media is going to be disrupted even more by um AI uh in general first and then maybe we can get into some like agentic questions after that. >> Yeah. Well,

I think from a just a broad media perspective, like if you're if you think about how I used to search to your point, I type in keywords and now I kind of ask questions of, you know, Claude or ChatgPT or whatever and they do the searching for me and they go get the information I need and get it back to me. What we're seeing, and this is public information, is a lot of big publishing companies are seeing huge drops in traffic. partially because of zeroclick you know AI summary results. Yeah. And partially because people are using you know these these chat bots to do a lot of their you know engagements. So yeah so it's really good interesting question. What's the business model for content? Now, if you're a creator on, I don't know, YouTube or Tik Tok or whatever, people are looking to be

entertained and there's an entertainment aspect of that content. And, you know, I think there's there's a clear business model there that if I'm watching Mr. Beast and you know, like YouTube can show four or five commercials while I watch. That's not dissimilar to me watching a football game and seeing you know commercials it feels like every two minutes and that has been a stable videocentric advertising experience for 70 years. I don't think AI is disrupting particularly but if you think about the newspaper and I I'll explain that to your listeners because they may not know what it was. There used to be a physical piece of paper like paper that came to your house in the morning and you would read it. I know that sounds very archaic. Actually, before that, they brought the milk to your house in the morning. There was this early

version of Door Dash called the milkman or the paper boy, you know. Um, but anyway, now we have, you know, kids don't even believe me. You know, we used to have phones with wires. Um, this is like in my lifetime, too, right? So now we're like I wake up and I turn on FT, Wall Street Journal, New York Times, go to hacker news. Like I'm not actually interacting in those ways. And so when you ask about the future of media, I think there's a large proportion of young people who get their quote unquote news purely from social media. And therefore the idea of a media company as being something that has journalists, you know, we may as a society want that very badly at some point because we've seen so much disinformation, misinformation, apathy, you know, confusion. Our politics are being twisted, I think, by

a lot of forces we don't understand algorithmically. But the bottom line is, as much as we may not like it, the reality is wherever eyeballs and ears go is where the advertising goes. And if you're a large advertiser, Nike, Craft Hinds, Nestle, Unilver, PNG, whoever, you got to go where your customers are. And so, and if you're a media company, you want to put content where those eyeballs are. like both sides, the the paid media side and the content production side are in the same search for consumer interest. And so to me, that future is going to be driven as much by my 15-year-old and where she decides to spend her time online in or whatever online means as it's going to be determined by any, you know, tech anybody, right? if she wants to chat with a AI bot on her phone that she

finds really funny and interesting and answers all of her questions. Well, is that media? I I mean, I think it is. And I have no idea what that means, but it doesn't seem that crazy that we could all have that experience in the very near term and that becomes our dominant form of of content consumption. >> Yeah. So, I mean, I just had some some thoughts uh based off of the conversation that you just like kind of brought forward. I mean, with you know, you mentioned New York Times um different uh news outlets that are directly on your phone. What I find really intriguing is that zeroclick search so to speak or zeroclick um information that's coming forward. There seems to be this trend where large mass output of content that then gets summarized into short snippets as kind of the the move forward. I'd

imagine you already see this on your summaries and notifications. I don't really see it as much because I don't use um I don't have notifications on my phone and I don't use like the news outlets on my phone as much. Um, I'm more like a video person. But what I'm noticing with Google search, when you see the AI generated summary, uh, chatting with AI, whether that be Google or or sorry, Gemini or Open AI's jetp or Claude, which had the search function come out a couple months ago, and even now on Apple devices, do you have an iPhone or Apple devices in general? Do you do you know I don't know if you're getting them yet, but even on just I'm using the beta for my iPad, my phone, they have summaries now of what you're receiving. Have you seen these? >> Yeah, I turned

them off. I found them really annoying. >> They are actually pretty annoying. Uh yeah, I'll give you that. They're not as effective as I'd like them to be. Um but that's kind of where we're moving and I feel like and evidently where it will probably end up being better and working okay. Um, as somebody who's working in media, what is your general thought and feeling about how we're kind of moving towards this? I don't want to call it shrinkified. I actually read an article about this um the other week, but like shrinkifying the or not shrink um I forget the term he used, but it was something between summarizing and shrinking the information um into like short sn maybe summarize like the problems with summarization. Right now we're not focused on like some people are into long form stuff obviously how people listen to this

podcast but what are your thoughts on how media is adjusting to this sort of summarized version rather than the in-depth um nature of probably what it used to be because uh you couldn't really summarize things that easily before because you have to take a long thing and really study it and then write it out. You know what I like AI has just immediately advented this summ summarization culture so to speak. Does that make sense? >> Well, I mean look like Cliff's notes, you know, were around when I was in college, you know, and fundamentally like I can tell you the entire plot of war and peace, right? There's a war and there's some peace. It's, you know, that's like the super cliffnotes version, you know? Thank you. And I think the the challenge is that there's a like I you could summarize our conversation and

you could say, "Oh, you know, Demetri and Brian had an amazing conversation. It was brilliant. You know, they talked about AI agents." You get nothing from that. If you sit through this entire conversation, you might actually get a few things that are valuable. And the way our brains work, I think there's a certain immersion into content, you know, and if you think about an article about, you know, there's a big shooting in New York yesterday. It was horrible, you know, and you know, I I I found myself seeing the headline which was like five people dead and I was like, "This is awful." You start reading it and these little tidbits start standing out and of course I've been to this building. I've been to that building recently. You know, I know people who work in that building and I'm just going to myself like

this is absolutely terrifying and I had an emotional connection to the story. Obviously, I live in New York and I know there's reasons, but I feel like there's something about storytelling that I am very obsessed with, which is this idea that for 10,000 years, people, humans, have been sitting around campfires telling each other stories. And that there's such a deep cultural collective unconscious where if I give you a word, I can trigger all kinds of fascinating connections in your brain. I'm just going to give you one. Thor. Okay, you just thought Nor Smith, but you also thought Marvel comics, you know, right? But but you there's all these different things. But think about all the people in the world who'd have a slightly different version of Thor, but I knew you'd say Marvel. And I don't have to go tell you the whole story. I

don't have to tell you he's Norse, has big ass muscles, like you know, he's he's sort of funny and goofy, right? like I can go down this whole path. But the reality is you don't you didn't think nor smith but at some level you know you've got Asgard you've got all these things you know his brother's Loki you know and you've got all this stuff that's like the Marvel myth is built on top of the Norse myth and that's built on top of other myths that we don't even know about or remember. And so there's something about that when you say summarization. I also think though that like the power of myth and these almost like keywords in our brains like Thor, right? There's so many things that come up for us that there's actually some value in in in saying like can we find

those resonant ways to communicate with few words and think about poetry. Poetry is elegant summarization in ways that that fire off so many neurons in our brains. So I would just say that the challenge here is great writing is cutting out waste and as a business person I will tell you that many of the emails and notion pages and things I get have not been turned into poetry. You know there is a lot of writing for the sake of writing and I think AI can do a lot of good in getting rid of a lot of that baloney. At the same time there's a real risk I think you're bringing up which is what if we summarize away the poetry? What if we summarize summarize away those nuggets that make our brain think and learn and grow because you know at some point we we

squish away the parts that you know let us learn and and connect. So I I'm broadly with you that it's not great but I'm also lazy and I don't want to sit through a bunch of baloney that someone sent me. Yeah, >> I respect the laziness. Yeah. If it's not if it's not good you know. Fair enough. Yeah, that's that's that's that's a no that's a good point. Um I guess where are um some of the interesting you know media media's got some other we're seem to be talking probably more textoriented right now. Uh just to talk a little bit more about what you guys are doing at scope 3. Um could you touch a little bit more on uh you know where you're seeing adjustments in not only like written media but like visual media as well um through Agentic AI. >> Yeah. I

mean isn't advertising platform broadly? You know we think a lot about you know for us like all these different media are just different places that your eyes or ears or both are connecting with content right? It's it's there's a billboard that's just ads, you know. Um, which billboard do you want to show an ad on? And if you think about like the olden days, some poor human had to drive around the countryside in a pickup and look at all these different, you know, billboards by the highway and say, "Well, I want that one and that one, but that one's not in a great position, right?" And now we have digital out of home, digital billboards. And in theory, you know, we have things like, you know, those Google Street Maps that you can see everything on your computer. We shouldn't need that. But there's still

a basic question of like, how do you analyze traffic data? How do you know that on Interstate 5 going up Oregon from Eugene to Portland, you know, that no one drives on Sunday mornings, but like tons of people drive on Sunday afternoon. And if you could figure that out and realize that a lot of those folks, you know, are hung over from going to an Oregon Duck game and you could get your perfect demo if you knew that, well, well, now you could figure out these really interesting connections of where to show the perfect ad for, you know, happy Duck fans versus sad duck fans. AI can figure that out, but you have to program an agent to understand all those things. And so to me, it's not about video versus text versus billboards versus phones. It's really about what's that sort of core insight

that you're trying to to to create or or exploit as an advertiser. And how do you go do that when there's so many different surfaces and so many different places to advertise and so many disparate data sets and so many platforms? It's just the world is super fractal, right? There's just complexity all the way down. So to me, if I could just write a prompt and be like, "Listen, here's what I'm trying to do. Go figure it out." And I could give a bunch of this information and maybe my login or whatever to an agent and have it kind of just work. Well, that would be amazing. And I think that would be good for the world because right now you get that with Facebook, you get that with Google. You don't really get that anywhere else. And so if I'm a lazy advertiser, which

you know, I mean, those platforms are super hard to use. I can't imagine any small business being particularly successful in most advertising systems that I've seen except YouTube where you just click boost and it works. Like to me, that's what I want. Why can't you just make it work? And I think that's what agents agents or agentic AI can do for us, right? is they can say, "Let me take care of the complexity for you. I got this." And that to me is is transformational. It's not just advertising. It's just the field that I know best. But I think it's incredibly exciting if we could have that's what we want. That's what I think we think ad agencies are, right? They're supposed to be my buddy who knows everything about a particular domain who makes it work for me. And if I could build the

best domainspecific agent in the world for advertising, and who better than me because I invented most of this tech, right, that should be an incredibly valuable, powerful tool, especially if you build it with all the lessons we've learned from the last 20 years around things like brand safety. You don't want to be on crappy content about sketchy topics, you know, about some of the games and manipulations that we see. you probably don't want to advertise to bots except to good AI bots, you know, but when you're talking to an AI bot, you might want to use text, you know, it might be like the LLMs.txt version of an ad because, you know, why spend tokens, you know, for the AI to try to like handle your crazy ad? Just be like, "Hey, you should buy my car. It's awesome. It's got doors." You know, that's

my attempt at an agent ad. So, I guess I'm just saying if you just put all those pieces together, that's what we're trying to do is is it's not an easy button because it's not ever going to be I know Mark Zuckerberg says he's going to make it all magically work. I I think reality is that advertisers who are just businesses, right, want to have some influence and and learn from the process and contribute knowledge, but they don't want to get mired into all the crazy details. And I think we're, you know, we're going to unlock a huge amount of creativity if we could make all of that complexity fade to the background. >> Okay. So, how large uh is your company? Like how many people are working at um scope 3? >> We're about 100. >> Okay. How has this impacted the the workforce?

Like um from like a staffing perspective, future plans, that kind of thing. And how uh how do you well well that was a kind of a double question my apologies. how has it impacted, you know, your workforce that currently exists, right? Like, uh, people, their their workload, that kind of stuff, like the advent of of these new AI tools, cuz I've been hearing some mixed um, and interesting things from varying companies on, you know, kind of how much time they've saved um, with the with AI agents. uh mainly companies like yourself that I interview, right? They tend to be pretty uh bullish on the fact that it'll save time and they have already saved time because they are early movers and whatnot in the space. But in general, I don't know if you're familiar, it actually hasn't saved like most companies that much time. It's

a very intriguing sort of thing where like in the same way because I feel like some of us don't realize we're in a bubble when we talk in this space like day-to-day companies, um, government workers, uh, your average mom and pop small business doesn't really like utilize AI nearly as much as us. How has it impacted you though as a company and your workforce um, to save time on work? >> Well, our company really has two modes, right? There's a bunch of engineers and writing code with AI is awesome. I mean, I I use Claude code. I'm Claude coding in the background right now. Um, you know, I gave I gave Claude a bunch of like to-dos while we were talking and when we're done, I'm going to go back and be like, "Oh my gosh, Claude, like why are you so dumb?" Um, but

I love you. And you know like to me that's incredible and that gives gives me a massive leverage as a product technologist kind of person because I can prototype and and give my team something that they can play with and touch. I can show it to to clients even and say what do you think before I spend the expensive engineering time to make it actually work right now. I don't think Claude's ready to write production code in most cases. I think there's still a lot of like complex integration work and things like that that you have to do to really make it, you know, an enterprise product, but I do think it's a huge timesaver. The other mode is that we have a big sales team and we're out talking to advertisers and talking to agencies, you know, and I was just joking the other

day that um you know, my team went to Detroit and they every year one of the big car companies has a a golf tournament for all their vendors and partners. and you know, I couldn't send an agent. I had to actually send a couple of my folks and they hung out and made connections and that will lead to business. Um, and not cuz they're good at golf, although they may be, it's because they're good at people and good at making those connections. So, I think for us, like there's a huge advantage to being able to connect in person to people who make decisions. And I think there's an advantage of being able to internally leverage these tools to dramatically speed up our operations, whether that's coding or that's some of our CRM type tasks. Um, I guess the third thing we do is a lot

of marketing and our marketing team's mindset is evolving to assume that people will use agents to do things like RFPs. So, you know, today if you call our salespeople, you can send them a spreadsheet with a bunch of questions. My assumption is more and more people are just uploading those questions into chat and saying figure it out. So all the answers have to be on our website which is very unusual for an enterprise company like you we all know like you go to an enterprise company it's like it's like pricing and you click and it goes talk to a salesperson. Well okay but anything that you can't get as an agent is going to make it harder for the agent to make a good decision. So, we have a radical transparency project going on to try to make as much of our internal sales support

data public so that Claude or ChatgPT or Gemini can become our salespeople or at least our marketing, you know, enablers. Um, and that's kind of mindboggling. It's just not how we've been trained or how we've thought. Um, but I keep imagining that that's our customer's procurement lead is one of these chatbots. So, we got to be really good at talking to him. >> Yeah, that's that's an interesting and again good point. My um thought on that is then like how does this impact your your thoughts for like future hiring and stuff? I I seem I hear a lot of companies saying they kind of plan on thinking of how they can grow their company um from you know like a tech standpoint and keeping who they have and whatnot, but like what's your kind of uh approach going to be moving forward? >> Well, I

mean we we operate in a really complex global business like advertising is massive and global and there's a lot of integration complexity. So, we work with a ton of different partners and a lot of those integrations, while people may publish an API, um, they don't always work and there's so many weird special cases about every one of them that a lot of our people time is actually spent clarifying things and working through these weird gotchas. And so, if you look at how many Slack channels we're in with our partners, like so much of the conversation is us negotiating. Like you might think that like Amazon has like fixed APIs, but of course not. They have 100,000 engineers who are building new stuff every day. Those those engineers want feedback from us. And if we're like, "Hey, this thing you built doesn't make any sense." Or,

"Hey, could you make this thing?" There's humans at the other end of that line that we have to talk to. So, if everything were like static, maybe AI could do my team's job. But my team is actually my engineering team working with humans. And I don't think that's most people's mindset for what code is, it's like, but the thing is if if you can just tell an agent to do it and it can build it, I'd call that easy. And if you can't, I'd call that hard. I want my humans to do the hard stuff and I want the agents to do the easy stuff and I, you know, figuring out how to do that well and the kind of technologists we need to hire who are actually better at humans. Look, we've messed up our engineers in a lot of ways over the past

20 years because we've been training people to write code, not training people to understand systems and interact with other humans. And I think it turns out that the actual skill set of being human of of negotiating of discussing of communicating of of collaborating will be the superpower and that you know the writing of code isn't but understanding code is really really important. So your computer science degree remains valuable but your codew writing skills aren't why if that makes any sense. >> Yeah it does. There was a there's a comment that someone made on a previous podcast about how you can tell an engineer in a UI add this check mark, right? And that they'll know how to code that in, but that doesn't mean necessarily that they understand why, you know, you want to add the the check mark to the specific spot. And it's

kind of an issue in that regard. And that's where the human component really comes in. >> Yeah. It's it's the classic of, you know, >> just the goal of the system. >> Yeah. It's like a doctor, you know, like, you know, cutting with a with a scalpel is like part of a surgeon's job, but like the vast majority of the job of a doctor is figuring out the problem and figuring out where to cut, you know, they have these new robot surgeons, you know, and that's cool, but at the end of the day, you know, that's mechanical. The the brain part is what you get all the money for, if that makes any sense. And you know, it's good to have good hands because I don't want you messing up the surgery, but that's not the thing that's so special. Yeah. The There's another analogy

to this that was that was showcased. It's um interesting. You know, we probably thought in AI that creative things would maybe be taken away later. Um and then a lot of these images that are happening are just beautiful from some of these generative AI tools. Uh, and there was a there's a reference on another episode where the interviewee basically said he was talking to a photographer and what he realized is the photographer went from being completely negative about what's going on and saying like, "Oh, I'm going to be out of work." to realizing he basically could go on photo shoots without leaving his home. you know, essentially he could type out things in a way in which your average person is not going to be able to do it. You know what I mean? Like there's no there there's no replacement for the mind in

that context. And that's the same thing we're probably seeing with code, I'd imagine, is that the engineers and and those who are interacting with them are going to be able to do a better job and quicker job of implementing uh things rather than they're not needed to make the things is the is the delineation. >> Yeah. Well, that's just composition, right? Um my my friend was here this this week and he and I were we we took a picture of this beautiful Tuscan landscape and then we all had a prompt competition. We all tried to see who could make a prompt to midjourney to make it make the most similar landscape to your point. Exactly. And so, you know, his 15-year-old son was here and, you know, it was actually a really interesting question of can you actually talk about what you see because, you

know, like there's olive trees, but there's also cypress trees and there's a pomegranate tree and some, you know, there's a valley and there's a road. But if you try to type that into a prompt, mid journey is like, what are you talking about? Like, it can put those things there, but it's not anything like this landscape. And the more we tried to iterate, the more we realized like how difficult exactly what you're saying. It is to describe all the things like the clouds in the background and we can see the you know the towers of Sienna over there and there's these beautiful. So after a while I was like okay let's try something different. Let's take this picture and let's just upload it to chat GPT and say make me a prompt. And so then we saw how reverse engineering the prompt how that looked

compared to our prompts. And so having chat GPT take the picture, prompt it to midjourney, and come out with a picture was really interesting because just it just pulled different things from it. And of course, it was a much more verbose prompt. I'm not sure it was better than ours, but it was an interesting question of like what it pulled out of the process. So I I guess I'm saying is exactly this this interesting interaction between, you know, yes, it's beautiful, but it's not what's outside my window. The funny punchline to the story is my my daughter and you know my my friend's kids were running down the hill and I took a picture of it and sent it to my wife because it was really beautiful and she said, "Oh, that's AI." I said, "No, that's that's reality." And she said, "But but the

colors are so bright." And I said, "No, like that's where we are. We're in one of the most beautiful places in the world. You don't need AI. That's actually what's outside the window." And the fact that her immediate instinct when she saw this picture was AI was was kind of scary, right? Like our our default belief if something is like extraordinarily superhuman, beautiful or whatever or supernatural anyway. That's just that's the moment. And in a hundred years, I can't even imagine what the world looks like, right? Like could you even imagine where we're going to be? >> No. and and yeah and that's that's pretty funny that your wife immediately and a lot of people do come to that assumption now um it's it's pretty interesting and you know we weren't I mean I try to keep things in reference right we I don't think

we were anywhere in a good place in regards to image or video generation when it comes to AI like a year ago like I don't actually think that's understanding it. I think a year ago we didn't have anything production quality like people forget GPT40 image generation is like 4 months old and that was like a huge leap forward if you recall like that that's four months old right like text and showcasing the text in such a way to where it was accurate upon the request was impossible in image generation like four months ago. That's how quick it's moving. So imag I I I don't Yeah. >> Oh yeah. Well, that's that's why this is fascinating and why I you know I tell people that you know I feel like I'm on the forefront of what's possible in my little world of of what an agent

can do. And uh I say 95% of the people I talk to think I'm crazy. Like they think it's like, you know, okay, cool. Like you're on the fringe. This is like so far away. But what you're talking about is the pace of change and where it goes from like an interesting idea to being better than anything you can imagine. It's happening in every single space. There's a threshold of, you know, pet toy project to, oh my gosh, this is the way I'm going to do everything. I've seen it in advertising. I am blown away. There was a point where I started just doing demos for customers and I would just not know it was going to happen. I'd be like, "Okay, Nike Ultramarathon shoes and I was kind of nervous." Like to your point about like if you type that into a midJourney a

year ago, you'd be like, "I don't know if it's going to be good." You know, it might have seven toes. Today, I'm 100% confident or like 95% confident I'm going to get a great response. And and if you I even put this demo live on the internet for people because I was like, "Just try it yourself. I'm tired of doing demos. Like literally, you can log in for free, try it, and see what you think. And it's been amazing to see the the interaction because it is so good. So, I think that's my biggest takeaway is while I have no idea what the world looks like in 10, 15, 20 years, it's going to have AI in a very central place because it does amazing things. And you can be a hater, you can be a denier, you can It's it's spectacular and and agents

are going to make our lives dramatically better and dramatically different. In the same way, you know, you used to get a newspaper. There's a lot of things that in a decade we're going to be like, remember that thing we used to do? I don't know which one it's going to be. You know, will we still have cell phones? I don't know. Probably. But I bet a lot of the stuff we use and do every day will be so obsolete, you know, and and we won't even believe that we were so dependent upon it. You know what I mean? Like we can't really imagine the fact that, you know, we didn't have smartphones 20 years ago, right? Like can you imagine your life without a smartphone? They didn't exist. There was no smartphone. >> Yeah. 2007 was the was the Steve Jobs uh keynote. Um yeah,

that's true. And it was and it went from him talking on that like really small stage uh to now if you ever seen like the summer what is the name of the summer uh Apple event every year? >> WWDC. >> Thank you. Yeah. Yeah. And I look forward to it every year and it's it's so over the top and in regards to the animation and stuff that they build and then I just think back to like oh yeah like Steve was out there talking on like a black, you know what I mean? Like it was Yeah. So we're we're in such a different place than we were. So, um I guess moving forward with with this, you know, it it kind of begs two questions. You know, you seem like you're really excited. Um my two questions would be, and you can go in whichever

order you feel like if you want to do the good news, bad news, bad news, good news order. What are you looking forward to the most with the advent of uh more AA AI agent capabilities in in what you're doing and in general? and the opposite like what what do you what what does make you maybe nervous if there is anything? I think what makes me worried is that this is a pretty fundamentally disruptive moment and that there's going to be winners and losers as with every transformational moment. And I think the winners are, you know, going to be, as it seems often these days, the the folks that are really technologists who are who are able to lean all the way into these things and and you know, take full advantage of this kind of technology. The other winners though, it's interesting. It's like

a U are the folks who may struggle with digital technology, who find the complexity of a computer or a browser, you know, like overwhelming. If you can just type or talk to an agent, it's interesting. It's like the top 1% is going to be incredibly effective. The bottom 10% is going to be incredibly effective because this is going to let them plug into technology in ways they couldn't before. But there's this weird middle and I think it's a very white collar upper middle class middle who survives in a world where they do a lot of you know let's call it medium value task work and I don't know that they're going to be as effective as agents and so I think my optimism is that I will arrogantly say that I think I might be in the top 1% of people who can leverage AI

I and and you know I'm a computer scientist. I'm an inventor. I have patents. I have all the resources to like fully take advantage of this new technology. I I know an industry really well. Like it's going to be good for me. Um and so that's good for me, but I do worry that there's a lot of folks where it's not going to be great and there's not going to be an obvious place for them to go. And you know, if you're graduating college right now, you don't have a lot of those like human connections into an industry. you know, you're not great at anything yet, probably. How do you get those skills? How do you learn? How do you grow? How do you get the decade of of experience you need when the actual stuff you can do now may actually be in many

cases more um cost effective for you know claude max for 200 bucks a month you know no benefits no HR costs you know might be you know destroying the electricity grid and you know using water that we can't spare but you know those are details because you know like It's so so that's that's that's my biggest worry about is the folks who are later in their sort of career choice in ways that would be hard to switch or the folks who are early in their career who just haven't had the shot yet to build those skills that would be more defensible against AI yet. >> Yeah, I think that's a that's a fair uh call out. My continued drum beat on this podcast in the last six, seven months has basically been I'm pretty sure in the next 3ish years, entrylevel knowledge work jobs um

will be worked out um or pushed out. and anybody who is in like uh that next step will be basically managing agents who are conducting the work and they're QAing it is kind of my my general theory. Um and as things are progressing uh my time I used to say like 5 years and I'm saying three years and I'll probably say like one and a half and like 6 months because it it's it's truly incredible how agents are progressing. Um, but and that's what and that's why I'm grateful like we're we're both probably in a similar position to where we understand how it works and we're using it effectively and whatnot, but I I do I do get concerns about that. I you know any family members I have or friends that are kind of at that age of like, oh, they're about to get

out of college. I'm like, okay, knowledge work is where we moved from after the factories. And I'm very intrigued to see how knowledge work adjusts and changes because I just I genuinely don't see where entry level knowledge workers fit, you know, if these agents exist from like you said the cost perspective. Um, unless everyone has a has an immediate freak out on the on the on the water usage, which is which was a good call out because that is actually accurate. It does use a lot of water. Um, yeah, it's going to be it's going to be interesting. So, uh, on a more positive note, I do think it's it's going to cause a lot of innovation that we're probably unaware of, you know, like that's that is the benefit of it when you quickly are able to add things to your companies and you're

able to give the people who have the ability to come up with these ideas and implement. you know, similar to the doctor analogy where they're instructing the robot. You know, it's like, well, there are situations, I'm sure, where the brain surgeon could have been more physically precise. However, he was limited due to his own motor skills, but he knew what to do. And I hope that, you know, kind of with what we're doing with AI, it uh ends up leading to have those thinkers who are limited by literally these guys and uh time spent on a computer to uh improve cuz I don't I don't know if you're familiar with it, but uh like the I think they're saying every I forgot what the metric was, but every like month or three months, it was one of those two metrics, we're seeing Oh, no. Is

every six months we're doubling the amount of uh compute length for like agents, you know, like how long can it continue to run, you know, I believe it was the number. Yeah. And we're going to get to the point to where I think in like two years they're going to be able to work for like a work week straight or something funny or in like a year and a half. So, I'm very I'm very excited to see cuz things are going to come out of that. We don't know like my my we can have these predictions um that I think some of us will be accurate in but I just like to think uh the whatifs are going to be um above and beyond what we imagined because as as we mentioned 4 months ago image generation couldn't generate text correctly. So we'll be and

we didn't think it was going to happen for a while. So we'll see. Um, is there anything that you'd like to talk about like or or plug obviously but with with what you're doing um just to close it out for the podcast? >> No, I think we covered a lot of great stuff and uh you know would love to show anyone who's in the advertising universe some of the stuff we're doing at Scope 3 and uh you know love the fact that you're continuing to educate and help people really understand what's going on in this crazy world. >> Well, thank you so much Brian. I appreciate your time here. Everyone, make sure to go to scope3.com. That is spelled scope and then the number3.com. Thank you so much for listening. Please leave a like and review for the podcast and we'll see you in the

next one. Bye.