GTMN

The AI Marketing Divide, Webflow vs. Claude, & PMax Updates | GTMN 18

• Pranav & Austin • Episode 18

Share episode

Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.

0:00 | 40:59

Is AI actually driving growth, or just making marketers faster at the wrong things?

In this episode, Pranav and Austin dive into a deep debate about the future of marketing careers in the age of AI. They also unpack Austin's genius workflow for running Claude Code remotely, the legal battle between Amazon and Perplexity, and WPP's radical shift away from billable hours.

SPEAKER_02

March 26th, we took a couple of weeks off here, Austin. I was traveling, I was busy with some stuff, which I can't talk about just yet. But uh I am super excited to be back. And did you say you switched back to Chrome?

SPEAKER_01

I switched back to Chrome, yeah. It's so sad. I I really believed that Safari could work for me. And you know, the thing that really drove the hammer home is I would be trying to log in to Okta for different client work, and like one password just would not load. And it's this well-known bug in Safari where like sometimes Safari like blocks one password. And all the other stuff I could put up with are podcasts not working from Safari being one of them. But the whole like one having to physically type in passwords in Safari was just like too much for me. So I abandoned it. It was a great experiment. I'm sad that I lost my original browser Arc, which is a Chrome-based browser. I think that the CEO of that company made a huge mistake getting rid of Arc. I guess although it is part of Atlassian, so maybe we'll come back now.

SPEAKER_02

Yeah.

SPEAKER_01

But yeah, I rip, I miss you, Ark. Ark was the best browser in the world. Um, so I'm back to Chrome.

SPEAKER_02

Oh no. Well, welcome back. I've been on Chrome my entire life, I feel. I I think I tried um OpenAI's uh Atlas, and that's also getting shut down, I believe. That that was news, right, this week, where open AI is like all focused back, they're killing Sora, they're killing Atlas, they're killing a bunch of other stuff and going all in on the same cloth strategy. Why do we start there?

SPEAKER_01

I know, I know. It's um, and for what it's worth, like I just don't know how much I need AI in my browser when I'm browsing. The thing that I really care a lot about actually is having my browser be accessible to Claud Code. And I don't actually have that solution figured out. I'm sure there's people maybe if somebody's listening to this and there's a great solution that you or they know of, they should email me. But what I really would like to have happen is that I can be working with Claud Code in an agent session and it has the ability just to not open a separate Chromium browser which has its own session cookies, but like literally just open the one that is on my screen. And the nice to have would be like if you could actually hook it up with a remote connection. Are are you familiar with remote connections with cloud code?

SPEAKER_02

I mean, how is that like that's basically a dispatch, or is that something else you're calling?

SPEAKER_00

Oh god.

SPEAKER_01

Oh my god, it is so much more than dispatch. This okay this is what's crazy, Pranav, is like things are happening so fast that I don't even think people know the obvious things out there. If you're in a cloud code session, there are two commands. One is you can tap dash dash RC, which so if you do like, you know, claude dash dash, it's kind of giving signals to claude code when you're in the terminal. One that people probably know is claude dash dash resume resumes a session. So you can pick the session that you have. And what's really fun is if you spend some time, you can actually have Claude build a session management tool for you so that you can explore your old sessions and pick them whenever you want. Your sessions are stored as JSON on your computer locally. But one of the things you can do is add a dash dash RC when you start a Claude session. And what that's is saying to Claude is, hey, this is a remote connection Claude session. I want this to be accessible to the Claude app. And so I start, I made a command that starts all of my cloud sessions with RC. So they're all accessible then inside of the Claude Code app, which means you can use your local machine to do work from your phone wherever. And so one of my favorite things to do these days is I start all my sessions in Cloud Cloud with RC. They all become accessible and I'll, you know, for the people who are Okay, I get it.

SPEAKER_02

I get it. So basically you're saying you're starting them on your machine, on your Mac, and they're available on your mobile app.

SPEAKER_01

Yes. But they don't use cloud services, they use your actual machine, Pranov. So all of your MCPs, all of your APIs, all your skills, all your directory that you've spent all this wonderful time building on your machine, guess what? It's accessible over the internet. As long as there's a little bit of work, you have to make sure that your computer doesn't sleep when it's closed. So I have a you can just ask Claude, hey, how do I make my machine never sleep when it's closed? Make sure it's connected to the internet, and now you have a server sitting in your laptop whenever you want.

SPEAKER_02

So so I did that this weekend because I was running into issues where I wanted stuff to like just always be running behind the scenes. And I changed my Mac setting to just never sleep if it's powered in. But I'm like almost tempted to just set up an actual Mac, full, full blown Mac desktop at home for this purpose. Right. That there's like it just the amount of productivity gains that that's gonna bring me is gonna be crazy. It's the same thing as like Mac Mini and you know open claw or whatever, but just using my claw. I don't need open claw. Why do you need open claw? Like it's just running in claw.

SPEAKER_01

The only reason for open claw is like, so here's this thing, right? Like you have to tell the computer to create sessions and you're creating those sessions, right? Now, if they're on your local machine and uses Claude Max, which is a fixed cost,$200 a month. Sure. If you now have a server that doesn't have Claude Max, now you're running API tokens. Now, maybe you know, Paramark is subsidizing your costs. You don't care if it's expensive. I mean, that there is a lot of people out there that don't care. They're like, hey, I'll just create sessions with the anthropic key and my API key and I'll pay for it. But I am a humble poor consultant who does not want to pay thousands of dollars for these tokens. So you can't.

SPEAKER_02

Wait, wait, wait, hang on, hang on, hang on, hang on. I know you are not poor. We talked about this right before this call about you are you are doing amazing work for some amazing companies and you are highly in demand. So, folks, if plug for Austin, he's being super humble. If you need help with your tech marketing stack, right? So everything that you need to do to get your marketing engine going, he's your guy. You need to call him. I'm pretty sure he's oversubscribed, has no time, but you need to talk to him. All right. It's false.

SPEAKER_01

More time because of Claude Code, but there we go. Um but but yeah, so so Pranav, put put a computer in your basement. The reason why you have clawbot is like, what if you want your machine to work overnight without input from you? Right. Like let's say, for example, you take take Saturday and you put five or six different projects on a linear board with all your ideas for them, and you spend a whole hour documenting the things you want to build, the PRDs for them. Like you spend a lot of time in detail putting these tickets together. And you're like, here are the five things I want to build. I want to build like a contact directory, I want to build like a health and training app, I want to build something for my investors, I want to do X. So you have these tickets, right? In today's workflow, we're speeding up really quickly, but you have to help Claud Code answer questions. It's gonna have questions about what the design architecture should be, where you want to deploy it, blah, blah, blah. And what that requires is for you to babysit Claude Code sessions. Yep. The whole purpose of ClaudeBot is to say, no, actually, if you give sufficient information and context and memory to your computer, you can let your computer guide the rest of the agents to continue the work stream, and then you just validate the work afterwards. So overnight, Clawbot is running its own version of Claude and it's directing Claude to do the work for you. I'm with you. I'm with you. Yeah, I talked about this in the article. The only limitation right now is cost because running that overnight for like an infinite number of queries is really expensive, which is the reason why people are seriously considering like, do I just go buy a$10,000 Mac Pro and put a local model on it, like one of the Chinese models, Quen. But hey, Quent, the latest version of Quen is as good as Opus was in December. So you're like, okay, now my my computer is running Opus five, whatever, and it's directing all my Claude Maxes to do all the work that I design. So all I really have to do now is like be thoughtful up front in designing the work that I need to do, and then keep guiding my little Mac Pro to kind of continuously improve.

SPEAKER_02

Yeah. No, I'm totally with you. And I think that all tracks. And it's weird. We started this episode of GTMN, all about go to market with uh Claude and OpenClaw, but that is the reality, right? Like you look anywhere, my my X feed, my LinkedIn feed, my YouTube feed, it's everywhere. And like Q1 has changed. I believe, I think this is like a fundamental pivotal change in all of GoToMarket. I was at a conference last week or the week before in Phoenix, Arizona. It was a bunch of B2B CMOs, and nobody, like, I don't think there was anybody who didn't talk about how Claudus changed their lives. So it's an interesting time. I want to start with uh a couple of news items that are interesting, and this is a couple of weeks old, but fun, fun things going on. So, did you hear about this? That there's a federal judge who has banned Perplexity's shopping agent from Amazon orders. And then there was an appeals court that reversed the ban, so now they're allowed an Amazon again. So this is an interesting sort of push and pull happening in the ecosystem. I don't know if you've been following, where what is an agent allowed to do on any platform and what is it not? And especially like Amazon, Amazon's like trying to fight it in a big way. So interesting, you know, future, like it's not obvious and clear whether this whole idea of like agents doing stuff for you is going to take off or not. All these larger businesses are trying to fight it. Any any reactions to this, like from your perspective?

SPEAKER_01

I mean, I just feel like I keep coming back to what you said at the start, which is like, it is so hard to predict what's going to happen based on how much is changing right now. And there's a part of me that thinks, like, okay, maybe we'll hit this plateau in the next quarter or two, and like this will all be a lot about nothing. There's also a part of me that thinks, like, okay, the step change from Opus, the last Opus model that was released in December, that is that along with a bunch of other stuff, is like the reason why we're experiencing so much change now and why I think agents now can be useful. So, what will happen when the next version of the model comes out? And not just like the next kind of incremental version, but like the next major model update in say six months.

SPEAKER_02

I'll tell you my my reaction to this. Like, I was thinking about my own personal shopping habits. And over the last like three or four months, I've definitely switched to asking Gemini for everything, right? So like I'm buying, I was like frustrated with some like skincare product, and I was like, you have all my skincare history for the last like two years. And I just asked it, like, hey, with all the conditions and the you know things that I have, here's what I need to do. It spat out three things. Now, this was interesting. It didn't spit out just one thing, it spat out three things, and then I went and looked at each of those three things on Amazon and then made the order. So the question is, does it get good enough where you can place the order within the LLM? Or, you know, and and that's and that's what this case was all about, right? Like, do you let Gemini, Perplexity, Claude make the purchase on your behalf from Amazon, or do you still force the user to go to Amazon and and take action? Because that then allows Amazon to upsell, cross-sell other things. Do you show you ads within Amazon and so on and so forth, right?

SPEAKER_01

So I think the consumer process, Pernal. I think I think before we see full agent buying, we see an agent approval button that uh pops up in GPT and Claude, where the agent is ready to purchase and they have your creds and they can make the purchase, but you just have to click a button right from Claude. Or maybe it's a a notification message sent to your phone, or maybe it's an email, but like there's probably still human in the loop for quite some time, but we just make it frictionless. And if I think about what I would want, I would probably want that until AI is good enough that I have full trust in the outcomes and we're just not there yet, you know? Yeah.

SPEAKER_02

And I can't imagine like Gemini, why wouldn't Gemini just allow you to buy directly from Google Shopping? Totally. They have Google wallet, like it should all just be one click. It makes complete sense for certain type of purchases. But yeah. So that yeah.

SPEAKER_01

So yeah, I mean, this there's two things that come to mind. Like one is the intermix between like agent permissions and human permissions is very fascinating. I have been experiencing this in my own life recently, where I've like, I've created a service token for my ClaudeBot and I'm giving him access to some tools. And like in my back of my head, I'm like, I should probably be more careful. But also he doesn't do anything unless I explicitly ask. So what is the risk? The risk is him kind of assuming I want to do something and using my credentials nefariously. So I think like that's a very interesting space around how can humans give full trust and autonomy with guardrails to like prevent really bad stuff from happening. And the second thing that comes to mind is, and I maybe you saw this on Twitter this weekend, but there was like a malicious version of a third-party library that is downloadable by Python. And what happened is probably what you're seeing when you're using cloud code is like it just suggests random libraries because it can do research really quickly on GitHub, and there's lots of independent libraries out there. And that's largely been like the way of code for the last 20 years is like build with little mini libraries of things that you want and need. Try not to build stuff that you don't want to maintain. And so I personally have gotten very accustomed to just downloading libraries of code that I then use for the applications that I'm building. But this thing that happened was like one of these libraries was like an inject, I think it was an injectable prompt. I don't know the exact details, but it was able to get like SSH keys, AWS credentials, like all this stuff that's tied on your computer because it effectively made a prompt to go scrape it. So like there are a bunch of people who are installing this library for something that's like pretty benign, and you know, I think it had millions of downloads, and all of a sudden this actor is collecting like your AWS key, right? Yeah. So how this ties back to the shopping example is like on one hand, I think a lot of us want to trust entirely, but the unfortunate side of technology is that there are bad actors who utilize it to make this harder. So it'll be interesting to see how this evolves with the major players. And and again, I that's why I also think like human in the loop will just remain for quite a long time until you can imagine like you know, 2FA and MFA evolving to this, right?

SPEAKER_02

So in that example, like Gemini doing something, but then there's a 2FA on your actual mobile phone that is not connected to those systems, right? And I imagine like tools like OnePassword or whoever else is playing in the space, they should go all in on you know that as their you know next big push agent permissions. Exactly. Yeah, exactly, especially for the enterprise. All right. Lots of other interesting things happening. So let's talk about this one. Google Ads is now mandating AI voiceovers for performance max videos. So essentially what they're doing is starting March 20th, so just like last week, they are going to layer AI-generated voiceovers. If you have a performance max video, they're automatically just gonna create that, the the voiceover for you. And um, you don't have to do anything. They're just doing it, they're launching it. You have to opt out if you want to opt out, otherwise it's gonna be done by default. So yeah, they're going all in on this one.

SPEAKER_01

Why why do you think that is? I mean, voiceover seems just like an odd place to be a hill to die on, you know.

SPEAKER_02

I think more and more you're gonna see like Google, Meta, Amazon, all of these companies make it easier and easier for these smaller brands to create, you know, these artifacts, videos, assets that otherwise would have required a lot more time and effort. And the first version is gonna be crap, but I think in like, you know, as we've seen with all the AI models, it's gonna get really good really quickly. And, you know, you've got, I don't know if you've seen some of these performance max videos, but they're basically like you drop all of your assets, right? So if you have a product, you have your brand guidelines, you have a logo, you have, you know, colors, and then it just like does some basic animation. So if it's doing that basic animation anyway, then you might as well, you know, tell tell them something about your brand and just says, uh, I'm just making this up, you know, uh the lightest shoes uh uh on the market for marathon runners, right? And that's it. That's the voiceover, and it's like a six-second ad in front of a YouTube reel and it, you know, very nicely produced, deep voice, you know, perfect for for for video. Yeah, why wouldn't I take that? So I I think it it works. The I was just surprised by there's a 10-day window to opt out, and otherwise you're just opt it in. Yep. And then you might have some funny voices. Uh by the way, have you been so I use the Google News app on iPhone and they launched a new feature which is two AIs basically create a 15-minute podcast with all the top news on a daily basis. So now in my car ride to work, I just tap the voice listen feature and it just like takes all of my Google news and gives it to me. It sounds like any other news app. It's perfect.

SPEAKER_01

So I know I think an a fascinating business opportunity for anybody out there who's looking for things to build is a dynamic uh morning podcast that's customized entirely for every person who's listening. So rather than have one pod, you log in and you get your own custom pod. So, like what you're saying, but it's a collection of the things that you care about. And so what you do is you actually spend some time looking through your existing sources of data, like the places that you listen to and get information from. And I talked to you about this maybe a week or two ago. I built a version of this personally. I have like my own consulting software called Taurus. I get a morning briefing, and in that morning briefing, there's a lot of data, like health data and action items and client stuff. But towards the end, there's a list of 20 articles from the prior day. And it pulls from Stratchari, it pulls from Planet Money, it pulls from Lenny's, it pulls from the Wall Street Journal, it pulls from the Atlantic. And the problem I was solving is probably what you're thinking too, which is like I have a limited amount of time in the morning to go drive to work and I have so many sources of information. How do I actually keep up to date on all these things? And so the first version of that is literally just like um imagine like a mega RSS collector, like with the only difference is that if you collected all the RSS feeds, it would be like a hundred articles. And so you use Opus to kind of think about your preferences and pick the things that are most relevant daily. And that's just like a daily brief I have that I get. It's not too hard to say, well, let's just productize that. Let's let people enter the names of the things that they already look at on Substack or other sites. And now let's run a prompt at the end that generates a two-way conversation, you know, BPN, you know, or whatever style combo.

SPEAKER_02

Yeah. No, I'm uh I'm 100% with you on that. All right, a couple of other things that are happening in this world. Have so Webflow, I'm sure you've heard of uh Webflow. Uh they acquired a company called Vidoso to launch autonomous content marketing. Had you heard of Vidoso? Like I had come across them, but I didn't know they were a big deal. But there's an AI like video company. And so that's going to be integrated directly into Webflow, allowing users to create SEO optimized video assets based on their site's design and data. So that's an interesting little play, but like going all in on video for web content. And I know why. It seems like there is there was a new update from Google that has made, yeah, there we go. It's called multimodal indexing. I don't know if you heard about this, but this is like now live 100% globally, which is enabling the search engine to index and understand video, audio, and podcasts at the same depth as text. So your video content, and you could be an early adopter here, if you're doing a lot of video or podcasting or any type of audio, that is a net new, like, you know, deep information that Google previously was not indexing as much as it does now. So these two things seem very related, like Webflow going all in on Vidoso, and then Google launching this multimodal indexing, which is fascinating to me.

SPEAKER_01

So I have a question for you. This is a great debate I've been having with a couple of marketing leaders and even with one of the teams that I'm working with, which is, you know, Webflow when it first started was revolutionary because it had a UI for people to interact with and build what was traditionally very hard websites in HTML and JavaScript. And I think over time, as the complexity of the product grew and it could handle more complex use cases with animations and stuff like that, it actually became a tool for front-end engineers. Because I remember specifically at Ramp, a conversation with our one of our directors of engineering was like, hey, we're actually the ones, engineering is the one that's making the code changes in Webflow these days, and we're the one approving branches. And so it's this interesting thing where like it started with a value prop to marketers, but then became so complex that engineers had to be the ones to use and approve it. And now we're in this place where I'm like, actually, I don't know how doing something in Webflow would be easier than just writing code with Claude, right? Even for technical, for a technical growth engineer, they'd probably much rather write code. They'd probably rather do TypeScript or, you know, React. And for like you, you could just write Claude prompts. So where where do you think website creation is going and how do you think about those problems? So we use Framer, big fan of Framer.

SPEAKER_02

Um and I'll I'll tell you what, and you know, my team was creating a new website on Framer for something, the project that we're working on. And we kept running into this issue where staging and prod would behave differently. And it's like it's like that that is the the complication with this whole space. Yes, you can ship code, like we all can just, you know, clot code and frame has like AI capabilities, so you can like do AI within Framer. But the problem is when you get in stuck on like complicated flows that just are behaving differently because of some weird caching issue, you're kind of stuck in in this like you know, hard spot. So I would say, yes, do all of the things with AI, but you still have to pick a platform that's helping you when you are trying to deploy from staging to production. There's like 10 people working on the stuff and you can break shit. And that collaborative stuff is where, you know, proper CMSs, whether it's a webflow, whether it's a framework, whether it's a sanity plus something else on top, right? Whatever whatever is your tool of choice, you're still going to need sophisticated tech at at that scale. But if it's a small operation, yeah, I don't think you need to worry about all this.

SPEAKER_01

Yeah. Well, I I just also think like, so a team starting from scratch, do you start from scratch with Framer? Do you start from scratch with Webflow or do you start from scratch with Cloud Code? And I think the answer is most people, especially solo founders, can just start from scratch with Cloud Code. I agree.

SPEAKER_02

It it it I'm thinking about like if I were starting a company, you know, again from scratch, would I do Framer or would I do Cloud Code? I'd probably do Cloud Code.

unknown

Yeah. Yeah.

SPEAKER_01

Yeah. Well yeah and I mean I I also just think about like most marketing teams are small in nature and scrappy. And usually it's growth engineering or some product team that owns the website surface area. If it's in code, it's already like you get the benefit of just being able to use cloud code directly. But if if you're in, you know I'm having this this struggle with a couple of folks I'm working right now, like the value prop of webflow and frame is that it's supposed to make it easier for marketers to update the website. But I'm like, what is easier to learn how to do frame and webflow or to teach yourself Cloud code? I think it's probably easier and better today to learn how to use cloud code.

SPEAKER_02

Did we did we just short the entire uh CMS market?

SPEAKER_01

Well no no so the the thing with the CMS is I think like even if you build a website in cloud code, you still need a CMS. And I'll cloud code the CMS. You could but same with CRM. Like yes you can but there's a lot of features of a of a CRM that are hard to build. Do you really want to stand up and manage a database which is what you need for a CMS?

SPEAKER_02

So this brings a really interesting point because I was looking at this company yesterday called 20. I'm sure you came across them they're building an open source CRM and giving you full like GraphQL and you know I forget like some other API access to all the objects and what have you and then the business model is you can deploy this on your own or you know you can deploy it on their cloud. So they're taking a different approach to the AI CRM market. But the reason I was thinking about this is like maybe that's what happens where you have a resurgence of open source and you pull that GitHub that they've got like 40,000 stars or 4,000 stars. I can't I can't remember this is like wildly popular on GitHub. Yeah just crazy. And so you can just pull that to your you know you can get download that and then just build on top of it. And then you've got the best of both worlds. You've got the CMS foundation built in and you can you know go crazy with it. So I think there's an option here where open source makes a massive comeback.

SPEAKER_01

Oh completely I mean yeah I I I mean it's like it's this trade-off between what do you build yourself that's easy and what do you build that's hard. And I'm like I'm increasingly finding like things that require databases you should try to find a third party for if things that don't require databases like front end matter easy to build with quad code and easy for you to manage in a GitHub repo. And so that's for just like for a a really classic example is like nobody's going to build a payments processor. They're just going to integrate correct because it's so easy. Yeah. And similarly like maybe with CRM like people are not going to really think about building the database infrastructure for CRM, the core primitives, but they'll totally like throw their own front end at the CRM, right? Yep. Yep. Like you can you can make it a single page, you can make it an object page, you can a multi-object page, but it's just like the Rails are already ready for you.

SPEAKER_02

And I had an interaction with another friend of mine who had built or was helping their friend build a storefront. And I was asking them like why did you build it? Why didn't you just use Shopify? Shopify is like so cheap and accessible. And he's like I just found it easier to build it rather than worry about learning Shopify. Exact same conversation. Right. So whether it's CRM, whether it's you know e-com, like the the world is just shifting majorly beneath our feet. Okay, a couple more things that are interesting um Austin. So both Salesforce and HubSpot back in the CRM world have restructured their partner programs for AI. So this is news they're basically now going back to all of their partner programs and forcing agencies to specialize in AI deployment. And so not just about platform support, but like essentially how do you drive AI usage and AI tooling within the systems. And I would imagine like every Salesforce and HubSpot consulting shop is now just going to be pitching themselves as like, hey, we know how to maximize AI usage on these platforms, which is fascinating to me. So you know different approach from an AI native CRM, but like hey let's put this entire army of professional services people into doing AI work.

SPEAKER_01

All the all the professional services companies out there that specialized in deploying you know resources to help you build against the SOAP API just became way more lucrative overnight from Opus Yep.

SPEAKER_02

And I think the same the same thing's gonna happen with Shopify agencies, right? So you've got all these web development and e-com agencies and they're all going to have to pivot to becoming AI agencies, basically.

SPEAKER_01

Well maybe I mean I think they an agency can still specialize, but I think what's important actually for buyers is to realize like these things are no longer hard to do. You're paying somebody smart to use Claude. So I like this is also a debate that's happening is like what who's going to benefit the most from the frontier of Claude. I think people with specialized knowledge in fields for a long time are going to maintain the advantage. As a great example like even with me and you, it's like okay I've been doing marketing technology for 10 years. I understand it really well. I don't think people are suddenly going to be using Claude to replace me. I think maybe people who are just starting can get a lot further without me. But for the biggest organizations who have money and want to pay experts for things, the value that I'm providing is that actually I can do it 10x faster and 10x more efficiently. Now whether that efficiency gets passed on to the customer or not is a different story. But you can imagine like say you hire a Shopify agency to build your you know your back end shopping portal for you, you care about the outcome as a business owner. If that outcome is achieved five X faster because you're using Claude, who cares? The only thing that will change is like if I have an engineer who says, hey, I can probably use Cloud code and build the same thing that this agency is building in the same time, well now I have negotiating leverage to go ask the question of the agency can you do it faster or can you do it better in some way and how? And it's just it remains to be seen whether like what scope you're working on is totally buildable with Claude or if the knowledge that you have from your specialty actually means that you're better able to babysit and produce an outcome that matters. And it you know it's unclear like I I think I told you this a week or two or two ago I built this Martech demigod agent that's a collection of all my work and a bunch of work that I care about in the Martech space. And I've been deploying that agent or collection of agents against a lot of problems that I'm solving. And like nine times out of 10, it does as good or better a job than I can. The question is like can other people produce that set of agents? And even if they could, could could they validate what it's saying to be truthful or not?

SPEAKER_02

Exactly. That's the the evaluation is the hardest part and you know how to evaluate the work of those agents but I can tell you that 99% of other people don't. And yeah.

SPEAKER_01

So I think for things that don't require specialty there's a great article I think in The Atlantic which is like why are all the bros in Silicon Valley suddenly into taste? Like this is what it means. I think for things that are not sophisticated or hard require technical expertise, we're using this term taste as reference to like basically good judgment. Like are you able to use your brain to look at what the output is and make a determination of whether it's good enough or not. And to be really clear, I think there's a lot of things especially in marketing that just require being thoughtful and smart don't require a specialty at all. And I think that's the reason why a lot of marketers are scared is because you know they have a specialty that is much more in the line of sight of smart people being able to solve with agents.

SPEAKER_02

Yeah, totally fair. Okay, on that note I I have a couple of interesting pieces of news here. So WPP which is a big agency one of the big four they have announced that they're shifting their strategy towards outcome pricing for AI services. It's a new billing model where instead of billable hours they're going to have essentially fixed cost performance guarantees based on specific sales or lead gen targets. I don't think I've ever seen an agency do this. So this is actually quite interesting and innovative. Yeah. So this is like one of the big ones saying hey our open AI platform they literally call it that WPP open AI platform is only going to charge you if we hit a certain amount of sales or lead gen targets. Wild. No time and materials which is the yeah which has been like the de facto for the last you know five decades. It's totally crazy. It's totally crazy. It's possible now which is which is you know fun but yeah. Couple other pieces that happened so this one's a fun one. So Amazon actually launched a creative agent that it's agents everywhere these days that automatically generates creative briefs and storyboards from product reviews and features that you've entered. So you load up your, you know you get like new reviews all the time. It's taking those reviews and the most interesting pieces and coming up with ads uh and then turning those ads into prime video. So these are for video ads not just for like what you're seeing on Amazon's you know mobile app and what have you. So that's interesting. And then our fund OpenAI team is now launching or they have hired a meta executive another meta executive to head up its newly formed advertising division. So this is interesting like I don't know if you've heard at all from like what's happening with ChatGPT ads. We have a couple of customers who've uh launched ChatGPT ads and I've been talking to everybody about like measurement. It is so behind right now. So what you get is UTM codes which is better than nothing but they uh essentially you get nothing else you get a CSV download once a week that tells you how much was spent, what the impressions were and what the click throughs were and that's it. There's no geo targeting right now. There's no you know any other type of targeting it's just all up to how they configure it. There's no ads manager there's no like you know UI and you still have to commit something like 200, 250K to you know get access to ChatGPT ads. So it's wild. But I'm guessing like you know with this hire they really are going to start to change things around and make a real ad product. But the initial launch has been pretty uh pretty underwhelming according to what we hear from from customers.

SPEAKER_01

Yeah I mean we talked about this a couple weeks back anybody who's gonna drop 250k on this should be treating this like a brand play. Yeah like totally trying to measure this like a deterministic channel is like a laughable waste of time. It will not be developed enough to make it deterministic for a long period of time. So treat it as a brand play and if you have the money to spend and which to be frank a lot of companies especially in the Fortune 500 have an inventory problem. They don't have enough surface areas to spend their budget on. So if I was a CMO of a Fortune 500 or a public company I would for sure be dropping 250K because like it's an unmarked channel that nobody else is using. Who cares? Completely agree.

SPEAKER_02

All right is there anything else that you wanted to chat about that you haven't or that you've noticed uh in in terms of what's happening in the market?

SPEAKER_01

Okay. Here is one not I feel like this session has all been about AI, but I think that's good because this is like the moment that we're living in. I'm kind of curious to get your take. I have been thinking and feeling for the last week that marketers who don't learn how to use at least something like cowork or Claude code or the equivalents across GPT and Grok are either going to get left behind or are going to be in a lot less demand. So I'm kind of curious to get your opinion like, you know, one, what are you telling your own marketers? And then two, what do you think the outcome is? And the outcomes can be one of three things either like one, marketers get replaced, two, you know, the marketers who are technical make more money. Or three, there's this big separation between technical and non-technical marketers and, you know, and that might have impacts on salaries and compensation. And I'm kind of curious like what you think is going to happen the next year.

SPEAKER_02

All right this is uh this is an interesting question. I'll tell you what I'm what I'm sort of seeing firsthand. One, if you are not selling a truly AI, you know, focused product so ignore the replets and the cursors and the clods and the and the open AIs of the world for a second. All right. Just like if you if you're not one of those companies, your growth rate hasn't increased as a result of the AI revolution. It's still very much you are in the trenches blocking and tackling, trying to get growth. And now what's happened over the last two or three quarters is you've adopted all these new AI tools, right? You were sold on the AEO dream, you adopted it. You were sold on some type of AI CMO, you know, this AI that's gonna like automatically manage all your Google ads and you bought it. The problem is none of those tools actually help you drive growth. They just make you more efficient at the tasks that you were already doing. And so the challenge that I see is there's gonna be very few market, all of these marketers are going to get very AI native. They're gonna, you know, post on LinkedIn incessantly about how I have done 200 things in one day. And that's great. But don't get me wrong there's value and efficiency and speed and velocity. The question then comes of like, okay, are they strategic enough to drive growth? What does that require? That requires understanding the unit economics of your business, understanding the TAM of your business, understanding the right ways to reach that TAM. And what I'm not seeing is any evidence that AI is helping you do those things better than you could have done a year ago. Now I say that also with you know the the caveat that the smartest marketers will use AI to figure some of that stuff out. But I haven't seen a sea change in AI's ability to drive actual growth. And that so so like how do I answer your question? What's going to happen in the last next 12 months? Honestly, the same thing that's been happening for the last five years, you're gonna see a lot of people pitch themselves as AI native AI first marketers. They're gonna get hired I literally had a LinkedIn exchange three days ago where somebody was like hey I was interviewing CMOs and this one candidate they didn't come with slides. They came with a fully built out website with full positioning and it looked brilliant. It was awesome and I was like wait this is a CMO right how many people are they inheriting what is their job what is the stage of the company oh it's PE back there's already 20 people and you're evaluating this person by their quality of developing a website right because it looks flashy. It looks impressive it's efficient it like gets the it just it grabs your attention I'm fairly confident Austin that that person is not going to be in their job in 18 months. And the reason is how fast you do something is not the driver of growth. It's do you strategically understand your business? And you know unfortunately that that is not changing with with AI and agents. So so so I don't think the distribution changes right if there's a if there if there's a curve of marketers and it's centered around you know there's a massive peak around like 50%ile like normal distribution and then like the top five percent of the people who run the world or or who run like the best marketing teams, that's not changing. It's still going to be the same. So that's my take um I think you need a fundamentally different way of looking at what is good in marketing. And I personally haven't seen yet one of these AI tools like change the definition of that.

SPEAKER_01

I think that's really well spoken. And if there was maybe a nugget that I took out of that it's like tools can help you be more efficient and more effective but intelligence and intellectual curiosity and thoughtfulness and str and strategy, all the ingredients that make current workers awesome those are the things that you can't replace with AI.

SPEAKER_02

Well no that's exactly right right I'll give you another example or or thought here all of the knowledge in in the world is now freely accessible to you right and yet if you go and ask even just in your client base like go go ask the 50 marketers that you work with hey how many of you can explain to me what correlation is and just watch what happens ask them like what does statistical significance mean? If I flip a coin a hundred times what do you think will happen right? Just like ask them these questions like ask them like hey what is capi and like 99% of people will not be able to explain to you what capi is why is that the information has been available for now like three years right like it's one prompt of so like you said the the intellectual curiosity and I'll also be honest this is also a structural issue and sometimes it does give me pause marketers are probably the least paid in corporate environments do you not I'm not talking about you. I'm talking about like generally right in a company like compared to engineers, PMs, salespeople, maybe support and marketing compete for like probably the the lowest paying jobs in my opinion. 100% and so there is a demand and supply here do we as marketing leaders as businesses attract the best people into those jobs? And if you don't then you give them all the tools you can but does that change the equation? So there's also a structural demand and supply issue here of like can you actually expect marketers to be the the best at this job? So so so that's my I'm speaking from the other side of my mouth here where saying like maybe yeah there are certain things where AI is just going to do a much better job. But I I actually think what's going to happen is there'll be a you know 10 or 20% of marketers who are going to crush it with AI and agents and they're gonna increase their you know compensation and the value add that they're driving for the business and everybody else is actually going to be the on the other end of the spectrum and you know nothing changes for them. So it's gonna be like the the the rich get richer the poor get poorer within like the marketing you know function.

SPEAKER_01

Love it.

SPEAKER_02

I agree. Love it. All right on that note I hope that wasn't too grim. We bring it to a close uh we're gonna get back to our weekly schedule so tune in next week no travel exactly hopefully it's getting warm in your part of the world and uh I'm starting back on my my half marathon training so everything is getting back to normal which I next Saturday let's talk half marathon I just I did a race last weekend and we can we can recast it and we can talk about your training as well we'll do that next time sounds like a plan. All right all right see you next week Austin have a good one bye y'all