#478 100x Outcomes, Not 10x Hype: AI Execution Strategies with Matt Leta

Episode 478

[00:00:00] 

Mehmet: Hello and welcome back to New President of the CTO Show with Mehmet today. I'm very pleased joining me, Matt Leta. He's the CEO and founder of Future Works. Matt, thank you very much for being here with me today on, on the show. Uh, I really appreciate that. [00:01:00] The way I love to do it is I, uh, keep it to my guests to introduce themselves.

So, uh. You know, I will leave the stage for you, but just as a hint to or teaser to the audience, we're gonna talk really about cool stuff, including future of work, indeed and AI and the age of ai. And maybe some of you you might be familiar with, Matt, he did something, um, you know, last year I think we're gonna talk about.

But without further ado, Matt, the floor is your. 

Matt: Thanks so much for having me Maed. Um, for a quick intro, um, well, I'm originally from Poland. Left when I was 18 to study multimedia. At the time I wanted to be an, um, an an interactive artist. I wanted to create art that interacts with people and, um, and I did, um, but I needed some money, so I went, uh, started working in marketing a bit.

Um, I had a kind of like a brief career [00:02:00] in marketing for, for a few months where I managed to save up quite a bit of money, started creating art, then I started making websites. And when I was about like 20 years old, um, I connected it all and built my first startup, which was a platform for artists. But they would come together, share what they create, and that was in the uk.

And, uh, since then I've lived in Spain and Chile and many places, Ireland. And, um, eventually, uh, sort of later in my twenties, raised a few million dollars and moved to California. And, uh, I eventually ended up exiting the startup, went traveling the world. And as I did that, I created a kind of like a digital product studio company that would help Silicon Valley companies with.

Innovation with digital product, and that went really well. I had that company for about eight years. It was doubling every single year. And, um. [00:03:00] On average. And, uh, and so that led me to kind of start seeking what the future may hold and, and trying to, um, navigate, you know, problems of the future. And so I created sort of a community around that called Future Horizon with a series of gatherings and a bunch of initiatives.

We helped create a number of nonprofits and fund them. Um. And, um, and with, you know, the future thought meant. We've been very close and working quite a lot with AI companies since sort of 2017, so quite a while now. And, um, and eventually, um, couple of years ago we realized that, you know, we can now actually use generative AI to help us deliver work.

And so I created future works. Which was an experiment. At first, it was just an experiment to see if I can co-create a company of [00:04:00] AI as my partner. Um, that was in early 2023. And from there it has grown to like a multimillion dollar, um, company that still operates with a kind of ai At its core, it's much more complex, much more advanced than it was then.

Much more custom, but it's still there. And as we did that, I noticed that it is not so easy. To get through ROI from ai and that's from our internal hurdles and from work with clients and that's sort of a primary are kind of mid-size US companies, but also a bunch of enterprise. And I noticed that, you know, there is more to it than tech.

And so I started really diving deep into the subject over the last two, three years. Interviewing a lot of people, reading a lot of material, doing, you know, programs at Harvard Business School and so on, just to kind of, to really understand how can [00:05:00] we truly drive change for, for businesses and ourselves, uh, using ai.

And so that became subject of my books. I think that's that. I live in Cambridge, Massachusetts. I live by the, technically on or within the Harvard campus. 

Mehmet: Cool. And thank you again, Matt. Um, I think people might remember, uh, and you know the hustle GPT, so this is what you were referring to, right? 

Matt: Yeah, yeah, yeah.

Mehmet: Right. So I, I, I remember, you know, this became like everyone wanted to also kind of, uh, take the idea, uh, um, so. You mentioned something about the ROI, I want to, you know, just start from there. Okay. What did you learn from that experiment? And, you know, now everyone's saying like, Hey, your AI can be your co-founder.

Like, uh, go try, you know, to, to, to let it do things with you or like, maybe I. Like behind or beside, I would say [00:06:00] especially 2023. It was the beginning of the whole, you know, generative AI and chat GPT, but behind all what, uh, you know, the lights and noise that were there. Really. I want to understand from you, can we really build businesses with ai or is it just like a good place for.

You know, maybe giving you some ideas, maybe just, uh, it can act as a, you know, mentor probably. What, what was like this experience like, and what, what's the biggest learning for you? 

Matt: Yeah. We absolutely can build, build businesses with ai. Um, and not only that, we can transform existing companies to have, and, you know, and become increasingly AI native or have like a digital core.

That uses quite a lot of AI to augment work, um, and achieve much higher outcomes than it would be, uh, with the sort of default. Like most companies right now that have an AI [00:07:00] mandate have some sort of internal systems, you know, but generally, uh, what happens is they. Have their employees use chat, GPT and perplexity, and maybe they implement, you know, few other tools that were pitched to them by one of the 5 billion AI startups.

And, and so, uh, and, and that doesn't really bring ROI, I mean, it's, you know, it, it does make our teams, our workforce, you know, faster to some degree. And accomplish more. And I feel like many of those, uh, are mentioning, you know, the gains that they're getting, but it's not going to be like a two x or 10 x and definitely not a hundred x.

It's more so like 20% or 5% sort of improvement on, on, on output from the team. And so, um. It's important to think, [00:08:00] to go deeper than that, right? Like rather than have people just treat this as a sort of ongoing advisor or strategy helper, and especially if they're using something like chat GPT or, or cloud, which you know, to today still love to hallucinate, right?

And I think there is just not enough attention paid to it. And those systems are not necessarily ready for our teams to, to trust in as much as they do. And, um, and so, uh, we sort of need kind of deeper solutions. We need to build the AI systems deeper into our workflow because if you have a, a scenario where people work, let's say they work based on tasks and documents, right?

They handle, you know, let's say emails, internal messaging tasks, documents. If all of this information, I. Is live, connected, uh, to intelligence, and that intelligence is sort of purpose fit to work with [00:09:00] those people, help them learn, but also learns from them as they go. Then that enables a much higher ROIA much, much higher augmentation.

And one person can suddenly start doing the work of two people and then five people and so on, right? That's impossible. If they just were to use Judge GPT. 

Mehmet: I, I was discussing this, you know, uh, with a friend the other day, and of course with my guests as well. So do you think like people, you know, they think like the ai is that just a magic wand and, you know, you just do like this and all of a sudden like, oh, you need, you can get rid of all, you know, the, the, the, the people maybe you have currently, so you can like do more work with less.

Or maybe some people they think, yeah, I will give it to the AI and let them decide what what they want to do. By the way, like the, the episode before this one, and I'm, you know, of course we are recording now and we're gonna, but I know the sequence of my episodes Uhhuh, so my guest [00:10:00] s yesterday, you know, when I was recording with him, so he was mentioning that actually.

People sometimes exaggerate, you know what the AI can do. And actually he mentioned about like, it's more about people should think about the automation use cases more than rather the AI as generative ai. Because generative AI at the end of the day is just like, you know, you, you give it instruction and it come back to you.

So from your perspective, Matt, now. I'm, I'm mentioning all this because I want to come to a point, uh, with all this being said, and we are recording this on 15th of May, and I think this news came either this morning or yesterday, that Microsoft are laying off like 6,000 people, which is like 3% of their workforces.

So someone might ar uh, argue and say, Hey, but look, it's working. So that means, you know, AI can replace completely. People's job. What, what's your take on this? Like, is it like different from a giant like Microsoft versus maybe a startup with, I don't know, maybe we are only 20 people, so, [00:11:00] and you have experience in both sides?

I would say, 

Matt: yeah. I think, uh, firstly to address that, that, that, that point, you know, uh, you know, automation is a good lens to look at it. Because generally if you look at incentives within the business, which is always to optimize, right? Always to increase margins so that you can outcompete and stay relevant in a market that constantly heads towards higher optimization, lower cost, and so on.

That's just how economy works, right? That's how capitalism works. Um, we. Uh, we cannot, we safely think, and that's nothing new, that everything that can be automated will be automated. Right. Period. Mm-hmm. Everything, unless there is legislation that somehow prevents it from happening. If we are in free markets, everything that can be automated will be automated because it's cheaper to end if the outcomes are comparable.

[00:12:00] Then, you know, whatever is a more expensive way of getting some outcome is going to disappear. Uh, right. And that has happened many times in history with, you know, electricity came and whatnot. And, um, and so I. Now, um, automation, you know, intelligence is so kind of automation, right? But, but you know, you can put it in those frames and have workflows.

A lot of the stuff that people call agents right now, there's not really agents. They're very narrow workflows that have some form of intelligence, like, you know, if this, then that and it writes something, comes back with some. Response, you know, based on some input, right? It's not really an agent, but a lot of companies call them agents because it sounds, uh, uh, sounds, uh, better for, for, yeah, it's easier to sell this stuff, right?

So there's a lot of hype and a lot of disappointment. There are the agents, but true agents are also coming, and this is very important to understand that then intelligence is a bit different to automation [00:13:00] because it can reason, it can decide, it can act on its own, it can choose its tools and there is like a global.

Network of tooling being developed with, um, you know, MCP and other approaches. And so tools are, uh, they're sort of opening themselves up to be used by those reasoning systems. And so to that degree, those systems are an automation in essence. It's not a person doing the work, it's the system doing the work.

But the, the, the, the path to achieving an outcome is not predatory mind. And I think that's, um, that's quite important to observe and get ready for. So I wouldn't narrow down, uh, it to, to automation. I would rather treat it as a spectrum where there are different types of solutions for different problems.

And the most important thing is to look at problems that we have that we can solve using intelligence or just software and data in general. And solve those rather than, you know, adopt [00:14:00] things for experimentation sake. Uh, or for, you know, just because there is a shiny tool. Because that's to the point of, uh, I, I, I, I did, I didn't quite remember the name of the, of the previous guest, but to his point, yes.

I mean there, you know, there's different, again, incentives. Sometimes you're just incentivized to, to report how much of an AI company you are. Often our clients. Our companies who announced two years ago, like now we've got, you know, whatever GPT, and we've got so much AI going on. But the truth is, no one's using that.

I mean, any company that had like a rollout of Microsoft copilot or stuff like that knows that no one's really using this stuff, right? And so, or Salesforce is equivalent and so on, right? And so everyone started going back to the drawing board, okay, why are people not using it? And that is the very important.

A limit. So, um, you know, a lot of hype, a lot of trying to pose as, okay, we are an AI powered company, but the truth is that as much as leaders want ai, [00:15:00] because again, um, that's how we out compete. That's how what allows us, our companies to grow. Our people don't really want it. Right. And, and, and, and there's always a percentage of people who are sort of.

Unsure about it and but would want to learn. Uh, but I think there's even a higher percentage of people right now who don't necessarily even want to learn or use it. They don't trust it, they don't want it, they don't think it's right. They should be doing their thinking. Why did they go to school to, you know?

And so that navigating that and the cultural aspect of it is sort of like a bigger challenge. So to see this all work, kinda like the technical side, so like automation, tooling and stuff like that is just like one part. It's actually quite quick to see results with that. But um, in order to fully build it into our organizations and have, you know, organizations become sort of AI native, AI first, we also need to get buy-in from our people.

Right. And I don't know what the [00:16:00] situation at Microsoft and how did they choose. Those people. But I know that I am personally guilty of realizing after two, three years of trying that some people simply will not open to ai. So it's not just that they are outperformed by, uh, you know, suddenly you don't need them.

Because I think a company like Microsoft is growing probably. I mean, I don't know the, the situation there on the inside, they're probably growing fast enough to reassign people. But my suspicion is. If you have an internal agenda to become an AI powered company and you realize that this is the future period in, in the business you're in, they're in tech, right?

There is a, a, a percentage of people who you can, we can think of them as like organic minded or, or so, you know, who will just not accept working with ai. They will actually sabotage our businesses. [00:17:00] So I've let go of over 20% of my team eventually, after two months, two years of trying and really throwing everything, trying to, you know, sort of awaken.

We've literally built a framework, uh, you know, that I published my first book on to awaken everything, everyone to this. You know, like, you know, to to, to innovate and to embrace AI and to become a part of those people who truly drive with it, but some just will, just won't. It's just, it's an internal conflict.

I think it's like a value-based conflict and we were not successful at getting everyone, uh, across. And I think maybe, you know, I don't have like correct data, but my assumption is that maybe every company will have like 20% of people. Who are simply just completely not uninterested. In fact, they're so against that.

They're going to try try to sabotage, even subconsciously, sabotage initiatives to bring in ai. And so I [00:18:00] think that the changes in personal, as companies realize that are going to be quite common, but it doesn't necessarily mean that those people are now. Um, replaced by ai. I think it's more so our workforce who use AI is as a whole more productive.

So let's say we can keep the same amount of people, but we can do 150% or twice more. If you're a smaller organization, we can get there and I think that's, that's very available. And of course, if you can't grow in the market, if you can't increase your revenues or what would that be? Then what? You can then, then, then you can lower the amount of the headcount that you've got.

But I don't know. I don't know. I would, I would imagine that it's not Microsoft's problem. Yeah. Know 

Mehmet: indeed. Now I, I, I gotta maybe mix two things together, but I believe they are related. So Yeah, this, we don't all, what we've seen in, in, let's say two years now, or two years and a half since, uh, charge [00:19:00] GPT came out.

So we've seen this, um. You know, initiatives that are coming actually from up into com. Big companies, like on board levels, we need to be an AI first company, right? And go. They go and tell you whoever is responsible for the technology, go and figure it out. This reminds me. Of what we still till today, we call it digital transformation, so it's kind of another transformation that's happening now.

I really have done some, some work also on that, and you have a, let's, let's say one hour per week transformation. It's like a thing that you, you say, now if I want to compare. The era where the digital transformation concepts start to come. Like, let's digitize everything, put it there, and actually people start to struggle.

Of course we know how, like they, they didn't know how to digitize the processes and so on. So now we have also the AI part of it. How much similarities there and uh, you know, [00:20:00] things in common you see in these two transformation, uh, Matt, but at the same time, what do you think the blind spots that people have in both situations?

To implement real transformation, not just, you know, to say, Hey, we are doing this transformation. We are, we are shifting to AI first. From your experience, what you can share about that? 

Matt: Yeah, it's a great question. Uh, well, you know, I've, I've spent a lot of time diving into the, and in fact I spent like a year asking people over 200 different executives, leaders, and so on.

Whether to call this digital transformation or something else. A lot of people like to call it AI transformation now and so on. And you know what? I've concluded that AI transformation is too thin because it focuses on technology only. If we focus on technology only, let's add it. Let's, let's sort of strap AI onto our organization and hope for the best.

That doesn't work right. In the last two, three years, I think a lot of organizations notice [00:21:00] that. Like they've had some really interesting shiny AI initiatives that there just didn't break much. Uh, ROI, again, going back to the result. And so, um, we've dug into this and. High level. My, my view is such, we've got three, uh, major sort of waves of digital transformation.

The first one was like from analog to computer, right. The second one was from just computers to networks, right? And the third one is from networks to intelligence. And each of those was like two, three decades, right? And the current one, I do believe actually will take us to three decades. Right. So it's not like we'll have a GI, we perhaps already have a GI, but, but it'll not be so immediate to see the effect in the world.

In isolated cases, of course there will be, you know, companies that fully outcompete with tiny teams, and there will be those who adopt faster than [00:22:00] others and wipe out entire markets from, you know, from incumbent players that are willing to change. But, but as a whole, globally. You know, uh, things will take some time.

And so I actually, uh, goes from that research. I got to fully believe that what we are doing now is essentially what we call next gen digital transformation. And I like this term. It doesn't come from us, it came from our friends at Ideal originally. When we discussed this about two years ago, and then we've kind of poked around it, tried to understand, you know, what is the best term that would sort of make everyone feel like they get it.

Like they understand what will be needed in order to achieve results. And so we've, we've, um, we've, we've settled on next gen digital transformation. This is what we do, and that has sort of three components. And if you order them by like time to value, like what is the quickest to [00:23:00] see results on its own, then AI intelligence is the fastest, right?

You can see results within days really if you implement something. But for that to happen, you need to sort of have like a ready infrastructure and so on. So then what? There is a middle layer, which we call software end data, which is something that came sort of with the previous wave. Right. And, and that's, um, essentially the, you know, everything that operationalizes our companies, you know, ERPs and CRMs and tools and, you know, the cloud stuff and so on.

And then, uh, we have people, uh. And now the, you know, the software and data takes like maybe two years to, to really upgrade it, right? Or sometimes, sometimes it could take less if, if, if you're able to move fast. But let's say it takes two years to really get to a place where you've got like clean data like houses.

You've got, you know, data sort of flowing out of silos. [00:24:00] You've got an interoperable composable. I, I, it's, it's a CTO show, right? So I'll be using some of those terms, but, but, um, you've got, uh, you know, you've got essentially an application ecosystem. That allows you to swap components, move them around, uh, you know, um, and, and build this sort of, uh, flexible fabric.

Um, um, you know, where data can, can, can, can freely travel and become accessible to AI and connect, uh, connect how people work and how, what ai so, uh, you know, augments or solves or, or again, going back to automations. Simply automates and, and, and then people are needed to actually do it anymore. And so the software and data component is still incredibly important.

What we've noticed, a lot of companies are simply not ready because they're locked in into some, you know, CRM on EER or ERP, I'm not going to name them, but everyone knows the big ones. And. And they, they, you know, they're, again, going [00:25:00] back to incentives. Their biggest incentive is to lock us in for many years.

Keep everything in there, become, you know, database or system of record and not let us look elsewhere. While their AI efforts are typically like an acquisition or acquisitions of some third party AI systems that are not really well integrated and don't really bring much value. And so to solve that, we, you know, ideally, uh, you know, want to.

Spend time to free, free, free all of that data up, making sure that we have it in some sort of data lakehouse somewhere where it can be accessed well, uh, by ai. And that's where we are starting to see a lot of value. And the third component that I mentioned briefly is people, right? And every company is essentially made out of people.

Right. Still. And I, I would say that very few will be able to be fully automated, right? And so we, uh, we want the people to sort of be open [00:26:00] to this change and welcome AI and co-work with AI and, and become augmented. And then we can start seeing each of those people suddenly, you know, deliver two 10 times more, even more.

Right? And, and, and we already are seeing that with many people. Currently. There is, like, if you compare one person to another, suddenly. One person is five times, you know, has a five times higher output than the next one just because they use certain tools and approaches and, uh, and, and they're kind of both into that.

But for that we need culture change. And culture change can take, you know, three to seven years, let's say, at the company, right? Mm-hmm. If you have not started yet, if you are not, if you don't have like an ongoing focused efforts to make your workforce, uh, you know, fully AI powered and ai, AI friendly, let's call it.

They will take their time, let's assume five years. And so it's so important to start basically as up if you haven't started, start tomorrow and, and start leading in that [00:27:00] direction because again, AI can take few days to value software and data are solving that. Uh, application ecosystem can take a couple of years, but people will take their time.

And, you know, we don't want to, you know, Microsoft fired 3% of their people not 97. And we don't wanna, we don't wanna be replacing, uh, everyone or, or using such adjusting methods because in the end they have the expertise. And they, uh, are the, uh, the, the kind of core fabric of our organizations. 

Mehmet: Yeah. This brings me to, to, to ask you about the book, actually, Matt, because Yeah.

I think, you know, the, the, the book title is uh, a hundred x, right? Yeah. And it's for executives. Uh, it's called the executive. I mean, the subtitle is an executive brief for an AI driven business results, and you, you, you just like did, so what, what? Uh, what executives can, can, you know, I mean, of course you'll not, will not say what's inside the book in details, but what they can expect, like what kind of lessons [00:28:00] they, they, or like, let's say motivations, uh, they can, you know, get from, from your book.

I. 

Matt: Sure. So I wrote a book last year or published a book last year. That's that. I spent a couple of years writing before that. Um, which, um, is a kind of like a step-by-step guide, how to awaken our people and how to get a lot of ideas. Uh, to implement AI and, and step by step kind of in an exponential fashion.

We're getting our organization to become AI first, but that book is quite thick. It's very detailed. It gives, it gives sort of like a step by step playbook with, you know, what to avoid, how to solve issues along the way and so on. I noticed that it's quite a read. And, uh, as much as, uh, it's adopted now by hundreds of companies, and I'm hearing great things actually, uh, about the implementation.

I really wanted to just, as I was finishing that book, I really wanted to create a different [00:29:00] one that's, that's just very short, very pointed, and can. Sort of help executives create a vision for, to, to even to, to drive this practice. So it's less methodical, it's less of a practical guide, but it's more so going over key concepts and elements that would allow us to sort of shape this AI first, uh, organization, right?

And help, help even shape our strategy. And that again, comes from my conversations over the course of over. Uh, two years with 200 plus, um, executives, leaders, consultants, you know, AI experts from the largest companies and down to some pretty fast growing, exciting startups and, and, um, and, and a bunch of other material.

And originally this was, I think the, the, you know, the material for this book was around 600 pages and we spent about half a [00:30:00] year. Crunching and crunching and removing and compressing and organizing until I got to where I wanted to, which is around 200 pages. It's, it's short. When you get this book, you are like, okay, that's, you know, it's a, it's, it's fairly thin.

And, um, and that was my whole dream to make a book that's very thin, but highly information packed. So that kind of like out of every line, you get some value. It was not easy to get there, but, uh, but basically it takes us through 10 chapters that cover sort of what I just talked about, but it much in much more, you know, detail and nuance.

Give some examples. Um, many examples pulled out of the sort of, um, 150 like, uh, case studies that we're aware of, uh, you know, that we've studied. Um. And, um, and out of, you know, around 500 that we've looked at, but around 150 we've studied deeper or, or are, were our own. [00:31:00] And, um, and we've, um, um, and it essentially kind of covers all areas from like, people, culture, technology, uh, you know, AI sales agents talks about, you know, actually demystifies agents a lot.

Where we are with them and, uh, where, where are we headed? And then talks about, um, ethics often overlooked, but incredibly important if we want to buy in from people. Um, and again, that's the biggest limiting factor to actually for us to become an AI first organization. And, um, and, and it sort of concludes with, uh, a kind of like a plan and a playbook, how to put it all together and how to, um, how to start, how to, you know, get it going.

Mehmet: Cool. Matt, I want to ask you something. Of course, like you, you, you've done it, uh, as, as I say, all, and you know, you've been in this startup, you've been like an entrepreneur yourself maybe. So people will say, yeah, Matt has been in this world all, you know, all his, his career, right? So [00:32:00] he started his first business, you know, you were 19.

Now, how we can as companies, you know, do this. And the reason I'm asking you, because when we talk innovation. And I'm a little bit biased here by, you know, the book Innovation dilemma, right? So 

Matt: yeah, 

Mehmet: it looks like, you know, at some stage, especially when we become like a large enterprise, so I can understand this.

First I was getting angry, but I understand like, so the bureaucracy is the destiny of large corporations. Yeah. And this is why things slows. But now how, how do you, you know, think or like, what's your expectations? These companies that are started now. Right. And you know, they, they've been built on actually on AI technologies.

How do you expect them to keep innovating moving forward? Are there destiny to stay like, you know, in, uh, you know, innovating all the time by keeping maybe their [00:33:00] teams smaller or there is no escape? At some stage they're gonna come very big and, you know. The innovation will slow down because of all the reasons we know.

Matt: I think there is like a sort of arbitrage opportunity in terms of outputs per person in short, medium term. So let's say there are several years. Where we, where we see a huge gap between individuals, a person that uses AI and uses it every day versus not. Uh, then next level would be a person that uses ai that is purpose fit within their organization for how they function.

That like multiplies that for much more and an entire organization using AI versus one that doesn't. I think there is a, there will be a huge difference. I mean. [00:34:00] The literal a hundred x right in is, is possible in terms of, you know, how much they can cut their costs, how much they can grow with in markets, how many customers they can reach, and so on.

Right? And we can see that with a lot of the, the new kind of. AI startups which have teams of like, uh, you know, 15 people and gen generate, you know, 200 million A a RR. Right. And, and there are more and more of them all the time. And in fact, the, you know, the percentage of startups that become unicorn is the highest now than, uh, and it's been spin, uh, despite the markets.

Right. And I think that will probably. I mean, no one knows the future because there might be some interesting curve balls thrown at us. But just the fact of, of sort of a SI super intelligence coming to play and, you know, that is very conducive to consolidation in the market. Like it is [00:35:00] possible that a company like OpenAI could suddenly be able to do everything.

Right. And, and acquire everyone else that you know, that that stands in their way. Like it's totally possible, right? End of competition. But you know, within the realm of our current economy and how things are, I think the arbitrage ends when everyone catches up, right? And so in the and, and I think either they catch up or they die.

You know, that's just how it is right now. Like if the company is not already embracing those technologies, if their teams are not becoming ai, augmented. They will go out of business. There's no other way. Right? Like all the companies that went out of business when they didn't embrace electricity and, or they couldn't because their market just got cornered by, you know, like ice trade for example.

You know, refrigerators came in and, and we don't need ice trade, right? And so all those, all of those companies went, went out of business. Um, but, but the, my bet would be. That the [00:36:00] arbitrage is gonna lower. Once things sort of like stabilize more in the sense that everyone will be AI augmented and so and so, there will be teams and the teams will be still growing, but the teams output will be much, much higher than would have been previously.

That's between, you know, two x to 200 x and more, depending on the industry and, and, and, uh, you know, and so we'll just see much more productivity overall, but the teams will still, I, I, I would imagine Will, will grow. Um, but no one will need as many people as it would, as, as in the past to, to achieve what was achieved in the past.

I think that's, that's, that's pretty certain. Um, yeah, that's already true, right? Like we have, you know, I used to have 80, 80 people at my company. Now I have nine full-time. You know, it's a, it's, [00:37:00] it's, it's, it's, it's, uh, it's, it's big, big, big difference already. 

Mehmet: Yeah, so just regarding, um, uh, you know, the number of companies, sorry, the, the company that they have less number of, uh, people.

Uh, so you just gave an example. I know, like maybe referred to some of them. So lovable for example. They, they, they have like, and I don't, maybe a hundred millions a RR now. Uh, and of course the bolts. And you know, and this is the reason I mentioned these companies because we start to see the vibe coating.

You know, movement of course. Um, we see it more on X rather than LinkedIn maybe. And with Vibe coding, do you see it like something similar to what you tried to do with Hassan GPT but now on steroids because now it can actually Yeah, totally dis design a full fledged app for you. Like, because Hassan, GPTI, I remember, you know people were talking about Yeah, you were just.

You gave the famous prompt, like, just give me instruction. I will go and do it. Nothing [00:38:00] illegal, you know, and I, you know, I will just follow the steps one by one. But now with, with the, you know, tools like Bolt and others, so, and versatile. So you just give the prompt and actually do the fully fledged thing for you.

Do you see some similarities in this? And we might see like more people trying to do similar, but you. We are trying to do, uh, you know, two years ago. 

Matt: Yeah, well I know personally a whole bunch of people who are actively working full conviction and I would say very high chances of success on building this mythical $1 billion company of one person.

You know, I think that's absolutely happening. I. Is this going to be like prevalent in the markets? I don't know. I don't think so because it takes a certain kind of person and, um, you know, it's still much nicer to, to work with people. Um, and, but, but, but there are, you know, people working on this and it's much more available than it was.

However, however, [00:39:00] something is really inter I realized something very interesting. Um. Uh, when we were discussing things at Harvard Business, um, a few weeks ago, a month ago, I looked at my post from two years ago, which sort of did a recap of my first initial 30 day run. Uh, when I started this, you know, this, the future works, essentially, which the name came even from ai, right?

Mm-hmm. And, uh, and, and, and this whole run of the documenting that process every day of how I'm building this company, everything was vibe coded there. That was two years ago. Right. Everything a hundred percent vibe coded. I didn't hand code anything. And, and that what, you know, in an app, you know, and a web app deployed and stuff, right.

And, and even client work was part vibe coded. And that was, you know, at the beginning of it, you could say all those tools didn't exist and Cursor came like three months after or something, like the first version that we started playing with and popularized like a year and a half later. Mm-hmm. But, um, but we, uh, [00:40:00] um.

Not much. I noticed that not that much changed. So GPD four when it came out was a very strong model and then it got like kind of dumbed down over time. And now we've got, you know, all those 50 different versions of models after that from open ai. Actually, they're not that much better, you know, they still hallucinate, they still have a lot of issues and, um, and they, it was, it was just a lot of infrastructure was built around them.

So now we've got this. Agents and applications and workflows and tools, but actually the ACT itself has not evolved that much. If you look at the, you know, that's, uh, there's a medium post. So if you look at mat letter and then on medium there is, uh, or just if you type like, um. Million mat letter medium, you would, you'll find this recap because it's like had like a $1.8 million pipeline or something within 30 days.

The findings that I've shared. Then [00:41:00] I was just, I was shocked our findings that I would write now, and so we think that we're on this like crazy exponential and we are to some degree. But the truth is that actually a lot of the modalities have not changed. And in fact, it was possible to vibe code then and, uh, it is possible to vibe code now much better.

You know, there are, there are all those, you know, there's MCP, there are new better approaches. You know, you would create first documentation library, then have something like Cursors or, or V zero work with this, these materials to, to code, you know, and, and you end up with results. And the results are, are better.

And back then it was like copy pasting. There was no IDE that had, you know, that, that would, could, could do this and, and, and, and, and, you know, and, and it would just approve changes. Right Now there the, the is that you just like, you know, vibe coding is just like you press. You, you press command, enter all the time.

Control enter. Yeah. And, but, uh, [00:42:00] and prompt, right? But, uh, I, I think a lot of people right now are kinda haha by coding is like, what is this? That's not, not gonna get there. Look at this code. It's garbled and not, uh, not accessible. And whatever would be the argument of those people, of those people that kind of want to imagine a reality.

Where AI is not actually taking over everything. Right? But that's not where we're going, right? Like if you just look at where things are going, this is not where we're going, but it's like a reaction again, going like from volume misalignment. It is. Coming from people who have been developers, let's say for a few years, they're like, aha.

It doesn't, it doesn't produce good code, right? But it's the same haha that we had last year where people were like, oh, look at, like, look how many fingers there are on this image, right? Everyone was counting fingers on images. It's like, oh, this is AI generated. It has six fingers, or whatever. The hands are a bit off.[00:43:00] 

But then a year later, a lot of, you know, main Street advertising video and stuff like is made by ai. There are five fingers and there will never, there will never be a return of multiple fingers. Right. The problem has been solved. We don't realize that this content is generated by AI anymore. It's just too good.

It's, it's real, it's, it's just as good as photography or design or whatever. If, if it's prompted well and if it's done well. Right. And, and it's the same with coding in my opinion. And the same with a lot of other things where like, it's easy to go to like, haha, this, this does not, you know, this is not as good as a good developer, but give it a year and it'll be, and then it'll be much better than any developer, you know, O three charge GPTs O three already.

Crossed, uh, you know, Norwegian Mensa, I think crossed 130. It's basically smarter than, than mm-hmm. Vast majority of people, and that's within one year. You know, I think last year we were at 70 or [00:44:00] something, and now we are at 130, and of course, that's, there's a lot of optimization just to pass those tests better rather than actually deliver better value.

But I think, you know, if you give it another year, we'll also be much further, it'll be smarter than basically anyone alive and so on. And so, um, it's really good to not to get trapped in the haha when looking at vibe coding, but rather learn it and learn it fast to be relevant in the market. 

Mehmet: Absolutely, absolutely.

Just a note, and this is not my own thought, you know, uh, regarding, because I, I have the same, um, you know, exper experience about the models that after a while I call them, they got lazy, right? Yeah. So, yeah. So I re I remember the, the, the first, you know, the, the, the chap chip, T 3.5, uh. After like maybe using it for two, three months, and by the time before the four came, uh, I [00:45:00] felt like yeah, it's, it's not, it doesn't want to work, I think.

Yeah. Yeah. Triple five. Yeah. Yeah. And same start to happen with the different four versions, because OpenAI keep, you know, uh, making us guess. Is it four o? Is it 4.5? And anyway, I, I, I saw like the same behavior. Um, I didn't see it much in Claude or the others, but even so, someone told me, yeah, they open ai, they do this on purpose now.

Is it true? No, I don't know. So they said because they want to manipulate people and they want to show that the new model is now there is 

Matt: progress. Yeah. 

Mehmet: Yeah. There's a progress. I'm not sure. Yeah. What, uh, what I like, you know, for example, and I figured out that so far it's, it's behaving good. So the other motives of course.

Still I use on daily basis chat, GPT, like, uh, because the good thing, it has a lot of information about me. So this makes it, uh, even when I ask for like, uh mm-hmm [00:46:00] some proofreading based on my. The way I write usually. Of course, I'm not native English speaker, so I need, I, I, it helps me in, in, you know, uh, fixing my writings.

So still good. But, uh, in terms of reasoning, and I've been interacting with it long time, I'm impressed by grog, honestly, I'm impressed by, um, um, also Claude, because Claude, especially if you have like a very, um, I would say complex task. It surprises me every time. Yeah. Uh, even, because I know people use it for coding, I use it for other use cases still, also very good.

So, but anyway, so moving forward, Matt, if you want to like, look into the future, you kept mentioning about the one person, $1 billion company and, you know, um, we discussed a lot. How far are we from that and, um, how. Like, how are we gonna make business in this case? Like if, are we going to be like one [00:47:00] person businesses interacting with each others, or like, is it the AI that gonna talk and we, you and me, we gonna be maybe on a nice speech in, uh, some tropical islands and the AI will, will, will handle the day-to-day operation, will report to us.

How do you envision the future? 

Matt: I think the.

So we, we, I, we and I had my previous co company. We, we were spending a lot of time predicting trends. We've analyzed reports. We had an AI system, in fact, that crunch like petabytes, it's like something we built for a client, but we had access to it that was used by BlackRock and other companies to predict mm-hmm.

Trends. And I write about that in my books as well, eventually. You know, when markets started shifting and we've started seeing the change that's coming out of AI in various ways, we concluded that we cannot predict the future. So my, so I'll just say that first, [00:48:00] that our current approach is not to only observe mega fundamental trends such as demographics or compute.

You know, the increase of compute, the double exponential as people call it. Around compute, but not really predict what, what, what will happen, but rather becoming highly adaptable to whatever happens because we have very large, um, sort of chaos in introducing factors right now. One is climate, which we don't like to talk about, but it's present and it's there.

It's a real thing. And, um, and, and, uh, you know, artificial intelligence, of course, technology as a whole. So, uh, you know, so between the two there, those, they're kind of like, you know, some people call them two super cycles. There is a lot of change, right? And the change cannot be predicted, I think, by anyone, basically, whoever, [00:49:00] whatever they claim.

You know, and, uh, and, but we can prepare and we can make some certain bets against that, right? And so the short answer is that I don't really know what the companies of the future will look like. But we can already observe that a lot of companies are sort of start getting stuck in a loop of like, we need to start, we don't start, we need to start.

We don't start. We need to. And that, and they're losing very, very, very precious time against those who are adopting AI and building it in. And we, for example, work with large real estate companies a lot. Mm-hmm. And we see gigantic difference. Gigantic difference in terms of what one can do versus another.

And that's in a very slow moving industry, you would think. But we see incredibly huge, uh, difference between like one company we've been working with for I think two and a half, three years now, helping them. They're so far ahead [00:50:00] of other companies we're talking to that they don't see how, they're not going to, you know, outcompete them on, on a very large scale.

And that's multi-billion dollar s.

I think this technology will really favor those who move and adapt, especially because of like this culture change. And, you know, learning from mistakes, just being able to, to, to, to track. And so, uh, you know, the important thing I think is to become highly adaptable and to become highly adaptable, to embrace something like the leap.

We literally created this framework for that. And, um, and be able to start moving and, and engage our teams in this like, ongoing sort of innovation of AI practice across the entire fabric of the company. I'm not so focused on the billion dollar startups. I think there will be some already this year, maybe.

Mm-hmm. But it's the valuation of startups, you know, it's all, it's all paper money, right? It is. The people we get excited for unicorns. But it's not real value. I mean, I [00:51:00] have many friends, including my own projects, that were worth, you know, millions, tens of millions, even billions of dollars. I have a dear friend whose company was worth $4.6 billion, and, and, and, and it got sold for parts, you know, when something went wrong in the market and then, and, and those things, those fortunes change very fast, right?

And so valuations are tricky, but I do think that this year we'll probably start seeing. You know, this, this, this, this actually come to fruition. And so, you know, like in any business, I think it's a, it's, it's a function of leverage. Like if you are able to big, build enough kind of orchestration layer around yourself using various AI systems and, you know, use AI to help manage ai, to help manage ai, I would say, uh, just in a simplified form.

Then you will get there. I mean, that you can just, you can get there. It's not a problem if, if, if you have a business that doesn't require you making deals in person. 'cause of course that's an impossible, but if it's a pro [00:52:00] digital product or something like that and you know, you can get there and I think that's gonna happen this year.

But I personally don't think that this will have a big impact on the market overall. I don't think those companies will outcompete, um, other companies. Uh, in that, in that space, you know, not to use dirty examples, but the company that has the highest revenue per employee right now, I think is only funds.

And porn hub still go, goes strong, right? It's not, you know, they're not necessarily going to, um, uh, to, to, to change the entire, um, the, the entire market in my opinion. But, um, and, and to that degree we'll still work with other people and other companies. There's an important aspect that's often overlooked as well, which is we like to work with people.

Right. And even Gen Z when you look at trends, they actually want to be in office now. [00:53:00] Right. And the millennials that I'm guilty of being, have, have been nomadic for or moving around the world for, you know, two decades now. And, uh, and, uh, and, uh, purely nomadic for three years just traveling around the world.

Right. And, uh. And, and, and I think people actually want to be with other people. Right. And when I started, uh, uh, future Works in the first month, it was just me and AI as my partner, and it felt lonely. I didn't enjoy this. Like, I didn't just like, okay, I'm talking to an system, you know? And, and so I think, um, we're not necessarily going to be just a bunch of people who are like solitary beings with.

I think that favors a certain kind of person that can put up with it that typically is quite angry with people as far as I know. Some people working on this and then not everyone is like that. Like few people are like that. And yeah. 

Mehmet: Super cool..

Great. . So thank you again, Matt, for, for [00:54:00] you know, this great episode, the great conversation today, and, uh, I really appreciate it. So this is for the audience, how I end my episodes.

Guys, if you just, uh, find out about this podcast by luck, thank you for passing by. I really appreciate it. I hope you enjoyed it. If you did, so please give me. Favor, subscribe and share it with your friends and colleagues. And if you're one of the people who keeps coming again and again, I really thank you very, very, very much.

You know, you are doing great things for the, for, for the show this year. You are like making us being selected on, on different platforms as one of the best podcasts. You're being, getting us ranked on the top 200 charts, uh, of the Apple Podcasts, uh, in different countries. Simultaneously. I really appreciate this.

This is not me. It's it's because of my guests and because of my also audience as well. So thank you very, very much, and as I say, always stay tuned for a new episode very soon. Thank you. Bye-bye.

[00:55:00]