← Startup Stories - Mixergy
Startup Stories - Mixergy
#2284 Why is Morning Brew’s founder selling “AI Transformation”?
Alex Lieberman & Arman Hezarkhani of Tenex
Loading summary
Transcript190 lines
- [00:00]
A
Hey there, freedom fighters. My name is Andrew Warner and this is a new series for me. It's called the Next New Thing. Here's what's up in this interview, then an intro, then we'll get right to it. What is AI transformation?
- [00:13]
B
We don't think most companies are going to cross the chasm from pre AI to post AI and we believe we can be the household name to do that.
- [00:19]
A
Give me an example of a project that you took on and what you were able to do that wasn't possible before Step exports.
- [00:26]
C
They were looking for a dev shop to build them a mobile app. Traditionally you would need a full design team, product managers, engineers, and a ton of money using AI tools. When we hear, oh, I have a request for a feature, we go over there, we prototype it and we hand it over to the client. We say, is this how you want it to behave? Now we can go and we can build production grade code incredibly quickly.
- [00:46]
A
At some point you've run out of ways to save people money. This doesn't seem like a sustainable business.
- [00:51]
C
Does a biz ops person ever run out of work to do in a company or does McKinsey ever run out of work to do?
- [00:56]
A
I'm a little skeptical. Alex Lieberman and arman hazarkhani run 10x, which does outsource dev work and helps companies use AI the next new thing. So I got the two co founders of 10x here in front of me and the thing that I want to understand is what is AI transformation? Because I see so many agencies building up this AI transformation practice and you both have been doing phenomenally with it. In fact, you actually are doing it as part of your business.
- [01:28]
C
So most of the revenue comes from engineering. But our fastest growing part of the business is AI transformation. It's very easy to implement AI as a person in your life. You can basically tell somebody, hey, Andrew, instead of using Google, start using this thing called chatgpt. You use it once and then you're like, holy crap, this is really great. And then you just very simply shift your behavior. It's literally one step. But to do that on a business, we call that like multiplayer AI. That's really, really tough. That takes studying, that takes change management, that takes tool building, that takes training, it takes the change of people, process and technology and McKinsey words. And that's where we come in.
- [02:10]
B
And one thing that you may find interesting is why we're doing like engineering as a service, as like a big part of our business versus just AI transformation. Arman and I Met five years ago. I had invested in his first business and we were catching up eight months ago at this point. And he basically had shared with me that he had to pivot his previous company and when he pivoted it, he had to lay off most of his engineering team and he just had himself. And he had one product designer who studied COMSCI in college but hadn't had hands on keyboard in a long time. And yet their output of production ready software 5x. And when I asked why or how, he basically said how he didn't have a choice but to basically rescaffold the entire product and engineering process from how PRDs are written to how code is structured, everything was redone. And when I heard him say this, that for me, even though I've experienced like two or three acts, whatever the improvement is of ChatGPT, that was truly orders of magnitude more leverage that I had seen. And one of the things I've realized working around engineers now is engineers are living in the future like they are seeing the greatest leverage of AI relative to all areas of knowledge work. And so as we were building 10X, you know, we had that in mind, which is like you have this incredible leverage that engineers are experiencing. At the same time you have AI consultancies or like you know, the MBBS of the world that are doing these gen AI projects. And there's all these stats either about gen AI projects just like not working right? There's like the MIT study that came out, around 95% of them don't work. But also a lot of this work is project based. It's one time revenue. And for us, what was exciting was how do we actually control the AI internally where we have engineers who are leveraging it to be far more productive than engineers at other companies. How do we use that to sign 12 month contracts? That is highly retentive, highly recurring. And we use that as the trojan horse to do transformation work after having that line of business.
- [04:17]
C
What's a typical scope from the engineering perspective? It's basically like one of the core beliefs is that the demand for software engineering is infinite. It's completely as price will go down, the demand will increase. And so the scope is pretty vast right now. Like we're building one of our clients, we built them a mobile app that hit 20th in the app store. For another one of our clients, we built an AI agent that helps the front desk at offices around the United States and at healthcare offices like doctor's offices around the us but is it.
- [04:52]
D
Fair to say right now I guess it's like a, there's a lot of companies out there that will do engineering. Hey, you need an app built, you need a website, you know, whatever, we'll do that for you. But you guys have now are now using probably Cursor plus your own agents that handle a lot of, you know, pieces of building those building blocks together. And I know Alex loves Legos, so you've got your own Legos that are proprietary to you guys, Legos that allow you to then produce bigger, faster, smarter, that same scope.
- [05:19]
B
The engineering as a service side of the business. Just think of it as like a high quality AI powered dev shop. And I actually love the idea of that line of business because dev shops have such an unsexy perception. We don't charge hourly. Like we have a fundamental belief that service businesses are going to move away from hourly work. And we think it's starting with engineering. So our clients pay us based on output and our engineers are paid on output as well. And the reason that is so valuable is we can hire truly the best engineers I've ever worked with. Because just as an example, like our best engineer will make a million bucks in cash next year.
- [05:56]
C
Yep.
- [05:57]
B
And, and, and it's not like stock in a startup, it's cash.
- [06:01]
D
Cash money.
- [06:01]
B
Yeah, exactly.
- [06:02]
D
And so it's like, Alex, you and I have talked a lot about the Red Ventures business. They're all about alignment. And Andrew came with me to see Rick the other day about a year ago and one of his favorite quotes is, show me how someone gets paid and I'll show you how they behave. And this feels like a really cool way to align the value based pricing orientation across, across the board.
- [06:20]
B
Exactly. And what that allows us to do, that, as you were alluding to like the big vision, is it not only allows us to do great work on the service side of the business, whether that's engineering or transformation work, but as we think about building IP internally, we truly have the best technical talent in the world to do it because like we are now as competitive in terms of pay and environment as a big tech company.
- [06:45]
D
Got it. Yeah, that makes sense. All right, what's the big vision?
- [06:48]
C
I was looking, or we were looking at the big hyperscalers and seeing that they're spending billions and billions and billions of dollars building data centers to build something that is being commoditized, where most of the value is accruing to the application layer. And so if you look at OpenAI, if you look at Andropic, so many of their releases now that are getting eyeballs are apps that they released and application layer tools that they release for the first time. Like, we were all really excited about GPT 3.5 and so when they released 4.40-4.5, we were all. That was like an Apple release. Right?
- [07:26]
D
Right.
- [07:26]
C
When five was released, no one like this was like an app.
- [07:30]
B
Crickets.
- [07:31]
C
Yeah, absolute crickets. But what that does is it opens up an opportunity in the application layer where for zero cost, you can compete with these hyperscalers. All you need is brilliant people in a room. And so in. When I was in like 10th grade, I read this book by Walter Isaacson called the Innovators. Basically the story of like the people who like invented the first computer and the first programming language all the way to like, modern Internet, AI mobile devices. Basically it, it becomes this, this study of how innovation happens. And it's a very simple equation. You put very smart people in a room, you give them freedom, but you also put freedom in the form of like financial freedom or somehow you give them fuel, but then you also give them a direction and you give them some form of commercial pressure. Right. In the form of whether that is investment, you have to get a return or I'm Bell Telephone company and Bell Labs is a research institution, but they have to output something for the company. Similarly, we think that if we can build a team of geniuses and we can give them freedom and direction, then we can compete with these hyperscalers. The question is, how do you do that? How do you build a team of geniuses? And this is how we're doing it, starting with the world's best dev shop, then the world's best consultancy. And as we're doing this, we're building ip. And by the way, as we do that, we have a distribution advantage not only from Alex, but also from the customers that we built trust with and we're actually already working with. And we think that the TAM for this business is basically the GDP of the entire world, every single business in the world.
- [09:05]
D
Sure.
- [09:07]
B
One other way I would just frame it is what would it look like if McKinsey Banner BCG was created today?
- [09:14]
D
Right. So, but is that, is that the end game? Do you want to become the McKinsey of AI or the Accenture even?
- [09:19]
B
Our view is there's this massive technology innovation that's happening. We don't think most companies are going to be resourced to cross the chasm from pre AI to post AI. And we believe we can be the household name to do that either through services or through product.
- [09:36]
A
Here's what I want to understand. Let's stick with the dev part for a bit and then we'll come back to the AI transformation. Give me an example of a project that you took on and what you were able to do and what your developers were able to do that wasn't possible before. Okay.
- [09:49]
C
Yeah. So as an example, I use Snap Exports. Okay. They're a social media company and the reason why I use them is because they were traditionally going with a. They were looking for a dev shop to build them a mobile app that is basically like. Their vision for it is like HQ trivia for sports fans. Traditionally, for a product like this, you would need a full design team, a full team of product managers, a team of engineers, and you would charge a ton of money and it would take. You'd basically scope out the entire project and then six months later you'd come back with a product and then it would either work or wouldn't. The way that we work is when we hear, oh, I have a request for a feature, we go over there and we prototype it and we hand it over to the client. We say, is this how you want it to behave? And then we have a brand book from them. And so we know what the UI generally should look like, we know what a button looks like, we know what an input looks like. And now that we know what the behavior looks like, because we were able to iterate on it in a vibe coding platform, now we can go and we can build production grade code incredibly quickly. The act of writing code is no longer the barrier. So no one on our team actually types letters into an ide. If I walk through a wework and I see someone typing letters in IDE, they're working in like 1990. Right. So we're using, we're pushing modern AI tools to the absolute limit so that we are scoping work down until the most atomic level, and then we're getting it done and we're architecting it such that we're really treating these AI.
- [11:21]
A
What does it mean to scope? To scope down to the atomic level?
- [11:24]
C
Yeah. So let's say you want to build slack, right?
- [11:27]
A
Yep.
- [11:27]
C
It's a gigantic platform and typically what you would do is you would manage, you. You'd have a product manager who would break that down into sprints and then they would break those down into tickets, right?
- [11:36]
A
Yep.
- [11:37]
C
And then at, to some extent, the engineer would have quite a bit of freedom around that ticket. The tickets are quite broad. We go, we drill. Our engineers drill all the way down until they're micro tasks. And this is really what, this is the difference between an AI engineer and a traditional software engineer. An AI engineer needs to think like an architect. They have to think as if they're, they're delegating all of their work to other engineers. And when they're delegating their work, how do they do that? Right. And not only are they delegating their work to engineers, they're very special engineers. These AI agents, they're incredibly high iq, but they make really stupid mistakes that as you work with them regularly, you can understand what they are. And when you work with these, you get really good at managing them and understanding them. And so most of our engineers have multiple different terminals open, each with a different coding agent. And they're just, they're trucking on all of them. So they're really acting as managers of these different agents.
- [12:30]
D
Are they, are they using voice or are they using, are they, are they typing?
- [12:33]
C
Frankly, I think that the voice thing's overblown. I think that when you work in an office it's like unrealistic to talk out loud. I think that the voice thing is like useful if you're an executive driving your Tesla to and from work and you want to, and you want to speak to an AI.
- [12:46]
D
So you don't think the whole world's gonna start talking to their computers and voice pilled.
- [12:50]
C
I think they might. But like in our office for example, it's impractical and I don't know how that works in a modern office setting.
- [12:56]
D
I think I understand. When you're scoping this piece of work, you launched a mobile app for someone you can be, I'm making this up. You can be cheaper to them than they would have paid otherwise because of all this stuff you guys have done. And it's probably higher margin business for you guys.
- [13:11]
B
Yeah, the way I basically think about it is it's, it's because you have.
- [13:16]
D
Like a 10 person engineer, 11 person engineering team, but it's really only one human being and 10 pieces of basically.
- [13:22]
B
If, if traditional dev orgs or dev shops are labor arbitrage, we're intelligence arbitraging.
- [13:29]
D
I mean it's interesting the AI transformation aspect of that is that just sizzle really. And then the real thing is actually just go get media engineering projects and, and focus on those. Like for example, today I would argue with you and Alex, this is, Alex is my friend. So we're going to keep this as informal as possible. Like I would tell you guys today, I obviously believe that AI transformation will be a hundreds of billions of dollars industry. But today, yep, engineering development is actually a multi hundred billion. And AI transformation is like a $10 billion industry.
- [13:56]
B
Yeah. I mean, candidly, this is something we talk about all of the time is what should that balance look like? Because the engineering as a service business, one, there are, there's really no one else doing it in the way we're doing it right now. And also it is such good revenue, like, it is just so retentive. It's latched onto the core technology of a company. Whereas to your point, like AI transformation oftentimes is project based. It's harder to make recurring.
- [14:21]
D
Well, not to mention a problem versus solution. Right. I mean, one thing you guys have to be careful of, there's this, there's this meme that was going around that was like, CEOs are like, we want AI. You must have shared it.
- [14:31]
B
When do we want it? We want it now.
- [14:33]
D
What do we want it for? We have no idea. Like, and that's, there's actually a lot of risk and wasted time and energy in terms of firm building. If you guys are building a firm off of this fickle revenue where nobody actually has a problem, they're just looking for a solution as opposed to finding problems, that's a much, much more interesting place to be playing in and solving the problem better and cheaper and faster than anybody else can solve it. Because here's. Here's another. I'll throw a crazy idea as you, Alex, with your platform. When Red Ventures bought Bankrate, it was a publicly traded company. They paid $1.3 billion for it. They paid 10 times EBITDA, roughly according to their math. Rick knew he had done the math. He had seen all the levers. He knew that within a year of closing it, I mean, his projections were that his effective EBITDA would have been five times ebitda because he knew exactly how he was going to double the EBITDA of the business. And he came in three months ahead of time. He got it done in nine months, right?
- [15:27]
C
Yep.
- [15:27]
D
You guys, if, if what you're saying is real and you get really good at it, you could go buy these dev shops and, and essentially you could increase. They already have books of business, by the way, which is, which is. Makes your life so much easier than Alex Twee all the time. And you go buy a $10 million revenue business with all the clients and everything already, and you guys can effectively rip out 70% of their cost, it sounds like, and deliver a better product. And now that's like the math there is actually more compelling because you could probably buy them at six times EBITDA effectively, but then you're actually buying them at three. And like anytime you have a real capability, that thesis is like very, very powerful.
- [16:01]
B
Yeah, yeah, you're basically just, you're buying distribution and you know, you can basically arb the product, like the customer experience.
- [16:08]
D
Not, not to mention, dude, you can finance the whole thing with debt. Like if you buy something at six times, you can borrow half that cost. But if you're effectively. Because you know you can double their ebitda, you effectively have no equity out. I mean, that's how like there's a company in St. Louis I found the other day that I got all excited about. You've never heard of it. It's called Perficient. No, it just sold for $3 billion. I mean it's been around 20 years, but it's literally like a roll up of one of these like Salesforce, Microsoft implementation companies. Implementation companies. And it just, every year it would like pick a different part of the, you know, the. I think I call it the function vertical box. Like HR for pharma, marketing for. And it just did a bunch of deals and like exited for $3 billion. So.
- [16:51]
C
Jesse, let me ask you this. I. We love all our children equally, whether you're an engineering client or a transformation client. You're a 10x client, you're part of the club. And let's say we can drive some amount of efficiency for our clients, right? You're talking about driving efficiency for a dev shop. But what if we can prove that we're driving efficiency for our clients? Right? How do you think about deploying that company to AI transformation or like general efficiency, like a traditional PE clay or focusing on dev shop and going that direction?
- [17:20]
D
Yeah, I mean I would, I would chase the problems, I guess. Like it kind of like, I mean, if you were to solve the same problem, let's, you know. And this is what our CTO tells me all the time. Adam, he's like, look, you just think about the. How do you. How does anyone decide how to spend money on dev? They do impact, cost and ease, Right? The same analysis. How big could this be? Money wise, saving us, how confident are we it'll work and how costly or expensive is it? And we, you know, at Ambush, we used to do that all the time. And like it was like, damn, should we build our own API pulls from Data's Facebook layer or should we use a third party or you know, should we not do it at all. And then everyone does the stack ranking and they go, well, we only have $2 million of engineering resources or $3 million engineering resources, so we should basically say no to everything below this hurdle. But now you guys know this, like, AI allows us to go faster and way cheaper, but it's like if everything gets cheaper and faster, then the amount of things people are going to do is going to go up dramatically.
- [18:18]
B
Yep.
- [18:18]
D
So I guess my answer to that is like, what are the real problems they have that like those, those and. And maybe some of your projects still are 90% non AI, but you're using all your AI tools to do it. Because as you guys, I mean, watch anyone coding with Cursor or any of these other tools and it's fucking insane. Like, yeah, my point is like the actual I ness of it is not actually that important if I'm you guys, based on what Alex said about the. Because, by the way, I'm a services business guy. Like annual contracts, aligned compensation models, margins improving. Like services businesses are hard to build for a reason because most people don't have some of those things. And even like, I mean, I was talking about this publicly. Ambush growth assistant. Very soon, like within months is going to pass Ambush's ebitda. Do you know why?
- [19:05]
B
The cost of labor.
- [19:06]
D
Nope. Retention. Ambush would get big and then two of our biggest clients would churn because, oh, we're going to take this in house. People don't take global talent in house for the most part. Some of our bigger clients have said, hey, we have some. But like that wasn't, oh, you're making a thousand bucks a month off. I mean like we just. The retention in growth system is amazing.
- [19:27]
C
Yep.
- [19:28]
D
Right. And so that's. I would just like tell you guys, McKinsey, by the way, and some of these guys have massive retention.
- [19:34]
B
Are you saying basically just because of the partner's relationship. They're constantly selling in new work.
- [19:39]
D
Constantly. Their prices just. And this is something we see with AUX too, is where we see a lot of like, oh, it's not. It's not like SaaS retention, but it is like they keep coming back to us for more and more projects. Of course feel good that we can actually manage and project that.
- [19:53]
B
So and to your question about is it sizzle? I think irrespective of what direction we choose, like what is the balance of engineering or transformation work? I do think positioning in a AI transformation way gets us more engineering clients than if we did not position in that way.
- [20:12]
D
Yeah, fair enough.
- [20:13]
B
Half of our clients have come through these free AI diagnostics. We do. And so we run these diagnostics. They take two weeks. Most of it's been automated. The only thing that's not automated is basically three hours of free stakeholder interviews. But half of those have turned into our engineering clients.
- [20:29]
D
That's smart.
- [20:29]
B
And so, yeah.
- [20:30]
A
So wait, let's. Let's talk about the transformation part of the business. Give me an example of what you've done for someone. What did they walk in with? How did you understand what they needed and then what did they buy?
- [20:41]
C
The first part, the understanding what they needed is actually the core. And the way we do that is Alex is talking about what we call our diagnostic, and that's the free version. And then we also have an audit, which is the paid version. It's basically anywhere from like 10 to 20. Interviews, data requests, a survey, a lot of deep research that we do on our own. Like, and. But the output of either of these, obviously the diagnostic is lighter than. The audit is a report. The report has our learnings. It's aligned with the goals of the business. Whatever.
- [21:12]
A
Pick a specific customer. You don't have to tell me their name. What did you do? What. What did you. What did you talk to them about? What did you uncover and what. What did you do afterwards after this transformation?
- [21:24]
C
So basically, one of them is a. Is a SaaS company that works in health care. We did an audit for them, and there are two types of interviews that we do in the audit or actually, let me start with the survey. So the survey, basically the way that we think about it is it goes to everybody in the company, or if. If the audit is specifically part of the, of the go to market order.
- [21:43]
A
But again, give me. Give me a specific company. I need to see the picture of what this is. So a billboard company finds you because of Alex's tweets, I'm assuming.
- [21:50]
C
Yeah. Okay, so alex posts on LinkedIn. Actually, quick, quick, quick kind of note here. All of our businesses come through LinkedIn, not Twitter. So, so basically there's a. There's a billboard company and there's. They do traditional billboard selling where, like, you'll. You're a big business and you'll call them up and you say, hey, I want to be on a billboard. And they, they'll put you there. But they also have, like, a platform that's similar to Uber for billboards, where you can go on, you can upload a picture, you can say, I wanted to go here. Click, go. Okay, cool. They came to us Saying, hey, I want to use AI in the company, I want to work, I want AI to be in the engineering platform, I want AI to be in the product itself. We're not sure where, but we know that there's opportunities. Here are a few ideas, but we want you to greenfield, go, go check it out. So we start with a survey. The goal of the survey sends to everyone in the company and the goal is to understand. We basically think of that as building a map for us. Right. It lays out the entirety of the company from an org chart perspective. But also it gives us an idea of where to look. Then based on that, we set up interviews. So so as an example, the survey really highlighted the product org. Okay, so we spoke to the product, or we had a couple of calls with sales, we had a couple of calls with other people just to kind of poke around different areas. But a really focused on product. We spoke to the product manager, we spoke to engineers, we spoke to the cto, all these different people on product. What we found is there were areas of opportunity in their product. So we dug in deeper. We use the product, we checked it out in the flow. Similar to like if you've uploaded a Facebook ad, you start with the Facebook ad, you set a goal, you upload an image, you set a caption, whatever, and then you click go. I want to run the Facebook ad. They have a similar flow for uploading a billboard. So you start with, I want to run it on these billboards in this area. I have this goal in this budget. I want to upload an image and so on. We asked about drop off data, so where are people dropping off? And then we notice that after this, we ask what happens after they click go? And there's two levels of moderation that are done by humans. Okay, so even just in this flow, we notice that there's drop off at either of those moderation levels. One of them is for like internal moderation. And then they send it to the billboard owner to moderate. And there's drop off at each of those levels.
- [24:02]
A
And so you're saying, look, we're losing users, it's human beings, easy for us to use AI to replace. Humans give a response quickly. Got it.
- [24:10]
C
Exactly.
- [24:11]
A
Okay.
- [24:12]
C
Likewise, in the flow of uploading an image and kind of building that campaign, we noticed that people don't know what a good ad is. And so there's a drop off of the creative step. And so we noticed that with the same algorithm that is helping do the moderation, we can also help give feedback on the creative that's uploaded to the platform and even built in the platform using AI. So basically we made these recommendations. We said, hey, we're going to build you an algorithm that's going to be used in the creative step to give feedback. We're also going to use this to replace moderation. We're also going to help you build AI powered like an AI powered billboard generator and so on. We built a. We made a few of these recommendations and for each one there's basically a lift versus reward. So how much is this going to cost and how much are you going to get out of this? And that, what are you going to get out of this is tied directly to the business's goals, whether it be revenue, ebitda.
- [25:09]
A
This doesn't seem like a sustainable business. At some point you've run out of ways to save people money. So does that mean that all these AI transformation companies eventually have to turn into engineering businesses where they could get ongoing work? Or is there something that I'm missing in ongoing work in AI transformation?
- [25:25]
C
Does a biz ops person ever run out of work to do in a company or does McKinsey ever run out of work to do with a business? I think that, like there are examples.
- [25:34]
A
An efficiency expert, does they run out of work?
- [25:39]
D
I think there's enough. I mean, Andrew, I think there's a couple of issues. You know, one is AI is changing at a rapid, rapid pace, right? And when I built Ambush, every six months, Facebook would change the way it worked. And it was just to our benefit because every time I changes, some new model is going to launch. And then the same people who you did work for a year ago, you're like, oh shit, we need to do more work here, right? So that's one thing. But then the other thing is I think the point of consultancies exist and they ebb and flow. And not to say there's not cyclical market issues they have to deal with, but there's always work for companies where they have problems and they have issues and you know, it's pretty perpetual.
- [26:17]
A
I'm a little skeptical. I'll be open with you because partially what I heard you all say, both Arman and Alex say earlier, is the reason the dev business is so exciting is there is ongoing work. You get embedded, you keep building with them. So maybe there is some place where AI transformation just kind of, if not dies, it goes down dramatically and you have to keep getting new business.
- [26:36]
B
One question I would have for you is, so I think what you're saying makes sense. Do you have an overall skepticism of management consulting firms?
- [26:44]
A
Interesting. I like the way you put that. I see you're saying just like in management consulting firms, there's always more management consulting, more ways to improve. Same thing here.
- [26:53]
D
Yeah. I'm asking ChatGPT right now. How big is the third party engineering business?
- [26:59]
A
Well, the engineering part I get. It's the AI transformation company. Do you? At some point you should.
- [27:04]
D
But those are gonna just blend, right, Andrew?
- [27:07]
A
Yeah, that's what I. That's what it seems like. Makes sense that at some point it does need to be ongoing engineering work. Otherwise you can't keep finding these places to cut human beings out. Eventually you've cut out all the human beings from the bottom.
- [27:18]
D
Let's stop using the word AI because the word AI is stupid. The real word is I can. I can build technology for you cheaper and faster than I used to be able to.
- [27:28]
C
And not just build it, but also.
- [27:29]
D
But, but like.
- [27:30]
C
Yeah, exactly.
- [27:31]
D
No, but that. Like, that's what you're. Actually. The whole reason this is an opportunity is because there's tools and like our ability, Andrew, to like, figure out content on our. On my posts and whatever or like any of these things. We just would not have said yes to them before. If we had gone to Adam three years ago and said, hey, Jesse's going to talk on. On, like podcast. You need to figure out how to take what he's saying and turn that into written content. He would have been like, what? I need an NLP this. And I need a database thing. This. And I. And it's like, dude, it would have been. He would have. Sure, I can do it for $3 million. Six months to do that, and now it's something they can do in a week because of AI.
- [28:04]
A
I see. So the ongoing opportunities to do work will keep coming up. So even if at some point you get rid of the efficiency, there are new things to be built. Alex, I'm watching you. You're. It's either my connection is really bad that's making you anxious, or there's something you want to say and I'm holding you back from saying it. What is it?
- [28:19]
B
Because I think your. Your gut reaction is a very natural reaction. Like, I actually, I think I've thought about this. Like, my, my kind of like, it's.
- [28:29]
D
Like the old Indian uncle reaction. I was going to say we buy ads. Ads on Facebook for people, the old Indian uncles. Like, why can't they do it? Tons of sales.
- [28:36]
A
What.
- [28:36]
D
What do they need you for?
- [28:38]
B
Yeah, I was going to Say like my, my, my, my Jewish OCD is acting up right now. And what I'm thinking to myself is like, yeah, I've had this thought before. And the way that I kind of think about it is similar.
- [28:51]
D
Like Alex's cold sweat dreams. You're saying them as a fellow member of the tribe and he's like, oh yeah, why do they need us?
- [28:57]
B
Andrew, the one question he wasn't supposed to ask. No, but I think there's a few things. One is I ultimately the way that I think about it is like consulting firms really just like are an index or like they latch onto being a growth and strategic partner for a business. And, and kind of the anchor for that is the partner who has the relationship with the CEO. And so the CEO, whenever the CEO has cold sweats at night, the CEO is not texting his best friend, he's texting his partner at Bain or BCG or McKinsey. The way I think about it is AI is this incredible wedge into being a company's growth partner and strategic partner in 2025. And so even if you're right, let's just say are my deepest fear. And your deepest fear is right, which is you run out of things to do. What I still think is there every business need support. Every business some size needs support to grow and needs help to unstuck problems. And there are a lot of problems to unstuck even if the solution isn't always AI.
- [30:12]
D
Just think. Here's a simple way to think. Well, first of all, let's actually play a game. I just googled or chatgpt this. How big do you think the I just said how big is third party engineering as an industry? Yep. And everyone think about your answers prices. Right rule applies. Andrew, you go first. Mr. Skeptical.
- [30:32]
A
I would say quarter trillion dollars.
- [30:34]
D
Okay. So that's 250 billion.
- [30:37]
A
Yeah.
- [30:38]
D
A year.
- [30:40]
B
49 billion with all of third party engineering. So like Deadshots Engineering.
- [30:47]
A
Yeah.
- [30:48]
C
Oh, over a trillion for sure.
- [30:52]
D
The right guy was right. It's. This has a reasonable ballpark is 1 to 3 trillion annually. So dude, it's, it's another way to think about it. Like you think about these cloud services guys, Andrew. And like they're like you I think you're getting so close to the problem you're forgetting the business need which is like why did someone need a Salesforce implementation shop? Well, the real reason is because they needed a CRM. Well, why they need a CRM? Because they need to track their customers and manage their customers and follow up with their customers. Then they needed a software to do it. Then they needed a service provider on top of the software to implement it. And so the point is these things all have to start with business problems. And I think what Alex was saying before about McKinsey is, do you think businesses will stop having problems? If your answer is no, then they'll continue to have a business in helping them solve their problems, which happens to be using that. The modality today is AI. 20 years ago it was, it was cloud services.
- [31:43]
C
Right.
- [31:43]
D
20 years before that it was like, we'll put up your mainframe and your thing and we'll do whatever you need there. And it's gotten bigger as there's been more and more stuff and everyone's using more of it.
- [31:52]
A
So are we all agreeing that AI transformation is the beginning of the relationship, but it's not the long term business? And then at some point something needs to come up. You either become the person who, when the CEO is sweating in the middle of the night, they call you, as Alex said, or as Jesse said, you keep finding other things to do and either way you're the support system and you happen to be using software or whatever that software is. Yeah.
- [32:13]
D
Or Alex, we've been spending a lot of time looking at the like as new tools and technologies pop up, there's going to need the equivalent of implementation businesses. And so someone's going to, some, some tool that we can't even think of today is going to launch on top of anthropic or whatever and everyone's going to go, wow, that solves this one problem for me that, you know, engages my employees. Now I need to get like, I need someone to implement that for me and so on and so forth. Sorry Alex, what were you going to say?
- [32:38]
B
I think, Andrew, to your question. I think AI is a tool in the toolkit that solves business problems. And ultimately if you want to build a huge business, it, my view is it isn't the only tool you're going to rely on.
- [32:53]
A
Armand told me early, before, Arman told me, before we got started. There's a lot of FOMO. You hear that AI can transform. You call 10X, you say, I heard it can transform. Go and look around my company and tell me what's going on and people will get a will, will then find ways to improve the business. That's, that's why this whole thing is so sexy in the beginning. All right, I want to do a few demos here. Can you both do a screen share and show me a few things here?
- [33:17]
B
Yeah, for sure. So first, what do you guys use to do you guys use eos?
- [33:23]
D
We have. Yeah, we use elements of it.
- [33:25]
B
Okay, cool. So something that I always found like I so eos just as just to give quick background. It's like okrs, it's a business operating system. My whole thing is I don't like thinking about the rhythm of running my business and the processes and I don't want to invent them from scratch. EOS is a system for running your business. We at Morning Brew we self implemented meaning we read a book called Traction by Gino Wickman and then we implemented what was talked about in the book. A lot of people hire actual implementers who either run your off sites, help you with EOS implementation and they charge a shit ton of money. Like you know it could be like $15,000 a day for being at your off site or with implementation. So what I wanted to do was basically build an EOS implementer because I'm implementing EOS at 10x right now. And so I'll just take you guys through this quickly. Okay. So EOS facilitator basically built a custom GPT. And the way that I think about custom GPTs is it's. It's basically just a, almost like a persistent ChatGPT instance. So it's the, you know, instead of me having a one off conversation with ChatGPT and then it falls into the abyss of all my conversations. If there's a conversation that I'm going to want to revisit frequently, if it's something, if it's a tool that I'm going to want to share with other people in my business and if I'm going to want it to operate with context that like on an ongoing basis, then I think about building a custom GPT. So we look under the hood here in this EOS GPT. Basically the way it's configured is I have uploaded the actual book Traction by Geno Wickman. I found the PDF of the book uploaded that I uploaded the VTO like all of like the frameworks you fill out to to implement Traction. What I actually haven't uploaded here, but it's something I'm going to do is take our knowledge base for 10x so our org chart information about our clients and upload it here. So then EOS or this implementer has full knowledge about our business and then the instructions are basically the way in which this custom GPT is supposed to interface with me. And just for context, the way that I always approach creating these instructions is I just have ChatGPT5Pro create a prompt. So I asked it to create a prompt that would act as kind of the instructions for my US implementer. And so just to give you an example of what this looks like in.
- [36:07]
A
Practice.
- [36:12]
B
And, and also like in terms of the model and this is something Armon and I talk about a bunch of. I just use GPT5 Pro on this. It's like the most powerful model that GPT has. It takes longer, but it's just very good at reasoning. So one thing that I was doing today with it is I wanted to create the accountability chart for 10x. And so what I literally did is I just had a conversation with my implementer. So I said, can you help me create my accountability chart? Said yes. And then it took us through the directions for creating the accountability chart. The accountability chart is like, what are the key functions in your business? It's about roles, not seats or people. And then it asks questions. And so what I like about this is it acts as like my facilitator because it's just asking questions to gather information to complete the accountability chart. So then I said, We're 10x, we have 12 team members. We have both visionary and integrator. Our core functions are sales and marketing products and client success and ops. It fills those out. Then it says, do you have five key responsibilities per seat? Which is what the accountability chart has. I had already created it, so I shared an image and so then it just filled in what the roles per every seat were. And then it asked a few follow up questions and went through the full thing. We have a full accountability chart now. And then the next step is you look at everyone in your company and you look at do they get it, want it and have the capacity to do it? Like, are they right for the roles you have in your company? And so then it created this whole table that now our leadership team at 10x is going to fill out to know if like everyone is right for their role at the company. So that's just one example of how I basically turned Traction, the book and like all of the guidelines into my facilitator. So this doesn't have to be like a single player experience. And one thing I'll say is I've found this very helpful because I am very much an external processor and so I find it way easier to have someone questioning me than me trying to, in my own head, self implement eos.
- [38:10]
A
I'm like that too. I need that conversation and voice ideally. Yep, okay. And I'm looking over on the left side of your screen. And I can see that you also have an AI coach which I think is trained on another book and this for different aspects of your business.
- [38:23]
B
Yes, this one was. This one was trained on Matt Boccari's great CEO within.
- [38:28]
C
That's cool.
- [38:29]
B
And then also info about 10x. So I, I actually use this all the time. Like we were just talking about. Armand and I are talking about like, did we want to give a bonus to a certain employee? What should that bonus be and what are the trade offs? I went through the conversation here and what it suggested is what we ended up going with and it's scripted what our conversation should be when we meet with that employee to share their bonus. But set the right expectations.
- [38:53]
A
That's fantastic.
- [38:55]
B
Yeah.
- [38:55]
A
Okay, Amran, you got a. You've got something you want to show too, right?
- [38:59]
C
Alex mentioned developers are living in the future and the tool du jour, like Jesse earlier you mentioned cursor. Our engineers would call you a boomer for using cursor. Like that is like so yesterday. And so the tools that are like the big ones right now are Codec CLI and Claude Code cli. So to those who've never seen this interface or think this is scary, this is basically what's called a command line interface. So if you're on Mac and you open what's called terminal, like you just go terminal, it'll open. And then on Windows, I think it's something similar term, if you search term, it'll open. So if you open up terminal and you download Claude then or Claude code, which you could just do from the website. Then when you type in Claude, it basically opens up this program called Claude code. And so you could just talk to it. You could say, I am Armand. Or actually here, let me. Who is Jesse? Did I spell that right? Okay, search the web and give me a brief bio. So it basically goes and does some research and it says, okay, I want to do web search. Jesse Fuji, bio entrepreneur. Yes. So this is like a very simple use case, but we'll get nuts in a minute. So cool. Serial entrepreneur, blah, blah. Okay, so let's actually say I want to build a competitor to Ambush. Oh shit. Please write me a business plan in a markdown document to do. So start by doing research. Spin up at least three sub agents running in parallel. One should study Ambush and its competitors, Andrew Buckle in. Another should study go to market business model to third one and how we acquire customers. And the third should study business model, use the Internet to do that. Then Come back and write the report. Cool. So first it's like, let me understand what this guy wants. And then you'll see in a moment it'll spin up three. So this is one subagent Research Ambush and Competitors. This is a second subagent, Research go to Market and Customer Acquisition. And then this is the third sub Agent Research Business Model Options. And now each one is running, each one is doing its own searches, as you can see. And I need to approve each of them. But you know what? I'm just going to say yes and don't ask again for web search commands. Yes and don't ask again. Yes and don't ask again. Yes and don't ask again. Yes and don't ask again. I'm just going to say yes because.
- [42:19]
B
It'S not listening to me because it.
- [42:20]
A
Wants to make sure that you're comfortable with the number of tokens it's using.
- [42:24]
D
Yeah.
- [42:25]
C
So you can change modes. So you can go to, I think it's Control Shift or Command Shift or something. Shift. I don't know what it is, but there's some command where you can change modes and you can go into planning mode where it won't do anything, it'll just write a plan. There is auto accept mode or like we call it YOLO mode, where it just says, screw it. I'm just gonna, I'm not gonna ask you for permission. I'm just gonna do whatever.
- [42:46]
B
And just so you guys have context, like, one thing I learned in just watching Armand and the different engineers work with Claude code is the reason it was such a kind of like incredible like piece of software. And the reason also, like, Was it Sonnet 3.7 that was incredible is because of its ability to have tool use, like to basically access tools, external tools to do work.
- [43:11]
C
Yeah.
- [43:11]
B
And it was. Other models before this were not nearly as good as doing this.
- [43:16]
A
Yeah, Tools like what?
- [43:17]
C
Like task. Like web search. Like it. So there is so much going, going on behind the scenes. There's a lot of people who've like reverse engineered cloud code to understand how it works. There's a lot of really cool techniques like injecting reminders to tell the agent, hey, don't forget this. Whatever. It's really, really involved. But it's really good for. And that makes it really good for coding because basically forces the agent to stay on task, but that makes it useful for other things too. And that's kind of one of my goals is to get people who aren't engineers using this. So as this is going I'm actually going to show you another, another example. Okay, so this is still running over here in this window. But over here I'm going to kill this, clear it and then I'm going to say Claude and I'm going to say open up LinkedIn. LinkedIn's website or webpage? Open up LinkedIn. Why am I speaking? Why am I being weird? Open up LinkedIn. Yes, I want to proceed. Cool.
- [44:24]
A
So now opened up a tab in a new browser window.
- [44:26]
C
Yep, that opens up a tab in a new browser window. Now I'm going to say search for Andrew Warner, founder of Mixergy. Okay. It's taking a snapshot to make sure it did the right job. But as it's doing that founder of Mixergy and send him a dm. Tell him oh wow. Or this was sent by Claude code. So now we're going to let me try to manage these windows. So I'm going to put them side by side here. Okay. So it's asking. Yes, I want to proceed. So it's going to click on search Combox but now that I resize that it needs to reorients itself.
- [45:18]
B
And Armand's using a tool called Playwright which, which basically allows you to interface with web pages.
- [45:24]
C
Yeah, which is actually, it's, it's an mcp. It's super easy to install and it's free so anybody can do this. So now it's going. It, it searched for Andrew.
- [45:37]
B
There he is.
- [45:38]
A
Found me right there. The one search result. Okay. Yep.
- [45:42]
B
That man is so handsome.
- [45:44]
C
Good looking guy.
- [45:45]
D
The machines are alive.
- [45:49]
C
What now see it's saying message button for Andrew Warner.
- [45:53]
A
Because it identified the message button on its own.
- [45:56]
C
Yep, exactly.
- [45:56]
A
And knew that's what it needed for the dm.
- [45:59]
C
And now it's looking at it like it actually looks at the code. So now just opened up the message window. See that? And now it just put this was sent by Claude code. And to round out our magic trick it's going to take a screenshot just so that it can wind up and it's going to send that I'm really trying to buy time for cloud code. And now it's going to click the send button and it's sent. So Andrew if you can.
- [46:28]
A
This is sent by Claude code. It went through from Alex to me.
- [46:31]
D
Where does all this go in in your guys mind, where does all this go? Like Arman, what's the pace at which I mean this stuff is obviously it's really cool. It's a little slow if you're trying to do something that takes a long time. But what's going to happen in like six months or a year, in your opinion?
- [46:48]
C
So what's interesting is that this example, technologically speaking, this was possible over a year ago, right. Why is this not the norm? I would argue that it's for the exact reason that, like, for the exact problem that we're solving. It's hard to implement. Like, this is a cool magic trick, but to understand how you implement this is actually difficult. So one of the things that, that Alex and I talk about a lot is when people argue, hey, Claude, code or cursor or whatever, AI is not as good as I am at writing code. That's what a lot of engineers will say. But a lot of people outside of engineering will say similar things. They'll say, AI is not as good as, as me at X, so I can't use it. To that I say basically, if you have a free or near free junior employee and you can't, and you have unlimited versions of this, you could spin up infinite numbers of these near free junior employees. If you can't find a way to generate leverage from that, that's a skill issue. Right?
- [47:50]
D
Right.
- [47:51]
C
Now, if we take that curve and we take the cost of intelligence, I would argue that this is intelligence, right. If you take the cost of intelligence and you asymptote that towards zero, what you see is people will 10x or they will go towards zero. And if people have the ability to architect, right, Whether it's code and they want to architect it, or in this case, if you are a really good recruiter and the only thing that was holding you back is your ability to message a ton of people or whatever it may be. If the only thing in the way of success was just time, you're going to succeed.
- [48:30]
D
But sorry, my question was a little different. Where do you see the technology going for like, is this going to happen in three, like in one second in the future now? Or like, is it going to. I mean, that seems like to be the obvious thing, that it's going to go faster, better, smarter. Like, how do you guys think about that?
- [48:46]
B
Yeah, I mean, if you read the. I'm forgetting his name now. We could link to it in the show notes, but there was just a piece written by a guy who was a is or was a researcher at one of the Frontier labs. And he's like, everyone basically talks about how like the models are slowing down in their improvement. ChatGPT wasn't that good. And what he basically posits is like, we as humans are just really good at getting used to things that felt like science fiction feeling like non fiction anymore. And the, the goal line just keeps moving in crazy ways. And you actually map out basically there's these independent orgs that do mapping of every new model. Basically how long can the model run for and still have a 50% success rate of whatever the task was that it was doing? And when GPT3 came out, I want to say it was like it was like three seconds or four seconds and now it's like it's at like an hour. And if you actually look at the charting of this, it is still exponential in how long that these tasks can run for. And the reason they look at that is it's like entropy. The longer a task is, the more likely it is that a model will start like spinning its wheels and all these things. And so if you just look at this like it is still exponential. And he would argue that basically you're going to be able to have a true digital employee, meaning it can do 24 hours a day of work without you checking on it by 2027.
- [50:14]
C
Yeah, I would be more bullish than that person. I would say that it's actually much sooner I think in terms of the, like to nerd out on this for a moment. I think if the way like, if you think about this in terms of like using humans as an analogy, the base models are, are appearing to asymptote, right? They're appearing to either asymptote or the growth of the speed of growth is slowing. Most of the growth now is coming at whether it be the application layer or the fine tuning layer or whatever or that type of like adding intelligence on top or adding ability on top of like core iq. And with the stuff that I just showed, like adding what I call arms and legs, right? And so like if you want to use the human being as the parallel, the base model is basically a newborn baby. They're born with a brain. But we need to spend a lifetime teaching that person how to be a human. They need to grow, they need to get coordinated, they need to be able to do shit, right? And that's what the models can't do right now. That is make that is happening much, much faster. I think that is the part that all the money and time and intelligence is going towards. So I think the next year we're going to see a lot of like when GPT 3.5 was first announced, they were advertising they could write a fucking haiku. Right now we're actually getting to the like let's actually do shit, let's get it to actually do work. And I think that that's going to accelerate over the next 12 months.
- [51:47]
D
What's your take on like human in the loop AI versus I mean again there's like this, some arguments like no, the humans are going to be, you don't need them, they're versus human in the loop. Like how do you guys answer that or think about that?
- [51:59]
C
People think about AI as like something that's out of our control. Like, oh, it's inevitable that AI is going to take over because of course it's going to. Right? Like it's the self fulfilling prophecy. I think that what people forget is that it's a statistical model that is in our control to make it better. And not only are we in control of making it better, but we're in control of making it better in the ways that we want it to be better. Right. And so if we as humans want it to be really, really good at something, it's going to get good at that thing by nature of free markets and intelligence and all these things. If we don't want it to be good at something, it's less likely to be good at that thing. And so I'm not necessarily worried that humanity is going to replace ourselves because it's in our control. I think I am concerned about different. Like some people who run these giant companies, I don't think care about the things that I'm describing here. I think they actually just want to see how far AI can be pushed and how much humans can be replaced. But frankly that's part of why I was excited about like starting 10x. That's part of the initial essay that I wrote is to compete against that and to make sure that humans are in the loop, but in a way that we want to be in the loop. I think humans are put here to create.
- [53:14]
D
Can you bring that down one level that the philosophical aspect of it makes sense to me? I'm curious, like there's a version of this where again the future is so automated and blah blah blah. There's another version where like humans make the AI better through the things that make them human. And then AI makes humans better through like output and stuff. Like do you believe that? How do you guys think about the like, you know, again there's that AI won't replace you. A human who uses AI will replace you. Like do you believe that? And I mean that more, less from the like your, your worldview and more from like a specific, the way Work gets done with AI in the future, now and in the future.
- [53:47]
B
Well, I mean my one thing I would say just specifically to the, the truism you just said is I really just think it is like, and like basically every executive is like on one hand being like how can I use AI in my company, on the other hand is saying how do I make sure my people aren't worried about it? And then behind closed doors is saying this is 100 going to wipe out a ton of jobs in my company. So look, I mean, Armand will have a smarter thought here, but my, my general view is like, if we view like, I view AI today as like this incredible junior employee that just like a great junior employee is high agency, high effort. You, you delegate. Like, you build up trust initially with like really defining parameters. And as it earns trust, you allow it to do more without you checking in on it. As intelligence gets better and better, I do think you end up getting like employees, quote unquote, employees that look more and more like senior employees. Which just means like at the end of the day, what's a job of an executive? It really becomes around like orchestration. And now where I'm not smart enough is what is the point at which it becomes intelligent enough that it is good enough? It, it can orchestrate well as well. I don't know. I don't know the answer to that and I don't know if anyone does. I don't know if you have thoughts on it. Like, if your question is like, are we out of jobs in the next five to 10 years? I, I would put it at a non zero chance, but until that point we're enjoying working on this problem.
- [55:24]
D
How are you guys thinking about the framework for your margins and your growth as a services business?
- [55:28]
C
What do you mean? Like frameworks, like what is our goal?
- [55:31]
D
It's a really, really important thing. So, so here's, here's what typically happens with services businesses, how they go south, right? One is pipeline. Like the amount of work getting done versus clients that coming in. Like you're selling people, right? And I don't mean that in a weird way, but that's what you're doing fundamentally. And typically what happens is deliver delivery and customer acquisition, they start to bleed in a bad way. Which means, you know, oh, like you have this pipeline, then the pipeline gets too busy because you still need to be involved probably in certain ways. And like, then that's one way you start to get into this, like this sort of swinging the other big way that's related is then margins creep, margins go down. Right. And margins typically go down because scope creeps up by the client. You're not, people aren't paying attention to it all. Like I see a lot of services business where they're just not even paying attention to it. I talked to an agency a couple of weeks ago and I was like, what's your POD level margin? They're like, we have no idea what are you talking about? And they weren't directly. They could do the high level number. They could say here's the total revenue, here's my total headcount. And then some clients are 10% margin, others are 70%. And, and so there's a lot of frameworks I think that, you know, it's important for you guys to build, to crush this, to start to think about some of these questions more directly. And maybe you have, maybe you haven't, but, but that's sort of just what I meant by that.
- [56:48]
C
Yeah, I mean we have the margin down to the day. Like, like we know our margin per employee because we're, we generate, we charge on output and we pay on output and everyone's aligned. And so we're very, very strict with thinking about that. And again, the goal is so that we can reinvest in the company and build IP for the customers. And so it's basically that flywheel.
- [57:14]
D
So what drives your margins up versus.
- [57:15]
B
Down one is that every additional time we do work there's less human, there's actually less cost of labor that goes into that work. So one example for that would be like, and this goes into like how we ultimately want to monetize the business different than typically service businesses. What does it look like if Andrew, let's just use an example for. I'll just use an example for. Let's. We're talking to a business, their sellers spend more than 50% of their time not selling. We're talking about building basically just like an a sales co pilot for them that does a lot of like SDR and sales ops work. While this business is like a health care business, 90% of the bones of this agent can be applied to every other business that has a sales function. What does it look like for us to basically one build this agent at cost as 10x and then charge them based on usage of the agent. But then also every additional time we work with customers that potentially need a sales agent built, it's effectively Lego blocks for doing that over and over. And so the margin increases every time we do that.
- [58:25]
D
No, I love that. Well, but so One of the issues with output based pricing, Right. Just to flag it for you guys.
- [58:30]
B
Yeah.
- [58:31]
D
What happens when the client like, what happens when the output isn't good and you've committed to the output? So all of a sudden you, you, you know, let's say your costs are 50 grand and you're selling it for 100 grand in this example. And they go, we don't like this. And you're like, I got to put another month's worth of work into it for whatever reason, client's unhappy, maybe they're jackass and they didn't listen when you were scoping it and vice versa, whatever.
- [58:53]
B
Yep.
- [58:54]
D
Then what happens is you spend another month and all of a sudden your margins went from 50 to 30%.
- [58:59]
B
It really is just a question of is it realistic to fully incentive align you with a customer? Because if you were to truly incentive align, I guess it would be like truly value based, where it'd be like, if this outcome happens in your business, we get paid. And I know like the big consultancies do this. I mean, I guess my question for you is like, do you, do you just think that's not realistic for service businesses to charge based on value or output?
- [59:24]
D
No, not at all. I mean, that's what we did at Ambush, right. I mean, we charge that way. It's just you have to build the infrastructure. There's a lot of pieces that are, that require like more thoughtfulness than you might think. For example, like, you know, one of the, one of the quotes like Rick would always say, and this is in a marketing context, right. Because we would take on people's marketing and we'd promise them better results.
- [59:43]
B
Yep.
- [59:44]
D
Right. And it's like, well, at first you're like, wow, these guys will take a risk and they'll. And it's. Well. But what's really happening is like, we're like an investor, we're underwriting the opportunity. Yep. So, and the line was like, never, you know, be like Floyd Mayweather. Never take on a fight you can't win.
- [59:57]
B
Totally. Yeah.
- [60:00]
D
Well, specifically though, in your, in our case, it was like we already knew the levers we would pull to be able to reduce their CPAs or grow their volume by a certain amount. And we didn't know it was work, but we were very confident more often than not it would work. And so for you guys, when you think about costing based on outputs, what you're going to find pretty soon is you're going to want to narrow the things you start to focus on. Like your example of the SDR thing because then you know, you can keep delivering it, delivering it, delivering it and you're going to have a built in margin model versus the more wide, like, yep, you know, wide industry wide problem set. That, that, that runs into that problem which is like you, if you can't start to predict it, it starts to really take a big hit on your margins.
- [60:37]
B
I think that, I think that makes total sense. One last thought, because I know you, like, you basically talked about acquisition costs and like delivery moving the wrong direction. You know, one of the things that I've wanted to do since leaving Morning Brewing is effectively build Morning Brew on top of a better business model. And, and so ultimately the goal is to build kind of 10x media on top of 10x where there's only one place that executives go to to make AI more actionable. And that's like the hook again. And over time it can develop into, not solely around AI. And so that's the idea is a combination of brand and personality driven franchises that make AI actionable for kind of the next generation of leaders that are trying to transform their businesses.
- [61:21]
A
Come on, guys. Now tell the audience your revenue. Can we say more than 10 million?
- [61:28]
B
Sorry, Andrew, you're cutting out.
- [61:30]
D
All right here.
- [61:32]
A
Gentlemen. Congratulations on the success. Thanks so much for doing this, Armand. I'm seeing you wince like you're regretting the friendship we even had just because I pushed it too far. I'll win you back later.
- [61:41]
D
I've got an awkward question for Arman. Why is this business not on your LinkedIn?
- [61:46]
B
It's.
- [61:47]
C
It's going on there, okay?