Who Controls Your AI Marketing Stack? (AI Sovereignty)

Does your experience with AI sound a little something like this? You've been prompting for 20 minutes. The output is polished, confident, and completely useless. That's what happens when AI doesn't know anything about your business.

HubSpot's AI works with your actual customer data, every interaction and every signal your team has generated over time. So instead of prompting and hoping, you ask once and ship it. Check out HubSpot.com, the agentic customer platform for growing businesses.

Did you see that talk with... You like CNBC. Did you see the Alex Karff from Palantir earlier this week? I did not see it, but it was going viral on Twitter. And I did not read about it on Twitter either, other than I saw David Sachs and a few others, Mark Andreessen.

He just responded today. Yeah. So, okay, here, let's talk about it. I'll give you some context here. So the real question is, who controls your AI marketing stack?

Okay. So when you have... When you have Anthropic that just updated their cloud design yesterday and they told Figma they're only going to have a basic version of it, and now it's pretty damn good and Figma's like, what the heck? Or when you have Anthropic saying that they're about to get into drug discovery and all these things, right?

What David Sachs is saying is that if you don't have any AI sovereignty where you start to give all of your context and all of your data to these companies, then... what, like, ultimately, you as a company, me as a company, we want to have our data protected, right? You need to have some tempo, you need to feel secure, you need to have some type of sovereignty from these models out there. Because if they, if you start to give them everything, then they're just going to start to build whatever they want, right?

Same thing like, oh, Harvey, which is that legal company that raised a bunch of money. Oh, Anthropic is now illegal as well, right? And so the thinking here is that Alex Carpenter is saying, look, you know, these are great models, these frontier models are great models, but they did a partnership with NVIDIA. So Palantir and NVIDIA, and they want to have these really open weight models, right?

And so there's been talk about how GLM 5.2 is really good. It's like Opus 4.8 level, right? Which is, as of right now, these are the really good models, aside from Fable 5. But again, like with, I'll give you another example here.

Claude released something called Claude Tag, right? So Claude Tag allows you to chat with your Claude inside of Slack, for example. You can have different connectors. The problem with that is once you come off of it, you've already fed all that context to Claude and they get to keep it.

You don't get to take it with you, right? And then they have limited connectors. You have to use their expensive models and things like that. So you don't have sovereignty versus if you start to put some of your stuff on open weights, then you save a lot more money.

Plus you get to control your stack. And then you don't need to worry about anybody coming in to start a competitor against you the next day. So that's why I think AI sovereignty is not just important for your marketing, but also important for your business too, because you don't want to feed them all your data. Yeah.

And the screwed up part about Figma, I was skimming the David Sachs posts. It was either yesterday or the day before. I believe their product person was on the board. Anthropics product person was on the board of Figma.

Oh, I didn't know that. He left three days before. I believe it was three days before they released their version of a Figma competitor. That's so bad.

That's bad. And I'm not saying the product person intentionally did it. Figma's, uh, Anthropic's a big company, so it could have been someone else working on it.

Um, but I do understand the issue here. I believe what we're going to end up seeing a lot of marketing data is on the cloud. You're seeing it in either AWS or Google's cloud or Microsoft's cloud. I believe they'll come up with solutions that'll help companies, uh,

Yeah. He'll allow us to, you know, get access to our data, make decisions from it. And we don't have to worry about people stealing it all. Yeah. And they all have NVIDIA chips.

So Microsoft actually launched something yesterday where they're putting 6,000 engineers against it. It's called the Frontier Company or something, but 2.5 billion invested. And it's all to help people deploy. And one of their key tenants there is, hey,

We're not your data is your data. We're not going to build on top of it. And that's really important, right? Because if you're spending hundreds of thousands, if not millions of dollars on this stuff, it's important.

And when we talk to people on the single brain side for these managed deployments, that's the number one question we get asked a lot. Hey, do you guys own the data or do we own the data? And we say, no, you own the data. We'll compound with you.

You get to own your memory and we'll build with you. Right. And they're like, oh, OK, sounds good. Yeah. No, people don't want you to own their data. When I saw the David Sachs expose, I literally sent a message to my team.

Be careful what we're sharing with open AI and Anthropik. Yeah. It's big. Like, and don't get me wrong. I use it all day and it's my default, but I just, I trust it less and less.

And then you see like open AI is like, Oh, considering giving 5% stake to the U S government. And then you see, you see now the government's controlling, you know, it's like export controls. Like, Oh,

You can't launch this one. It has to be slow launch. And then Fable 5 came back, but it's a nerfed version. Like I was using it the other day and it found some security vulnerabilities in one of my things.

I was like, hey, do security sweep, right? Once it found it, because it used the keyword security, it bumped me back down to Opus 4.8, right? And so these things are, I'm just like... You know, it's becoming a little dicey to the point where you want to have your own sovereignty.

So having your own NVIDIA chips or having and then using these open weights, it's going to save you a lot of money. And I think you can sleep better at night. And it's going to help with your marketing, too, because you know someone else doesn't have your data and they can't just steal your messaging. Yeah.

And I believe all of this is just going to cause NVIDIA to grow even more. Yeah. Yeah, you know, there's a thinking. I was listening to Gary Tan and Naval on a podcast today.

And it's like, oh, I think Gary, apparently Y Combinator did some math on this saying, oh, maybe this stuff is, you know, still 90,000 times more upside to human usage with these things. 90,000 times? 90,000 X. And I'm like, that's a lot.

So anyway, I don't know how that pans out, but I do believe that we're still in like the first or second inning right now. Yeah, I think it's really early. Yeah. So anyway, I mean, that's how I'm thinking about it.

I mean, like when I sit in these AI chats, like we're all talking about GLM where you can use open router for this, by the way. So if you want some AI sovereignty, you can sign up for open router. You can sign up for GLM 5.2. It's like five to 10 times cheaper than using like a cloud.

I still just for myself, I just buy a bunch of cloud max and open AI subscriptions for myself. Then I go to town on it. I just keep switching the keys. um but for your company if you want predictability there i would say you can do that and then i am prepared to start paying um a api cost again for fable 5 because as of next tuesday it's going to switch back to usage only and then i i made a little video game the other day did you see that um no did i made this i made this video game where it has like have you played like age of empires or starcraft or any of those games

No, I played back in the day. Do you remember Star Wars? It was like a pod racer. We have Anakin Skywalker on like a little, not a race car. What do they call them?

Like the pods, the pod racing. Do you know what I'm talking about, Noah? Dude, you play the most garbage games, man. I saw this guy playing like Fruit Ninja before and all this stuff.

No, you've never seen me play Fruit Ninja. I have seen you. play from ninja. I've never done one for ninja. Play from ninja. Okay. No. Sudoku, yes. Monopoly, yes.

This game, here, I'm going to, here, Neil, just take my phone. I'm going to explain it. So Neil's looking at it. So that's basically like a city, right?

That's dummy data and it shows your city and what's on fire and your city and what needs help and what needs attention. And then you can basically create quests for your employees to take on and you can award bonuses for those quests. Is this real data of your company? That one's dummy data, but I just added real data this morning.

And then now it's building right now and I'm just like, okay, well, some stuff I want to keep private from some people in my company, so maybe I want to shade it out, but then I want to give them quests. And then maybe a quest is repeating and you can award them in stable coins and smart contracts. You can award the money, you can escrow the money and just award them. But then

imagine the world can just grow right and then you can have other maybe i want to trade with your business or maybe someone from your business that needs extra work and then they're working on quests in my business right so my point of saying this is that these models are so strong where you can start to make previously stupid things like this that actually can help your company because if you gamify the company it actually makes it fun and you can see who's doing a good job and who isn't doing a good job so anyway that was a little tangent for me um but yeah it's only gonna get crazier You should make the fire bigger in the game so people can see it.

Like burning? You know, when I open it, it's actually on fire. Oh, cool. And you can see people moving around and stuff. And then the stuff that's going well, you see gold coins coming out of it.

Yeah, you should see who's working on putting the fire out. And you can create a little character that's throwing water against it if they're actually working on it. Or let's say your character becomes really more powerful. You level up in the company, right?

You start to look cooler and cooler with your gear. And so I think there's a whole thing because I looked into this. Do you know how much World of Warcraft was making at its height like per month?

Profit or revenue? Revenue. I don't know, like $500 million billion a month. Wow. That's a good guess. This is a massive company. World of Warcraft at its peak was making $180 million a month.

And so I'm like, one day when I have more time for this, if you make a World of Business Wob instead of WoW, I'm like, how much is that going to make? Because at World of Warcraft, you're charging $12 a user per month. Was that it? Because didn't Microsoft buy Activision Blizzard for like $80 billion or something crazy?

They did, but that was when World of Warcraft was already out. Oh, so they have more games than just Warcraft? Way more games. They have Call of Duty.

They have a lot of franchises. I just know Activision Blizzard as Warcraft. Yeah. Well, no. Blizzard's also Diablo. It's also StarCraft. Diablo is the game Elon plays?

Yeah. Yeah, yeah. So he plays Diablo 4. But anyway, all that to say is, I want to say something real quick. I noticed something. I've, from a recruiting standpoint, because we're still talking about AI, I've been talking more and more about this forward deployed marketer.

And I don't know if I explained it well here, but I've been posting. And these posts didn't get a lot of reach on LinkedIn, but I've had a handful of people from large companies reach out saying, and these people are like, these are companies doing billions a year in revenue, right? And some of them were leads for single grain. And they're like, hey, you know, after 20 years of working at so and so multi-billion dollar company that everyone knows, I like what you're doing right now.

I want to come work with someone at the, you know, talking about the front end, right? Like at the frontier. And then I was talking to someone else. He's working 15 years at this company doing billions a year.

And his comp is like, let's just call it like a million a year or something like that. And then he's like, yeah, I know I'm probably gonna have to take a cut if I work with you. I was like, yeah, you are. And then and then but he's like, but I want to I want to be in the next wave or whatever.

And then all that to say is these people are so motivated. But then the challenge is I almost have two companies right now. This is where I'm going with this. I have one side where everyone is super motivated, having fun and just cranking all the time, right?

Because they want to crank. And then the other side, it's a lot of, I don't have the time for this. I can't do it. Oh, it's I'm too overwhelmed. And but the other the side that's super cracked out, they see the other side.

Wait, is it two different companies or two different cultures, two different cultures, two different cultures. So you have one culture that's super, super AI native, super AI. You have the other one that is like, you know, you start when you start to hear the clients that that work with them, right? You start doing I just I just asked my gong, like, what are they saying?

not proactive enough, right? Not coming up with strategies, not fast enough, right? And that's a cultural thing. And that's why I keep harping on this AI native piece thing because AI native for us means it covers all four of our core values.

Be a relentless learner, be biased to action, be unreasonably resilient, make bold long-term bets. If you're not AI native, you violate all four. And so the challenge right now, and I'm seeing this in these CEO chats too, that these people have thousands of employees, they're seeing the same challenge. And then when we have new people coming now, they're just like, dude, Eric, like I've seen the training stuff.

I've seen the committee stuff. It doesn't work, right? I'm like, so what do you do? It's like, you know, I'm a fan of not trying to keep trying to fit square pegs into round holes and just make it's better off if we find them a better spot to work at.

And that's what I'm seeing. I don't know what you're seeing. Yeah, either people culturally want to embrace it or they don't. It's too hard to get people to embrace it who don't want to.

If you don't have a role for them, switch them out, as cutthroat as that sounds. And do you want to know why your forward-deployed marketer is gaining traction right now? Really? Yeah.

Because so many people are funding money into Ford deployed. Engineers. Billions and billions of dollars. Anthropic is doing it. OpenAI is doing it. Microsoft's doing it.

I think you just mentioned Microsoft today is $2.1 billion. A few of the private equity funds have been jumping on board as well. It's just, you know, everyone's just like, oh, companies are going to struggle to integrate this. Let's spend a lot of money.

These are their new versions of Accenture or McKinsey or Bain, but they actually do more hands-on stuff. And yeah. And then on a side note, did you see Accenture stock? It's been taking for a while.

I've been watching. I was wondering if you looked at it. Yeah. But dude, so one thing, one of our mutual friends, Syed, likes to say that AI is IA, which is intelligence amplifier.

But I also think AI is TR, which means truth revealer. Never, never has it been easier for me to have... like informational density or the informational advantage where I can see everything in the company, right? Yeah.

And so if I try to hide from you, you smell it from a mile away. And like you already probably have a good nose because you've been doing entrepreneurship for a while, right? But now it's like probably 10 times faster. You can just get to the point.

And so I think this is more for anybody. Like one, if you're super AI pilot and you want to be a forward deploy marketer, we're hiring. So is Neil. Neil's hiring too.

But I would also say that if... If you're not down for that and you're working at a company right now and it's uncomfortable for you, I think you should find something else. I think you should find something else that fits you because it's better for you and it's better off for the company as well. Don't make it hard for the company.

Don't make it hard for you. And I think if you run a company right now, it's hard to rip off the Band-Aid, but you're better off putting people in a better place.

We're seeing AI, one of the biggest disruptions it's causing to marketing that people don't talk about too much is valuations of agencies. So some agencies that are very AI centric have extremely high valuations, which I don't believe will last. I don't see a service-based company. People could disagree with me.

There's some AI forward ones that have been trading for over 30 times profit. I think that is very excessive for an AI forward marketing type of company.

And the very traditional ones who haven't adapted, their valuations are getting cut. And a lot of people in AI right now, they see the opportunity as, hey, how do I take this technology and implement an organization? Which definitely is an opportunity and is a massive one that people should focus on. But another opportunity that a lot of marketers and agencies aren't focusing on is, oh, there's a lot of marketing agencies that generate a lot of revenue, have lower margins, and

Their growth has slowed down because of AI and they haven't adapted. I can pick these companies up for maybe two times profit on the front end and do an earn out. And they may not, you know, make a large chunk of the earn out. So net net, you may have paid maybe only 3x profit for a very stable company that has low churn and that you can fix.

And maybe your net cost after you fix it is one and a half times, you know, EBITDA. which is just dirt cheap. You can arbitrage that with bank money all day long. I think it's the euphoria right now that's with this.

There's a lot of euphoria around four deployed engineers right now, for example, and thank God to Palantir for putting that out there. I didn't know they're the ones who came up with that. But they have a different version. I think it's become like this new, like people just morphed it into like growth hacker.

They've morphed it into their own definition, right? So that's why when I say forward deployed marketer, a forward deployed marketer is someone who's a really smart marketer. That's a good strategist across the board, T-shaped, deep in one area, go across, but they can speak strategy. Like I tend to think that you and I, when it comes to marketing, we can kind of talk around almost any subject around marketing, right?

Yeah. But I think when you're working with other engineers, you're paired with an engineer, you can build solutions that compound. The key thing with a forward deploy marketer is you build it for these clients and you have your own book of business. But then you bring those learnings back to the product team and you say, hey, I'm seeing this over and over.

We need to productize this. And then you just keep scaling the scaffolding. And then like, by the way, when we talked to because these single brain deployments right now, when I talk to the clients, because it's my CTO and I were on these calls. right?

The difference is it's constantly like, oh, this is amazing. Oh, this is so good. How do you do more? This is great. Can we do this? Can we just, oh, hey, can you add this person into this lab?

Can you do this over? It just, it grows, right? Same thing with ClickFlow too, right? It's when we talk to customers there, oh, this part's broken over here, but this is really good right now.

I love it, man. You move over to client services side, not fast enough, not proactive, not this. And I'm not saying that's like that across the board, but you don't hear those complaints in this new area.

And my point of saying 40 point marketer too is you need someone who's humble, hungry, smart that can scale with this stuff. It's not for everyone. Yes. Yeah.

And generally you're common on hungry. If you don't, find people who are hungry in all positions you they tend not to work out unless you want mediocre results you know you know what i'll say so think about this in your org right i think i would venture to say 99 of the people that we hire both you and me i'm speaking for you you can push back on this are nice people right but Yeah.

Okay. I would say most of them are humble. We tend to find hungry people, but sometimes like what I'm seeing with AI stuff right now, because it's an intelligence amplifier, the smart piece is a little more difficult. Recently, I'm always talking about the last 12 months or so.

I think it depends what kind of roles. Yeah, it does. I'm only talking about marketing roles and digital marketing, for example. No, no, I know. But what kind of roles inside digital marketing?

I'm talking about, let's say you have people that are client-facing and these are people doing the work. So they're the strategists that are also kind of account managing. Yeah. Usually for us, people who are client-facing aren't doing too much of the work.

We've never really had account managers being the ones who do the work. We found that it doesn't work well with enterprise. That strategy does work well with SMB and mid-market though. So we...

My whole take, like these new people that we're bringing in, like some of them are used to that model. And I'm like, I'm here to just, you know, we're going to try to flip that model around. Because my whole thing is, if you have a book of business, you don't need a ton of clients. If we're talking about scaling these exponentially, five to 10 per person, right?

And that's all they need to handle, then they should theoretically... When you did ACS consulting back in the day, and when I was consulting back in the day, I was happy to work with a customer. Typically, they want to talk to you directly. But I do think if you're in a different environment where you have a large team helping the enterprise, you do need that layer, that account manager.

But maybe there's a smaller version of that in the future. I think that's going to evolve as well. I don't know how much, but I do think it's going to evolve.

I think it all comes down to client size. But for big corporations, I don't know how much it's going to evolve. And I would say for big corporations, it's probably one of the things that's going to evolve the least. The reason being is big companies have too many complexities.

So I'll use a real life example. We have a client, one of the largest financial companies globally. Everyone's heard of them. And I think their logo is on our website.

So someone can try to figure it out. They operate in so many different countries with their solutions that we have to have an account manager that's just dealing with them directly. It's too hard for an account manager to go and do the work as well. Because let's say if you operate in 15 or 20 countries for this client,

you have to go and get all the data from all of these different regions or countries, aggregate them, make sure the strategy is going well, everyone's aligned and continually communicate, not just with the people in region doing the work, but also with the customer. The customer doesn't want to communicate with 20 plus different contacts. And in that scenario... Account manager is extremely busy.

So I agree with you. That's what I'm saying. It's going to evolve, but I don't think it needs to be as big, right? And we haven't talked about how big it is, but I think when you have a bot that sits in, the client can ask questions across all the divisions, right?

I think they're going to be able to get what they need faster. And I don't think there's going to need to be as many people on that team necessarily. So I think, again, I don't disagree with you. I think you need someone to manage the complexity.

I think they're going to be able to manage more with less is my point. And we also see in marketing, working with companies varies drastically region by region. So I was with my managing director of LATAM. His name's Javi.

We were in Medellin, Colombia, I don't know, like a week or so ago. And when I was chilling with Javi, we were going back to airport because he lives in Miami. So we were both flying back to Miami. And then from Miami, I went to LA.

And he was just telling me some stories of his past life. And how Latin works, even though we've been in Latin for a long time. I don't understand the culture anywhere near, you know, what he understands. He's a native Spanish speaking person.

He was born in Spain and lived in Mexico for a very long time. And for most his career, he dealt specifically with Latin America. So he's like, man, I remember doing some work for Samsung and they were so happy with me that I gave my work on time. And I was like, wait, what?

And he's just like, yeah, he's just like, I submitted the work on time for marketing and Samsung emailed us and told us they're so happy with us. I'm like, what do you mean?

Do you like the work quality? He's like, no, no, you submitted it on time. People don't submit work on time. And I was like, what? And he's like, yes, this is a lot of Latin America.

And he's like, you don't understand how Latin America works. I'm like, explain it to me. So he's like, let's say, and Eric and I both don't believe in this philosophy of doing shit work. But the example that Javi gave, and I'm not saying we do shit work, is he was just giving me a generic example that if you work with a company in Latin America and your work is mediocre subpar,

And you take them out for drinks and events, you're much more likely to retain the contract than if you did really good work and you did not take them out for drinks and whining and dining up. And I'm like, seriously? He's like, yes. He's like, that is the culture of Latin America.

It's about relationships. It's how well you treat them. And he's like, a lot of times you'll go out for lunch meetings and you'll finish after dinner. The same meeting at the same place.

People keep drinking, eating. They like being wined and died. That culture, from my understanding, you know, I wasn't born in America, but I've lived here almost my whole life. I've never seen that culture in America.

It's a very transactional economy. We're like, what's the results? What's the ROI? Mm hmm. very different in certain countries from marketing dude if if you and i were running a company in latin america startup and our team was paying marketing professionals who are whining and dining us but producing mediocre results and wanted to keep those i would want to fire my employees and fire the people doing the mediocre work i'll just be pissed yeah

You know, as you say that, that reminds me of Japanese culture. So Japanese culture is, you better do good work. You better also Hawaiian and Dinos as well, because it's a heavy, they drink a lot. But it's like, that's why you see the people like, actually, no one knows this.

No one lived in Japan. So you hear about kind of the salary men, right? They work really hard. They don't get paid much. And then they're required to also drink out, like go out and drink and party, wine and dine with their customers.

So Noah's nodding his head. So you get double a gym hat. So be grateful for your Japanese employees. Yeah. And I know from Japan, they have a saying in Japanese.

I don't know what it is, but I know the US version is death by work. yeah like they'll work themselves to death but i didn't know that you have to drink like crazy we have a mutual friend andy lu you've met him a few times right yeah andy lives in taiwan does a lot of business also with in china and just chinese people in general he was just like if you ever want money from chinese people you have to drink and you have to drink heavy yeah but he's like but you can have someone who drinks on your behalf i was like what He's like, yeah, when you're doing business, if you don't really want to drink, you can have someone who drinks for you to build that relation.

So I'm like, wait, I'm networking with someone. I'm trying to build a relationship with them. They want to go out to drink. If I don't want to drink so much, I can have someone with me or pay someone, you know, to drink for me.

He's like, yeah, usually it would be someone in your organization. But yeah, that's what they expect. And I was like, anyway, that's it for today, guys. And we'll see you tomorrow.