# GTM AI Teams: The 3:1 Dev Ratio and Why It Works
Guest: David Arnoux — Fractional GTM AI Strategist · Founder, Gen AI Circle
Published: 2026-02-23
Series: Business AI Explained
Canonical: https://www.elementsagents.com/podcast/davidarnoux-ai-advisor

SPEAKER_00 0:00
Today, I'm meeting with David Arnaud, who's advising GTM leaders and founders on how to implement AI in go to markets. He's been a founder and marketeer himself and more recently also a builder using tools like Cloud Code and Cursor. So in this episode, we're going to unpack what he sees in go to markets these days and how and where the the biggest wins are with AI. What works, what doesn't in some of the major companies. So super excited to have you, David, and thanks for Cool. Good to good to be here, Vlad. Perfect. So usually, the best way I like to do this is to understand a little bit how you are defining your role today. You're working on a bunch of things. So could you just walk us through a little bit, like, how your day looks like and what you're actually working on these days? Yep.

SPEAKER_01 0:40
Yeah. Sure. My core, core role at the moment is helping startups and scale ups with AI implementation, specifically on GTM. Call that a fractional role if you want because it's just what I love. I love to be able to dive into GTM motions and try to optimize them with this new technology at the moment. So I sit inside companies and I actually build this stuff as well, not just advise on it. So it's kind of a cool dual vantage point which we can talk about a bit if if you want. Yep. Yeah. That's great.

SPEAKER_00 1:10
So talking about and expanding a little bit more on your the last row. Usually, how do you go about advising those go to market leaders? Who are you talking to? Could you tell us a little bit more about, like, how this partnership looks like, how you're actually helping them out as a fractional adviser? Yeah. So it's pretty boring, actually.

SPEAKER_01 1:31
A lot of GTM leaders or actually CEOs of scale ups reach out to me or actually of enterprise, and I just do a basic process where we run an audit of how the company currently works. And then we start to identify based on that audit, is there a technology gap, is there an infrastructure gap, what do the use cases look like, is there a skills gap, etcetera, etcetera. We identify the best use cases to start building a proof of concept around. We implement that one, and we get more traction to actually roll out an entire GTM AI engine. So, of course, that depends on the size of the organization, on the team I'm working with, and on how fast they wanna move. But that's that's basically how that works and then I sit as sort of a fractional role to basically build the backlog, drive the backlog, and then sometimes bring in partners for the building side or we find internal champions to do the to do the building. So that sounds all very corporate y speak, but essentially, we're just trying to get proof of concepts out and build on stable infrastructure as fast as possible with this exciting new new tech.

SPEAKER_00 2:38
What do you think is today's, like, number one priority for, like, CMOs, CROs at, like, big companies? What's, like, the one most recurrent project that you see, these people,

SPEAKER_01 2:48
like, requesting? It it depends on the strategic orientation of the organization. It's usually not CMOs. It's a lot of just c suite in in general. So if we're talking to ops teams, they're looking for operational efficiency. And if we're talking to more, like, customer facing roles, chief revenue officer, chief marketing officer, we wanna increase pipeline. We wanna increase conversions, increase ARPU, increase retention. So the Mhmm. First question to ask yourself is really, like, what's the one metric that matters at the moment that we're trying to optimize for? AI implementation, it's not about the technology. It's really about the it's really about the use cases, working use case by use case. And typically, you kick off with a high impact, relatively high ease use case. And then after that proof of concept or that quick win is established, then we can go into typically the very high impact, slightly more complicated slightly more complicated use cases. And it really depends which motion they wanna focus on. Is it on awareness? Is it on demand generation? Is it on deeper in the funnel on conversion? It it's really it depends kind of

SPEAKER_00 3:50
So usually, they already know which level of the funnel they want to influence. They already know which kind of KPI they wanna impact, and then they ask you to do, like, the audit to come up with, like, ways to to improve those metrics?

SPEAKER_01 4:04
No. Everybody wants to optimize everything at the same time. So that real that usually takes part in the that's in the audit phase. What should we focus on? Where's the biggest bang for your buck? And it's really like a two by two matrix with impact and ease. And, typically, the higher you are in the funnel, the further you are away from product, the easier it is because you have less constraints, internal or external. Mhmm. So if we wanna go really easy but high impact, we go on demand generation, something customer facing high in the funnel. If wanna we have a little bit more impact, typically, we go a little bit deeper in the customer journey or in the funnel. And it's really just a question of the leaky bucket, where is the bucket the leakiest, where are we leaking most opportunities, more the most money, what should we be focusing on. But no, they don't usually have the answer. They want everything optimized, so that's usually a discussion or you workshop it. What should we be focusing on first? And I'd say that this is important for any type of adoption, whether it's for personal AI implementation adoption, team AI implementation adoption, or or company. It's really about what's the first workflow or use case that we wanna focus on first that drives the energy, it drives the focus, and it's a step by step iterative process. Don't try to boil the ocean. You know? There was a big trend around Cloud Code recently, using Cloud Code for personal productivity. I think we all jumped on the bandwagon. I'm running a 25 multi agent framework through Cloud Code at the moment. That's the new that's the new GTM expert. That's the new marketeer. It was difficult the first three, four days. You know? Mhmm. I needed to onboard myself to this new way of working and onboard all of the use cases, the context, do all of the MCP integrations one by one. It's the same thing with any project, any change management project. God, I sound really sound like a a consultant now, but it's it's really about building up an urgency around a specific use cases and taking it use case by use case.

SPEAKER_00 5:53
Yeah. Yeah. I think we're gonna untangle a little bit and get into the specifics of these workflows and, you know, what all of these things mean in in Yeah. In practice. You mentioned buckets and leaky buckets also. Could you maybe tell us a little bit more about, you know, how you're thinking about those major buckets of things that you can actually do with AI today? Yeah. Okay. So I would say at this point, like, almost anything's possible at quality and at scale, and it's kinda scary, to be honest.

SPEAKER_01 6:20
So it really depends where the strategic focus of the business is. One that does keep keep coming back on the b to b side typically is how can we build a better engine behind which customers are right for the taking at the moment. You know, in b to b, typically, the number is that only three or 4% of your total addressable market is ready for your solution at any given point either because they have other imperatives or they've already bought a technology or it's just not the right time. So it's identifying who's right for the taking at which time. And typically in b to b, there's also a lot of outbound motions. There's a lot of requests for outbound motions and or account based marketing automation. In b to c, it's usually about how can we create a bunch of assets for our performance campaigns, how can we make those performance campaigns auto improve themselves? And maybe a third one is still around content marketing and showing up in SEO and in GEO, generative engine optimization. So in LLMs, those usually show up both in b to c, b to b. So that's kind of like pattern matching. Those are like the two, three that keep that keep coming up. Mhmm.

SPEAKER_00 7:32
Okay. Honestly, I'd love to to dive a little bit deeper also in the ad creative workflow and, like, this loop because I think a lot of people are very excited about this space of analyzing images and tying this back to, like, the, you know, the conversion rates and, like, impressions and so on. Can you maybe tell us a bit more about how this looks like? Do you actually, like, just throw, a Gemini model at the the creative to analyze and connect it to the metrics? How does that work in practice? Okay. So we're gonna geek out specifically on this one. So first of all, it's a question of attribution. How do you define what success a successful campaign? Are you still relying on last touch attribution

SPEAKER_01 8:10
from the platforms? So are you relying on what Meta says as a conversion and on what GA says as a conversion? For example sorry. Not GA, but just Google says as a conversion. So first of all, your definition of what a conversion is. Typically, sort of the most sophisticated today, it's like, okay, multi touch attribution done correctly, marketing mix modeling done correctly, and if you unify those two together, it's called unified marketing modeling. So are we actually treating conversions correctly? That's sort of sort of number one. Mhmm. Worst case scenario, you're not doing it correctly, but at least you're getting last last touch correctly. What you can do there is you can start so feeding all of that data through APIs into a central engine that's essentially optimizing your campaigns for you. I don't know if you saw there's a really cool tool that came out, I think it was like a month ago, it's called GoMarble. And Mhmm. GoMarble is basically a performance marketing optimization engine, which I believe you can still white label and all it's doing is it's looking at the campaigns, it's looking at the assets that are associated with those campaigns, and then it's making recommendations for how the campaign should be optimized, whether it's on the text or whether it's on the creative. Now feed into that an engine, like a tool like Weavee or Dolly three, whatever you're using. So Weavee is sort of a Mhmm. Image orchestrator tool to create image generation workflows. DALL is specific DALL E three is specifically an image generation LLM. So you can auto generate visuals from winning concepts. Mhmm. I don't think a human should be doing that. I don't think a human should be going through Excel sheets or campaign reports to find out what are the most successful campaigns, and then based off of that, making new recommendations for what we should do next. It's just too much data. Right? It's also too boring. Mhmm. And what you can do then is you can have auto generated visuals based on winning concepts, a kill switch to pause bad ads before they waste budget. And so the secret here, it's really a feedback loop so the system gets smarter with every dollar spent. It refines its own prompts based on what actually converts. And essentially, you're turning like a week of creative strategy into a fifteen minute autonomous loop. Now you might be thinking, yeah, but the the images I've seen online, they're not production ready. They're not production quality ready. Actually, they are. You're just not doing it right or you're not doing it with the right models. You can make amazing images today that are production production quality ready. And you might be thinking Mhmm. Quality of the copywriting won't be as good as if a human does it. I would argue that most campaigns actually don't have good quality to begin with, copywriting quality. And if anything Mhmm. It's it's working off of the initial prompts that you give it. So if you wanna be more bold, more provocative, it'll go off of that. More creative, it'll go off of that. If you wanna be standard and compliant, it'll go off of that. So I I hope I'm making sense, but you need to imagine this virtuous self improving loop for performance marketing.

SPEAKER_00 11:20
Yep. No. I think it I mean, it makes total sense, and I tried to also share, like, tools, landing pages, to give more, like, a better understanding of, like, how they actually work. You touched on something at the very beginning about the quality of the data that you're starting all of this from, and we all know that garbage in, garbage out is kind of the the, the buzzwords. What are the stuff that, you know, people should try to fix before they get into, like, sophisticated and fancy AI projects?

SPEAKER_01 11:51
I think that with regards to sort of AI implementation, GenAI implementation, Two years ago, like, prompting was a competitive advantage. Today, it's become kind of table. So Mhmm. The advantage now is those two who have pretty clean data with kind of a strong semantic layer on top of it and those who can actually delegate answers. The the teams that I work with are the ones that are moving the fastest now. They're not the most technical anymore. They're usually the teams or the organization of the people who have learned to delegate to humans in the past and who are now applying the same muscles to machines. And I'll explain what I mean in in a bit. The quality of your taste and eval evaluation, but also the quality of your repository is now dictating the quality and the quantity of the work that you're generating. So what does that mean? Like, level zero to two, it's like basic chat, cloud browser, desktop, GTM application. It's like ad hoc research, quick copy drafts. Right? Something like that. Level three, four, I would say it's like you've got your projects and your connector set up. So you're connected to your CRM, you're connected to your Marketo, you're connected to your HubSpot, you're connected to your Salesforce. So you start to have persistent context across campaigns, live data feeds. So, of course, if the data in there is complete junk, then the output will be complete junk. So there's usually a lot of work there around data unification, a little bit of data cleanup. Although LLMs are good at sifting through messy data, you can actually train it as you go. Right? You can actually train it on messy data and train it to train you and to clean up your data. That's been the case a few times. Then it gets exciting. Then you've got, like, levels five and six, let's call it. That's where you start setting up your internal skills, repeatable workflows, outbound sequences Mhmm. Automated reporting. I mean, today, it's possible to do, you know, business reporting at the click of a button based on certain data feeds and based on certain rules and constraints that you've given it. And and the output is either a report or a deck or a spreadsheet. And that used to take us hours as humans. And I don't think as humans we were ever meant to build spreadsheets and build decks and sift through data. And now it's possible with us humans in the loop Mhmm. To validate that the output is actually correct. It's possible to do that at the at the click of a button. If you've reached that level I was talking about before, that that level three, four where your repo is actually clean. Yep. And then after that, you you start going into like full stack GTM, programmatic SEO, ad creative loops, customer CMS management, things like that. Mhmm. And the level even after that, that's the cutting edge that I'm seeing is twenty four seven optimization, autonomous campaign management, and even creative ideation hand in hand with the the engine, not just coming from the not just coming from the human. And that's the brave new frontier, I would say, that's that's really cutting edge at the moment.

SPEAKER_00 15:09
Yep. You you you used a couple of words that I just want to clarify and maybe, like, expand on a little bit with you. Three words, semantic, repo, and evals. So just to make sure everyone is on the same page, like, here's semantic. We're basically talking about a database where you can look for a similarity when you do a research. So instead of looking for a specific ID in the database, you're looking at something that is similar to a sentence. So the best clothing line for sports, and then it's gonna return you, like, the words that are, like, the most closely like, the most similar to that query. So that's semantic. Repo, we're talking about GitHub repo where some people are using it with files where they have maybe, like, some prompts that they can reuse to to basically, like, like, ask questions about their best practices. So let's say you have, like, ton of voice in a specific file in your repo. You can just use that code and, you know, and draft an email basically saying like, hey. Look up my ton of voice file, and you can draft an email that is consistent to your branding guidelines. And then finally, you talked about evals. And essentially, evals are just it's just a term to say, okay. Whenever there's an output, do we know how to evaluate how good it is? Is it good? Is it bad? Can we, like, rank it from, like, one to five? Can everyone do it? Is it only one person and so on? Is this is this how we're thinking about this? Is this how you're looking at this as well? Yeah. Yeah. So semantic layer, a 100%.

SPEAKER_01 16:41
On repo, yes, it just doesn't have to be a GitHub repo. It can also be whatever repository you're you're using. By the way, teams and organizations who invested in clean data lakes, clean data warehouses are probably, like, have a big, big smile nowadays because it means they have a very clean repo to tap into with very strong data unification. And I think the most important is probably the it's the term evals. So I'll give the complicated answer and the sort of simpler answer. Here's the dirty little secret evals. Most marketing teams couldn't evaluate their own human produced work before AI showed up, would say. They were already shipping mediocre campaigns with no feedback loop, no systematic way to say this is good, this is bad, here's why. So AI didn't create sort of the eval problem, it just made it visible to us all. Just think about it and I think that the greatest marketeers today and the greatest business people or users of this technology are people who actually have good taste. Eval is like having good taste. Is it do I wanna publish this? Yes or no? That's basically yes. Is this good? Yes or no? It's evaluation. Could be as simple as, yes, this is the blog post that I want to write with the help of AI. Yes or no, ship it. Right? And we're still the human in the loop that decides, yes, this is something that I wanna push. So I think a lot of people are have increased the quantity of the bad eval that they had in the past or increased the quantity of the good eval or the good taste Yeah. Good taste that they had. And that's why I think things like co cultural curiosity, analogies, historical metaphors. Just having this cultural curiosity is more important than ever before. So that you know, you know, you you maybe have a good taste for what might be good, what not be good. Having a lot of this vocabulary around art, cultural movements, maximalism, brutalism Mhmm. You know, Bauhaus. I don't know, like, Ernest Hemingway style of writing, Josh Pirac style of writing. Just by having those cultural references, you're able to also prompt a lot better and and guide the campaigns and whatever you're doing a lot better. And that's where branding comes back into play and creative copywriting come back into play. So, you know, you've got local maxima, global maxima. Today, I think that the evals cooked into the LLMs are really good at local maxima optimization, slightly improve this campaign, slightly improve this copywriting, slightly improve these hooks, slightly improve the quality of the account based marketing strategy that we're doing. However, on the global maxima side, something like, hey, we want to be way more provocative and bold because we want to stop the scroll. I don't think you'll get that from the LLMs yet unless you're asking for it and that's where the human is still in the loop defining that that that brand aspect or that that global maxima of where you wanna of where you wanna go. I don't know if I'm if if that's making if that's making sense.

SPEAKER_00 19:46
But No. I think it makes sense. I was thinking that maybe taste is, like, a super critical skill that you either have it or you don't, or it's something that you can acquire over time by just being curious, traveling, accumulating experiences. So, basically, go touch some grass to become a better marketeer. Yeah. Could be, like, an interesting way to to, like, think about these things. Yeah. Or not because

SPEAKER_01 20:08
there was these there was this work done a while back that Comic Sans MS is actually one of the greatest fonts to get conversions online or that ugly websites actually convert better than pretty websites. So actually it could be that bad taste is good taste. I don't know if you remember this website links cars or even just go to booking.com and you still see the user interface that booking.com has. It's not pretty but it's good taste in the sense that it works. So it also really depends which industry you're in. And in that sense, maybe the LLMs are better than us at identifying what works, what doesn't work. But sometimes you do need to pull it out of the sort of pull it out of its local maxima to try more creative approaches. I hope that makes sense.

SPEAKER_00 20:48
I think there are different reasons why some websites do it like deliberately to have like a like a like a bad copy just because it's nostalgic. It's always a trend, nostalgia.

SPEAKER_01 20:58
You just have to go back thirty years and just reuse that. And by the way, Vlad, so I think that there's definitely I can't remember if it if it was, you know, it was in one of the brain trust that we had, we did twenty twenty six predictions. And one of the predictions that came out was they were gonna we're gonna start to have two Internets. We actually already have two Internets. One for humans, one for AIs. So I don't think a human is meant to pick the best flight tickets for a trip that they're about to take. I don't think a human is meant to evaluate. No. It's like, is a human meant to evaluate? If you're looking at MarTech for example, are we as humans meant to evaluate 150,000 different MarTech possibilities or tools that we could actually be using? And even if it's something really simple like lead enrichment. Are we as humans meant to sift through all of the comparative tables that make Mhmm. One lead enrichment tool better than another lead enrichment tool? No. It's just too complicated. So then we revert back to Daniel Kahneman thinking fast and slow. We think fast. We use gut. Mhmm. We use a referral or we let somebody do a a very extensive assessment, and then we just want the executive decision. Right? Mhmm. So I think actually, there is gonna be a second Internet that's gonna be an agent to agent Internet, and it's already happening. Mhmm. Like, if you're setting up Cloud Code today and you need to use an MCP for something that's not official, what are you gonna do? You're gonna say, hey, LLM. What's the best MCP I can use? It's gonna go look at the Internet a little bit. It's gonna go look at some, it's gonna go look at some, some goo Google queries, and it's gonna come out back with, five or six GitHub repos that seem to have the most stars. Right? Mhmm. And then you're gonna rely on that decision. So more and more, I think we're gonna rely on the agents for everyday mundane or not so mundane decision making Mhmm. In the sense that it's just too much information for us to handle. And more and more, I think there will be a second Internet where it's my agent interacting with the service provider's agent to make that decision, even a purchase decision. So probably in the future, maybe you'll be on your, you know, ChatGPT, but that's 2023. You'll be on your Claude, and you'll be asking Claude to pick, you know, the best tool for you, and maybe it'll have a wallet, and it'll actually buy that product for you directly, whether it's, you know, a piece of grocery, whether it's a flight ticket, whether it's an off the shelf tool, whether it's an API call from a service provider. More and more, there's gonna be the human Internet, which is gonna be for doom scrolling, watching TikTok videos, you know, being on YouTube, having a bit of fun, and then the a to a Internet for much more complex decision making processes. So then the question is, all the marketeers are thinking, how can I game this? And I've actually got a buddy who's building a tool that allows any website today, any SaaS to be agentic ready so that an agent can actually navigate it. And then in SEO, the SEO experts, they're trying to gain the LLMs by building listicles because LLMs are not smart, and the first thing they do is they look for top 10 lists and they look for recent ones. So there's a programmatic SEO strategy of building a bunch of listicles at the moment and you try to, refresh them continuously. We're actually doing this internally and it works. Right? Because the LLMs are not smart at the moment. SEO is smart. Geo is kind of silly or crazy teenager. Of course, they're trying to patch it, but they're not patching it fast enough. So, you know, the new UI is actually no UI to a certain extent in the future. It's gonna be a lot of the decision making, is gonna be taken off your plate. And I'm so happy this is happening because ever since the Internet exploded, we're being asked to be travel agents, car buying agents, insurance experts. We've removed the human from the loop, but we're the human in the loop, and we're asked to go on Google and, you know, for the last ten, fifteen years, make these decisions ourselves. Go to g two, go to MarTech comparison engines. Mhmm. And and it's just not what we're meant to do as as humans. Right? Yeah. As humans, we're meant more like creative puttering and Mhmm. And sort of big decisions. So Yeah. So I think this should be an interesting sort of, you know, change, paradigm shift.

SPEAKER_00 25:18
Yeah. I was having a conversation with a salesperson at UiPath last year, and he was saying that salespeople won't disappear because of AI SDRs, but because of procurement agents. So, basically, at some point, procurement is gonna be so standardized with agents that will be kind of objective in evaluating all providers that no matter how good you are of a salesperson, you're not you're not gonna be able to compete because these agents will, like, be completely insensitive to your sales skills, essentially, which is kind of an interesting way to to look at this. Fair enough. Mention you you mentioned a couple of skills that people should be able to have these days. One of them is obviously taste, being resourceful, being open, curious, and so on. But then, obviously, we talked about repo, eval, semantic, all of these skills. So if you think about building out those capabilities in go to market teams, what's your, like, dream team? How do you see, like, those teams to evolve over time? And what's the ideal team you think every b to b company, like, should be planning to have, like, this year? Okay. That's a tough one. That's a really, really tough one.

SPEAKER_01 26:34
Oh, you we're talking you only wanna talk about b to

SPEAKER_00 26:37
b b? Because you mentioned If you think it applies to both, then let's do both. If you think they're different, let's go over through each of them. I think it's interesting for Okay. So

SPEAKER_01 26:47
ideal b to b GTM AI team, I think that you really need a ML engineer, AI ML engineer, a full stack developer, and then some GTM strategists sort of layered layered on top. I think that's kind of the basic the basis at the moment. Just to keep things really, really simple. So I remember when, like, growth marketing became a thing fifteen years ago, we always, there was this growth engineer profile. Mhmm. And I think that's more important today than ever before. You need a dev in the team around AI orchestration and and potentially also around architecture. Mhmm. So you're looking at a full stack dev, an AI ML engineer, and potentially, you know, somebody really good at orchestration, whether using an off the shelf platform like n eight n or somebody who actually can actually hard code this stuff. Now, that doesn't mean that they that needs to be a developer role. Right? They can also vibe, figure it out along the way. Mhmm. But in that sense, I think the paradigm was changed quite a bit. I actually think today, potentially, devs could potentially make better oh, I shouldn't say things like this. Are the necessary co pilot for the modern day marketeer because we're talking more and more to agents rather than humans for the decision making process. Does that Yeah. Does that make a bit of sense? If we take maybe, like, one client

SPEAKER_00 28:22
that has been, like, implementing AI like crazy and has been incredibly successful at it. Maybe we can just go over, like, the team that they had and why do you think that worked.

SPEAKER_01 28:34
One very strong GTM strategist strategist. Mhmm. It could be you or me. Usually, if it's a small team, it's like the founder Mhmm. Or it's the head of growth, for example, or or the CMO. And next to that, you have a technical ish team. So it also depends on the size of the of the company that you wanna have. So the one I'm thinking of now, so it's a 400 person company, a very, famous European unicorn and scale up. And in the GTM team, the ratio of developers to marketeers is three to one. So you have three devs for sort of one marketeer. Crazy. Well, I don't know if you wanna call it devs, but AI native sort of engineers, I would say. Three to three to one. And they're working mostly on observability of what we're of what we're releasing, making maintenance to make sure that all the workflows are working, etcetera, etcetera. But the strategy side is the, yeah, is the sort of 25. If you have a ratio of three to one, it's 25%. One strategist for three sort of a for three sort of builders. Having said that, the marketeers can also build themselves up into these automation experts as well because it's not that complicated anymore. Even me, like, I have a really weak dev background, but I'm able to run sort of 25 workflows simultaneously at the moment that work quite well across our different ventures. And then from time to time, something breaks and Walid needs to come in and help me to sort of fix it because the GitHub repo I used was was not good and didn't have enough stars.

SPEAKER_00 30:19
But

SPEAKER_01 30:20
that that's kind of what happens. They're there a little bit as the as the adults adult in the room. Mhmm. So I would say you have, like, one GTM strategist who sets the direction, owns the taste, owns the evals. Then you have, like, one builder. It could be a full stack dev, could be a technical architect, someone who can actually ship things, connect the APIs, set up the infrastructure, build the automations. And then, honestly, like, this is the paradigm shift. Maybe it wasn't a three to one ratio, maybe it's a two to one ratio. But Mhmm. I think that the AI is the third team member. Mhmm. And that's not a joke or an exaggeration for Buzz. Yeah. In the teams I've seen move the fastest, in the teams, it's like literally two or three humans and then a stack of agents doing what used to require like eight, twelve, to maybe even 20 people. Like, I'm working with one enterprise CMO of it at the moment and his core team for AI implementation, it's essentially him as the strategist, one technical architect, and then a bunch of agents and automations doing the execution. Mhmm. You know? Okay. Yep. And, like, how do you and then how do you keep control? Mhmm. Like, on my I have I have a ultrawide screen Mhmm. That I bought, like, two, three years ago. And everybody get an ultrawide, please, if you can't put a bit of money into it. And get your get your boss to unlock a budget for it. So I have like my clot code on the left, two thirds of my screen. And then on the right, I have what's called my my lookout. It's kind of like a, you know, a house. It's where I keep a lookout on everything that's happening, and it was built through ClotCode, and I just basically built my cockpit up there. The the and everything's plugged into it. Everything's plugged into it. CRM's plugged into it. Our databases are plugged into it. The meeting transcripts of the meetings that I'm allowed to transcribe are plugged into it. Notion's plugged into it. The the the the the Google Workspace is plugged into it. Emails, calendar, Slack conversations are plugged into it. Not for clients because I'm not allowed to, but Mhmm. More for our ventures, and I have oversight of everything that's happening. Absolutely everything that's happening. So that's where you maintain control. But then it's the AI that's doing all of the execution work and sometimes even a bit of the decision making. So it's really that third person that's equivalent of a 100 people. And I think that's the biggest mindset shift for GTM leaders right now. Your job is no longer to manage a team of 15 people producing content and campaigns. Your job is to be the editor in chief of a small team plus a fleet of agents. Mhmm. And the skills that matter the most isn't technical, I would say. It is ish, but it isn't technical.

SPEAKER_00 33:10
It's taste. It's knowing what to ship, what to kill. Yeah. I I there's this guy who posted on Twitter that building with AI is basically like the new product manager, and fine tuning and training LLMs is like the new dev work. So, basically, a PM has good taste, knows what to kill, what to ship. So all of this thing around road mapping, basically, like, the development work linked to that falls under the PM, and so all of these AI native people that can actually ship and build. So you basically have, like, a technical PM across each function, or you have rather three for one strategist steering all these people to, like, execute. Yes.

SPEAKER_01 33:48
Just think about content marketing. Content marketing is like a programmatic job now. Even if you're doing content marketing without having a dev in the in the room, you're using PKI, for example. So you're actually already doing programmatic SEO. You're already programmatically creating content Mhmm. Across the board. Mhmm. The last two, three things we haven't really been able to automate yet are in person events and meetings, podcasts like this, and really rich content like video, even though they're already trying. And at b to c level, the uses UGC generated content is hard hard to discern from from actual human content. And, you know, if anything, I think that authentic we will come back to authenticity and everything that's sort of human, that's high signal, that's hard to build. Friction is value now. So everything that has a lot of friction to build will have a lot more value in sort of authenticity and and brand building. Mhmm. So we'll we'll come back to that because everything that's easy can be automated and will have a lot less value. So I think it's gonna be the revenge of the nineteen fifties, the real brand people, in person meetings, in person gatherings, closed communities, people you trust, referrals, etcetera, etcetera. Think that will make a bigger comeback. Even receiving something in the main that you can actually touch, that's so fantastic nowadays. So there's definitely a moat and a moat there still. Everything else is becoming kind of a commodity. Mhmm.

SPEAKER_00 35:11
Yeah. So, yeah, I think there are, like, so many different topics. This could go, like, in so many different directions in terms of how can you actually, like, leave human breadcrumbs along the way and just, like, you know, almost leave typos on purpose to show that it was a human or or or just, like, have those long form content where it's obviously, like, human or yeah. I think it's it's it's very interesting, and and we'll see what people actually care about. And only time will tell because maybe there are certain things that we take for granted or that we assume to be true, but maybe they're not. I think we'll only know over time. Just to dip a little bit deeper into, like, Claude, you mentioned that you learned everything in three, four days, which is fairly fast when you consider, like, the power of the tool. And probably it's still, like, something that you need to learn and improve on, like, continuously. But when you think about these topic topics around buy, build, hire, contract, how do you think about these topics with clients? You know, do you tell them, okay. You need to put staff three engineers in each team so that they can, like, build out all these all of the things that we need? Do you suggest to, like, go with, like, a consultant in the first couple of months? Yeah. Can you tell us a little bit more about how you are thinking about this, if you have any framework or any approach

SPEAKER_01 36:29
that you apply? Yeah. Yeah. It's really simple. I I posted about it on LinkedIn a while back. It's a really good question. So there was a simple framework that I talked about a while back. Basically, on the decision to buy, it's when you have a proven workflow, the team isn't technical, and the speed matters. So, like, play for enrichment, Gamma for Decks, HubSpot AI features, for example. That's where you should sort of buy, Mhmm. You know, a solution off the self solution. When you wanna build, if it's a core differentiator, proprietary data, feedback loops are needed. So, like, programmatic SEO, ad creative loop, custom ABM automation, and then when to hire, it's really on the, you know, really long term important stuff. One AI architect, fractional or full time, not 10 prompt engineers. So data unification layer, repo architecture, strategic oversight, that's when you should be hiring. Of course, can use a freelancer for the first three, six months. And usually, in the projects I do, like, they have internal talent, internal champions, we use that, of course. First, we put them in the task force, then we roll it out, and then people's job roles change after a while. And if we don't, for example, I work with you for that. We work together with the client on getting those first POCs out the door and the first base layer established, I would say.

SPEAKER_00 37:49
Yeah. Okay. Maybe just one last question for you, more on a personal level. You live in the countryside. Yes. You travel a lot, and you're still, like, advising on AI, like, some of the, you know, fastest growing companies. Do you think we can do all this shift remotely, and what's the importance of, like, being in person in the this entire transformation of, like, a bunch of industries?

SPEAKER_01 38:14
Yeah. So that's a tough one. It depends on the company culture. I like the hybrid approach, basically. So it's cop out answer. 90% remote, 10% in person for the big creative moments. Mhmm. That's how I see things. Okay.

SPEAKER_00 38:28
Perfect. Hybrid it is then. Perfect. So

SPEAKER_01 38:32
where can people find you? Where should they reach out if they wanna hear more about anything that you Just on LinkedIn. Check me on LinkedIn. I post a lot there. I also have a newsletter called heyarenew.com. And if you really wanna be cutting edge, try to join our Gen AI Circle, www.thegenaicircle.com.

SPEAKER_00 38:50
Awesome. Perfect. Thanks, David. I'll let you jump into your other call.

SPEAKER_01 38:56
Take care. Thanks, Vlad. Thank you. This was cool. Laters. Yeah. Thanks. Ciao.