Leveraging AI

23 | Achieve exponential scalability by becoming an AI first company with Adnan Boz, Stanford University AI business course instructor, and CEO of the AI Product Institute

• Isar Meitis, Adnan Boz • Season 1 • Episode 23

Share episode

Do you want to unlock exponential growth for your business without significantly increasing resources? This episode of Leveraging AI with Adnan Boz is a must-listen for business leaders and AI enthusiasts alike.

In this enlightening conversation, we deep dive into how businesses can leverage AI to enable exponential growth independent of resource requirements. We discuss the concept of a digital operating model and the fundamental shift it represents in traditional business structures.

🎙️ Topics we discussed:

  • 🌊 Creating a "Blue Ocean" by reimagining business models with AI
  • 🚀 Achieving exponential scalability using an AI-first strategy
  • 🎯 The role of AI in product management, development, and market fit
  • 📈 The transformative potential of AI beyond marketing
  • 🧠 The unique insights from Adnan Boz's experience with the AI Product Institute

Adnan Boz is the CEO of the AI Product Institute. With over three and a half years of experience in the field—well before the current AI craze took off—Adnan focuses on integrating AI into product management, product development, and achieving product-market fit. Connect with Adnan on LinkedIn to stay updated with his latest insights on AI in business.

About Leveraging AI

If you’ve enjoyed or benefited from some of the insights of this episode, leave us a five-star review on your favorite podcast platform, and let us know what you learned, found helpful, or liked most about this show!

Isar Meitis:

Hello and welcome to Leveraging ai. This is Isar Metis, your host, and this episode is going to be really exciting if you are in leadership position in any business, or if you are just enthusiastic about where AI can take businesses in the world. We are going to talk about how can businesses enable exponential growth that is independent of the amount of resources that it requires? This is done through creating a digital operating model for a business, and we're gonna talk about what that means and how to create the different layers in order to make that exponential scalability possible. What it means is that CEOs and business leaders must completely reimagine the entire business model and not just try to optimize using different AI tools that exist today. And we're gonna talk on how to do that as well. But if you do that, if you are able to reimagine the business with AI tools, with a digital operating model, you can create blue ocean and exponential scalability for your business, which will allow you to grow dramatically faster than your competition and dramatically faster than you are doing right now. At the end of the episode, like every week, I'm gonna share some exciting news that happened this week Let's dive in into creating exponential scalability in businesses, using an AI first strategy, Hello and welcome to Leveraging ai, the podcast that shares practical, ethical ways to leverage AI to improve efficiency, grow your business, and advance your career. This is Isar Meitis, your host, and AI can improve efficiency and get better results in almost every aspect of a business today, and for a good reason. A lot of the focus these days is going to marketing because of the literal magic, the generative AI can do for more or less every aspect of marketing. But limiting the use of AI just for marketing would be a big mistake because, You can benefit from using ai, like I said, in almost every aspect of the business, is just knowing how to apply it properly for that function in order to get better results. And our focus in today's episode is gonna be product management, product development, product market fit, and how you can use AI in those aspects in order to have the best product market fit in the least amount of time with the least amount of resources, which is how you make your company more successful. Our guest today is Adnan Boz. He is the CEO of the AI Product Institute, which as the name suggests, he focuses on exactly that topic, but different than a lot of quote unquote experts who popped up in the last, few months as the craze started with ChatGPT, he's been doing this for three and a half years. He's also the AI business course instructor at Stanford, so he understands both the practical side of this as an entrepreneur working in companies, as well as the theoretical side of this. Deeper than probably anybody I know. And hence, I'm really honored and excited to welcome Adan to the show. Adan, welcome to Leveraging ai.

Adnan Boz:

Thank you, Isar. Great to be here. Thank you.

Isar Meitis:

Adnan, I wanna take you back before we dive into the actual process of product management and how AI applies to it. You started your company three and a half years ago, what was the aha moment, or if you want the process that led you to understand that AI will transform the way product development works?

Adnan Boz:

Thank you. So that's a really interesting question and, correct question actually. A lot of people asked that, and it goes back to around 2010 when I first start getting into ai. And at that time I already had, I was already more than two decades in the software industry, and I came across these trainings from Andrew E in 2010, and I learned about AI and machine learning and I was like fascinated. But at that time I was working on high performance computing and use, utilizing GPU to actually provide solutions in the advertising industry. Then I started teaching at that time, so back in Florida I was. I was holding these meet ups at that time actually with NVDS help and teaching people how to utilize these GPUs and machine learning. And then later, when I start implementing all these solutions for the next, couple of years, I realized that the whole development life cycle of an AI product is completely different than the way I had been doing for two decades at that time. And, I was at Yahoo at that moment. I was a software architect. And then I moved to product management at Yahoo I, and then, I started seeing more and more patterns how to do AI product development different than software product development. And the main difference is that regular software products are deterministic, where you decide for the program, you plan it, you write the specs, then you program it, then you deploy it, right? But AI product development is different because the nature of machine learning models is not deterministic. It's probabilistic. So think about it, in the past you would just write, you would know the outcome, and the outcome would be a hundred percent guaranteed. The way you wrote, of course there are bugs, but also the use case, maybe you forgot, but still all the use cases you wrote and you will get the outcome. But it turns out that with AI projects, it's not like that. The outcomes are probabilistic. So in a probabilistic world, you start, running these projects different because it can take a probabilistic amount of time, right? You can come up with different probabilities of projects, use cases. So the whole project turns into some unknown results. And here is the, here's the problem. Data scientists are really good with probabilistic, models where, or probabilistic approaches where they really experiment more rather than defining more, right? Whereas software engineering, where I was coming from is really good at defining but really bad really not very good with Things that are not deterministic.

Isar Meitis:

And I'll add to that, business people at somebody who ran three different companies myself. Business people are not really good with probabilistic things. You want a plan and you wanna know what the outcome is going to be. Correct. It's not always the outcome, but at least you have a plan and you can follow the

Adnan Boz:

plan. Correct. Planning, And ambiguity. It kills actually the performance of these people. business side, software engineering side. So I started leaning toward more, towards more data science. So I started working on training models on my own and working with data science get into that area, even start a company, move to ai back in 2017 and 18 I think. During that time it became more and more clear when I started talking to a lot of companies that people really don't understand the nature and how the whole AI development life cycle works. What we call today ML ops. At that time there was no ML ops. We didn't call it anything. We just called it ai, the open lifecycle. Then in, in the last couple of years, the name AI Ops OLS started archit surfacing, but Basically explaining those to people became a job at the point, when I was with my own company. Then, at that time I started teaching. I said, okay, let make sure teach people this. what is what we call to the ML Ops. And I started teaching. I realized then, this is value. People really like it. Then I created a curriculum. I applied for Stanford's continuing studies, and they accepted it, because it was such a unique topic there, ai product management. And in fact, I was talking to my admin at Stanford, and he said at that time, He said, are you sure? This is such an important topic that we can even drill down to product management level. it's just still like unknown, right? And I said, let me show you. Then I went over and they were like, they really liked it since these are nothing new to companies like Yahoo, Google, Amazon and I worked at Yahoo in, in platform teams and deep embedded into these technologies and I seen that they have been using it for, at that time, for the last, last decade. So it was nothing new for large tech companies, but the rest of the world, outside of Silicon Valley E even in the Silicon Valley, smaller companies didn't really apply these techniques. So my goal was at that time to just tell it to everybody to get as many people know as possible. So then I also started in 2019 AI Product Institute to do it besides Stanford continuing studies, on my own. Of course, my corporate career continued, that wasn't overlapping. I worked at eBay Nvidia because it's, since it's training, it's not overlapping with their core business, proposition. And and I had more than thousand students. They're happy and getting, actually getting their hands on the AI product development and utilizing this information to deploy their own products.

Isar Meitis:

Fascinating. I wanna add one thing to that will make, explain maybe why this is so important. In my eyes, in the very near future, and definitely within a few years, there is not going to be a single software product that does not have AI functionality in it. Because if it will be, it will cease to exist because the competition will have these tools built in and hence, You will run outta business because the competition will be built, will be able to do things that you cannot do, or that your software or that your product cannot provide. So if I am within a, let's call it a traditional software company today, and I understand what I just said, that we have to start infusing AI functionality tools, backend capabilities, into the product that we're developing. What do I need to do? What are the steps that I need to follow in order to be best prepared and get the most successful outcome in that process?

Adnan Boz:

Yeah, that's a good question. A lot of people are asking it right now in the generate AI era, right? and the answer is actually twofold. Even maybe it's threefold. So first of all, We are in a disruptive time, so every market is being disrupted, right? that is one thing. last until last year, I would say until December, that wasn't the case until last December. In order to get into ai, you had to implement something named digital operating model. If you read the book from Professor Ian City competing in the age of ai, they actually did a lot of analysis on many companies and they figured out that the reason these large tech companies are large tech companies, that they're growing exponentially, is that they have something called digital operating model. So if you just keep out this generative AI disruption, if you wanna take your company into the field of AI and become AI first company, AI first, this is a term Google came out about five years ago. So they said, we are a AI first company. And if you wanna move to from mobile first, 10 years ago MO Mobile first was the thing. Now then, five years ago, the AI first started, then you have to implement a digital operating model. And professor, they're from Har Harvard Business School, they did this analysis and they figured out that a digital operating model gives you the exponential scalability. in, in economics, you, when you start scaling your company in the economics theory, you hit a point where you cannot, where scale doesn't make any sense anymore. those are disa economics of scale, because you have human in the loop, you have actually physical interaction there, and physical vault doesn't scale exponentially as digital world. and they looked at companies like Amazon, I don't know, Facebook to be these big companies and they have a digital operating model where they scale exponentially, digitally adding another user. Has zero marginal cost, right? For Amazon, right? a new per a new customer comes into Amazon. It doesn't cost them extra money like a customer comes to another company, right? And here's also the interesting part. They add to the value. So every user they add increases their capacity to serve other users, right? Because their machine learning models suddenly becomes more richer with extra users. So this, that gives them this. A cycle where they can exponentially grow their value to the customers as well as extract exponential value back from the customers. And so first thing, companies really have to jump onto Expon, jump onto that back bandwagon of digital operating model. So a digital operating model it used, be able see in the book looks like, and we had it at Yahoo actually even, 15 years ago. so you have this data layer, it's horizontal, and all the data goes into this data lake or data layer. And then on top you have this, API layer or some technology layer where you have this machine learning models and other things. And then there's again, a team is holding that. a team is delivering that. And then on top you have this application API level that provides all these different APIs. Then you have these different use cases from different teams. This could be vertical, especially in large, larger companies. You have these multiple product portfolio products and you hooked it up, although maybe there are different use cases, but here is the interesting part. Every piece of data getting into this digital operating model becomes suddenly a citizen of the whole company. And you can utilize it in finance, in marketing, in logistics, in your product development, in your engineering, wherever you wanna utilize this and. In non-digital operating model looks more like a small silos where finance team has their own data input. They just use it there. Nobody else can use it. And I don't know, marketing team has their own data source. They could, they generated and then they could use only there. so once you apply these to, to your company and become actual digital operated company, then you can start thinking maybe going into generic AI and utilizing it so you can actually utilize this exponentially growing, exponential scalability in that field as well. So this is the second section of my answer, but before I go there, no, this was,

Isar Meitis:

this is fantastic. I think it's a, it's really eye-opening and I have a very tough question to ask, which I'm sure doesn't have a simple answer. But I think people who are. Again, if you're a C E O today or a head of strategy, you gotta understand this. You gotta understand that without going in that direction, you're doomed. Then you might be doomed in a year, might be in three, but you're doomed. You have to go in that path because your competition will and then will run circles around you. But what you just mentioned is, okay, you gave it a name, but it's a very significant architectural change on the underlying layers of everything that you do. So if you're running like a traditional software company, like you're saying, you have a database for this, a database for us, a database for this. Each one of them serves an application layer directly, usually, or in most cases through an API layer. But there is no, that one layer of data that everything feeds into and one layer of machine learning that rides on top of that under everything else. What's the process companies have to go through in order to. Make that transition, is it even possible or do you have to rebuild on the side kind of thing? And eventually kill what you have today. and again, you've been doing this for a while, so I'm sure you have some answers. I dunno if people are gonna like them. I'm sure you have some answers.

Adnan Boz:

Yeah. This is great question. Out of companies asked it and before that to just step back and there are really three things you need to change in a company, especially in these disruptive times. And the first one is the infrastructure model. Yeah. You want to maybe develop these layers and create new actual technology there, which can serve all your products as well as your company and this infrastructure model. Then you have the operating model. And the operating model is the way you do business, right? The way you create your products, the way you bring your, the products to your customers, the way you actually, charge them and everything you do in the operating. Then the third one is the business model. And the business model is what value you provide to the customers. Now, this is your value proposition, and that's your sales and all that stuff right. Now between infrastructure model, operating model, and business model, the easiest thing to do is actually infrastructure model, because you start modernizing, you start optimizing your system. So if you have, let's say 50 databases, 50 data sources, let's say going into 50 databases, then you start, merging them slowly. You actually create bigger database. You don't have to do it at once because that will create a lot of out of disruption, out of problems internally. So you start merging them one by one. So you first merge two, three, and then go from there, and then you go to your machine learning layer where you have all these different smaller SDKs, APIs working internally, maybe a lot of miles. Then you'll start to figure out, Hey, how can I create standards here? So can I have a natural language processing N L P system that solves all our N L P problems across finance, all the way to legal, all the way maybe to, to marketing, right? Yeah. So one single N L P system that can solve everything. Let's say you hook up to open AI APIs. Yeah. but then you think, Hey, can we have a image recognition, system like image recognition solution that can solve all our image recognition problems? So you start gr putting this group of solutions together and eBay did it really good back in time when I think it was 2019 when I was at eBay. They split it in this way. So we had the whole NLP group, even at Yahoo, we had the whole NLP group was working on this 50 step NLPs at the time. but separating like this gives you actually this platformization layer where you have these different platforms and if you go to any. Any cloud source product like aws, Amazon, aws, Microsoft Azure, IBM Watson, or g ccp, you will see that they operate in a similar way. So you go there and you will see there's a whole section about language processing. There's another section about image processing. you create this layer. It could be really some cloud provider or you may be create internal. It doesn't really matter because that is more about buy versus build versus rent decision that you should provide to your company. The all these different layers. and then you create actually the API layer, and that goes about the infrastructure. So you can do the infrastructure slowly. This is fine. The more little more difficult part is changing the operating model. And when Ian City is talking about digital operating model, and of course there is flavor of infrastructure model where you have to do your homework. the bigger problem is that they're talking about is the operating model. Now, the way people think has to change to your culture is to change a little, because in a classical operated company, you must probably have been there as well. people, as we said, they create these deterministic solutions. even maybe marketing department doesn't really. Think exponential way they may think to write, I don't know, 10 blog posts or maybe do these ads to do the, those ads, right? Maybe 10 ads a week. So can you imagine doing 10,000 ads a week, different ads a week, or can you think of being able to write 10, thousand varieties of blog posts? So the mindset has to change a little towards exponential rather than the limitations of the alt vault where you think do, where you do things sequential, as well as maybe you have to introduce different aspects to operating model to, to remove certain bottlenecks. So people have to start thinking, Hey, what is the bottleneck in my process, where I cannot grow exponentially? And the simplest, for example, marketing right, is trying to figure out running some experiments. They're trying to figure out what is the best message to my customers. And we had the problem Beckett, eBay at fault and eBay it out of marketing at that time. And the problem was, okay, we are sending these emails, right? But which email subject is the best subject? And there's maybe billions of possibilities. You can really write that. because they were leaning towards AI a lot when the operating model was changing at that time, people start thinking, Hey, instead of the writing is email manually, how about we remove it poll and we turn it into an experiment, hook it up to AI that can generate these subjects, and based on some, heuristic, some logic of course. And then we run, let's say, I know 10, 20 experiments about this in, in, in a month. And then see which one is gonna be the best for that particular time in that context for that product. and then we go with that. So people have to start thinking in a different way. That is, I think, the biggest change in terms of operating mal.

Isar Meitis:

I, I agree a hundred percent. I wanna add one thing that connects to this very nicely to what you're saying right now. I say a lot and people have been listening to this podcast. I've heard this, but I say this on lectures that I give on stages and in the courses that I teach, removing from an era of improved process efficiency to an outcome, meaning if you think about everything we developed through human race until now was about how do I make the process better? How can I do more of this thing? And let's use the example you give. I need to write blog posts. So how can I write faster blog posts? Okay, so there's, as your research tools that make it faster, and there's templates that I can use to do things better. And there's, WordPress, plugins that allow me to put in the pictures without having to wear. there are ways to improve the process. And what machine learning and AI enables us to do is to get the outcome. I want 10,000 blog posts and you will have them. And especially now with the era we're moving into agents, and I'm not gonna dive to what it is if if you not look it up, but you'll be able to get the outcome you want at any scale you want at seconds or minutes or hours instead of months and years that it would take before and without the human overhead of, okay, if I want 10,000 blog posts, I need 5,000 writers to write those 10,000, all of that goes away. So the mindset, going back to what you said, the mindset of needs to shift from, how can I manage a better process to what outcome am I looking for on every aspect? Whether it's the business aspect, the operating aspect, the tech, what's the outcome I'm looking for And find the solution for that. Because right now, that solution is either exists or is at reach when it wasn't three years ago.

Adnan Boz:

Correct. And you have a really great point there. And you know how it started and our head is still there. we had industrialization where we turned everything, industrialized, everything, let's say car manufacturing or any other agriculture. And then from there we moved into globalization. And with globalization, we got all these frameworks like Six Sigma, right? Yeah. Try to improve efficiency. and Japan, is really good at it. So they could really reduce the marginal cost of everything, of production of, in, of any industry. actually down to, down to the very, very low margins now. But that changed now I think, and. Not just, I think, but I read that they hit you hit a limit after a while. Yeah. It doesn't scale anymore. Let's say you have a car manufacturing factory and or a plant where you maybe reduce everything to the bare minimum with using six Sigma or something like that. You really brought it to a point, but when you keep doing that, keep trying to optimize it as you said. Now, if you just still think that logic, how can I really optimize these one by one, then there is a limit you will hit. You can never reach that point of exponential growth. So we have to put that hat down. We have to reach to remove that idea and start thinking outcome based as you said. Okay. I wanna really assume that I don't know anything about how to do it, the how. I just want, I just know what I want at the end, right? Yeah. And what I want. So by giving this what to AI algorithms and it can figure out the how part and that is the main logical change and extra maybe cultural change for a lot of companies to, to adopt. For sure.

Isar Meitis:

so you started talking about the problems, the operation side. What about the business side? So what's, what has to happen on the business side to support this process? Or what are the limitations or the difficulties that you see companies going through?

Adnan Boz:

Yeah. Business side. That is the golden question. Now, Jeffrey Moore says that no established enterprise can easily change their business model. If you think about business model is your company. It is, of course operations and infrastructure and all that thing in it, but that is the value proposition to your customers. So everything starts with your business model, and then you align everything, everybody. You even hire people based on that business model, right? And now think about it, your whole business is being disrupted. Take for example, search the whole, web search business. So the web search has its own ecosystem. There are advertisers, a SEO companies like think about it. All those people create websites, put it up there. Now think that search business is being disrupted and turn into some chat conversation where you just ask questions and get answers. So don't, going

Isar Meitis:

back to our thing, I don't care about searching. I care about knowing. exactly. I just wanna go to the outcome.

Adnan Boz:

exactly. Why do I then, why do you even have seo Then? Why do you have all that ecosystem out? It's very similar to moving from CDs to streaming CD times. You had actual CD cleaners. CD recs, CD drives, a CD holder cd. the CD ecosystem around CDs was bigger than CD itself, and now it's the same thing. The search ecosystem is actually bigger than the search itself now. But think about it, it's now dying, and you're moving to a new business model where people just wanna talk to somebody to get information. They sometimes, they don't wanna get information, they just wanna say, Hey, I'm, I wanna book, I wanna actually write a book and just help me write it. Yeah. I'm not searching actually for topics, eh. More often than not, we actually try to go use the search to accomplish steps in our process to get to the outcome. Now you can just tell the outcome and it'll give you the results. Yeah. and that is a business model change. Yeah. And in business model disruption like this, like generate ai, the companies, companies really apply three or four steps. So Jeffrey Moore has a book named The Zone to Win. And In Zone to Win, he explains this much more in detail. so he devis every enterprise into four. And that's coming back, that's coming from this Three Horizons framework, from McKinsey. you have this horizon one to three. Horizon one is your main business, core business, horizon two, horizon three is your research going forward. Then Horizon two is actually intermediate place where you transform your business, transform this research into your core business, and he divides into four because. Usually the productivity organization in a company who that will develops this product is different than the performance organization that sells the product and provides the value. Yeah. Now in this Four Horizons framework, he explains that in order to change actually your business model, you have to go through these, through this loop where first you incubate something. Incubator idea, and you have to ha you have to incubate for a while. It's not just one day, right? We are talking about, about maybe three to four years and the way Microsoft invested five years ago in OpenAI. That's incubation. And, the way NVDIA invested, about, I dunno, seven to eight years ago, or maybe 10 years ago, into AV autonomous vehicle and the way Google invests into certain things the way meta actually invested 10 years ago into vr. now these are all incubations. You try to grow in your company. Then once it hits, Size, some significant size. It's usually maybe 0.1% of the overall revenue. Let's say if your revenue is a$100B so we are talking about still$100M revenue. It's not small for a lot of companies, but you have to grow your incubation to that point. Then he says that you have to transform it into your core business, and that is where all companies fail. He list like 50, 60 companies who failed from. And there are many names I don't wanna name them now, but, so the main problem is that when a company thinks of, incubation, if they're not experienced, what they do, they steal some resources from their current core business. their current core business is a revenue machine that is generating, we reset hundred billion dollars. So how can you. Steal from it. While your investors are asking from your 10% year over year revenue, how can you take out 1000 people from there? If you have 10,000, 10% of the people and still expect the company to generate that amount of revenue, you cannot. And here's the interesting part. Even if you steal those people, most proud days, people cannot work on your incubation new business because they were hired to work on the current operating model, the way they do the business. So when you create this new incubation to change actually your business model, you need new people. You need actually a trailblazers. You need people who are entrepreneurial. You need people who want to jump on topics, create take high risks. Whereas the other people who are in your current business, core business, they don't take risks because you deliver. Year over year to your, pro promises. So you hire this new team. That is why a lot of companies have these groups separated physically, even building wise in another location. And then you incubate it, then you start transforming into a business line item as Jeffrey Moore says. And then I wanna

Isar Meitis:

pause you a second and ask you a very interesting question. Sure. Because I never worked at a huge enterprise, like the largest company I worked for, which was pretty big, was like a 7 billion, dollar company. And we had, I don't know, 10,000 employees around the world. So not a small company, but not a huge company. But most of the companies I worked for were startups, and we had between 30 to a hundred people. and the company I worked with in between was a travel company. We had less than a thousand people, so I had 800 people. It's a lot harder than to do this thing. So if I'm now running a tech startup that is doing okay, like I have good margins. I'm selling, I'm growing, but I'm looking two years ahead and I understand what you and I are talking about, that I gotta infuse AI into this thing and I gotta go through the process that you very clearly described. But I don't have the manpower, the bandwidth, the resources to now incubate five different ideas. I maybe can bet on one, maybe. How do companies do that? what's the right process to figure out? The most likely thing to, you gotta make a bet because if that bet fails, it might be the last bet you make in that business. Correct? Correct.

Adnan Boz:

Yeah. That this is a question a lot of companies ask. That's a really good question. And the answer I've found by working with lot of companies is a strategy called Blue Ocean Strategy. And these two professors, they came up with this logic, I think 10 years ago, and Tab Book is so famous, actually, it's a bestseller. they have their third book now. and it explains that, In order to grow into actually a new business, you have to find the blue ocean. Yes. The way companies really operate when you go through the product open life cycle, and when the actual technology goes through technology, adoption, life cycle, when the categories mature, everybody is fluctuating around the same idea, same offering. let's take for example, e-commerce. go to any e-commerce website today. They're all the same. they're maybe 10% different up and down. They have all recommendation systems. They have research, then they have these listings. They're all the same. Now, how do you really create something more valuable to your customers that will actually separate you from your competition? Especially with ai, you employ their framework. So they're really great frameworks. Active by utility map, four actions, framework, strategy, canvas, that's the last one you actually utilize. So utilize those tools to identify a blue ocean. That means a place for you that is different than the current competition. And remove the, remove the. product features and other things that competition forces, but brings value to the customers, unprecedented value to the customer. So that y your customer, those customers will come to you and they will also have high barrier to entry because you change the business model so much, they cannot change the overnight. So if you apply Blue Ocean Strategy, that will allow you to find the niche idea that will be, that will keep you apart from the competition, as well as add unprecedented value to your customers. So it's the

Isar Meitis:

first, I love what you're saying a lot because it reverses the sequencing. We, we spoke about things. We spoke about, technological infrastructure, and on top of that, the operating infrastructure. And on top of that, the business infrastructure that talks to the customer and the real thing here and what you're saying. Is if you wanna do this and you wanna do this in an effective way, especially as a smaller business, the trick is to really understand the needs of the customer and the value that you can bring in a unique way that nobody else is doing right now. And then reverse the process, then say, okay, in order to do that, our business model need to be this, which means our operating model needs to be this, which means our technology needs to support it. So the starting point of this process, and I agree with you a hundred percent, the smaller you are, the more critical this becomes. Because if you Google, you can make a hundred bets on any given point. And they do, that's why they change it to Alphabet. They have all these other businesses because they knew that search one day will die. And that drives, I don't know today, 80% of the revenue, probably more. And so they have eggs in all these different baskets. But if you're a smaller company, you probably have enough resources for one of those bets. And what you're saying, Is the key to all of this is to really understand the need. What can you solve? How can you provide very unique value and invest everything into that? Where now it'll be harder for people to catch up because now you have a new category of business that did not exist before in which you are the only solution, at least in the beginning.

Adnan Boz:

Correct? Correct. said. And here's the trick, right? and it says maybe not so common sense. And also people try to utilize common sense and they think Blue Ocean strategy is just really understanding the user. And I can go talk to user, but I see it again and again. I and I teach it in my classes as well. We do workshops on Blue Ocean. now the biggest impact of Blue Ocean is to have you step back. Step back, look at the whole end-to-end life cycle. So when we work, and that happens actually to, out of companies who are getting new into the ai, what they do, and they, because they live in this closed loop area, closed loop solution, they cannot think outside of that box. So for instance, if you're doing e-commerce and you think, okay, what can I do with AI in e-commerce? Let's say you wanna apply AI there, and then you start thinking, oh, how about I add a recommendation engine? Or how about I make the search better? Maybe I should have a chatbot, right? All in the box. All in the box. You're thinking really from the problem. Okay. Blue Ocean suggested, okay, step back. Let's see. How does this user even start thinking to buy something? Let's go back and let's go back to the point where they actually sit in the porch with their neighbors and they see a barbecue in their neighbor's yard. Start there. All the way there. Then they see it and they create the need. And then they actually create the want. Then based on their income, they create the demand and they go to the website, they search. Maybe that's not the only thing they do, but they want to really buy this barbecue because they sold it in their, neighbor's, backyard. And developing this understanding of that, everything starts with a need, then turns into want, then turns into a demand based on your resources. It is really important. it's really important. So when you go back to the need and figure that out, how the even started, then you start coming up with very unique solutions. And maybe instead of creating a recommendation engine, maybe you're gonna, slap a smart, sticker on, on the street. Could be even that I like, or maybe you put a sticker on the barbecue devices on their neighbors so they can see the brand and shop directly. Instead of going extra to website to search. Like you can come up really with interesting solutions, which no competitor thinks because competitors are locked in that box. oh, how can I improve optimize my website, my e-commerce website? That's all they think all day long. And with ai, that comes even more important because what AI brings is far beyond than just optimizing what you have right now. It requires you to have a new way of thinking sometimes, like for example, Google search it, let's assume you are actually Google search and you are, the VP there and you're thinking day in, day out in the 2000, how can I optimize search you? Maybe make it faster, more beautiful, maybe more different colors. That's all you think all day, right? And the algorithm gets maybe better but you never really think. Hey, can I use it here in a natural language processing a large language model to turn it into a chat? you won't think that because it's not in the realm of your solution at that moment. So in order to think about that, you have to really step back and ask yourself, Hey, how did the user even think about going to Google? So what was, why did they go there? So you think, okay, oh, they're at home, they're friends at home and they wanna cook something, and the guy is thinking to, maybe order something or cook something and then they go Google for something. Or maybe they're curious about a movie, they just watch the movie. And then they go Google upload it. So if you go back to the point where their need starts, then maybe you can even just remove the whole search and say, okay, the guy is just curious about a movie. Why don't I put actual icon on the movie on Apple tv? They can just click there and get the information right. You suddenly eliminated search. But here's the thing, if you are the VP of Google search, you would now think to eliminate your whole organization. yeah. that, that actually is the Archus heel of large organizations. So in a, in an foreign, an organization to change large organization like Google, or we are talking about 10 thousands of people to change their business model, they have to think to kill themself first. and nobody I know would, would actually think, Hey, can I just close my department fire all, 1000 people, that's not part of their goal, right? They won't think that. So that is why you need product managers. You need CEOs who can see the bigger picture and think of even changing the whole organization to a new organization. and you need that higher level view. And only CEOs can do that, or sometimes product organizations. In some companies, the product organization, the, it's connected to the, VP products and C P o Chief product officers separate from the technology org in other places, and that gives them the power to make those changes. But in some other organizations, the product organizations connected under. I don't know, tech CTO maybe, or maybe marketing somewhere, which is irrelevant, but they don't have that power. So at that moment, the C has to make that, decision. So it goes back to the organization and how you also structure, as you can imagine. I think

Isar Meitis:

these are really valuable points. I think they are, especially now with the level of disruption we are going to see because of ai, it becomes even more critical. And like you're saying, the understanding that the business model that you have today may not exist. All of it. Correct. I'll give you a few examples. I, I'm consulting to different companies on how to approach ai and I had on conversation with the CEO of a very large, legal company, one of the biggest in the nation, and I told him, did you think about the fact you may not be able to charge per hour, at least not for paralegal work moving forward, which is, I don't know, 50% of your income, sometimes more. And he is no. I'm like, okay. Think about it, all paralegal work will disappear. Or will be minimized to minutes instead of have a team of five people working on your case, researching, collecting information, putting it together, analyzing it, and they're gonna work for three months. So that's gonna be$1.2 million worth of paralegal hours. that all goes to, six minutes of an AI machine that can to do all of that. So that 1.2 million is gone. what do you do? And so this is the paradigm shift that, that, like you saying, people at the helm, CEOs leadership teams have, that's the way they have to think this way. And there's a lot of other examples, not necessarily directly ai, AI related, electric cars. So it's, I think it's obvious to everybody that internal combustion cars are gonna be a luxury that people who buy Ferrari may be able to afford and everybody else will move to. To electric cars, and yet you don't see Ford saying, okay, I'm gonna stop making all the cars I'm making today and start making all the electric cars because that keeps Ford running. So they have to find this interesting balance between how much do we cut the branch we're sitting on by selling something that right now we don't know how to make it at least not profitably and eating away our own market share and cake by giving away, the stuff that we know how to sell today. So I think we're live in this very unique moment in history where things like events that happen every now and then, like Kodak disappearing. Kodak was a global monopoly that very few existed. As I own 80% of a global market in something. That disappeared to nothing because they were not able to go through that process. And I think we're gonna see a lot of that happening in the next few years of companies valuations being spread to other competitors that have been around forever. Do you agree?

Adnan Boz:

Yeah, definitely. And I suggest every company to, to think that what if the marginal cost of the product I'm providing or service I'm providing is zero? What will happen in the market? So what will put the new equilibrium of supply and demand? That this is the question they have to ask themselves. Because once the marginal cost is zero, that means you, you will have all kind of competition. Even, I don't know, a two man team, company will be your competitor. And in fact, that is happening right now. For example, Grammarly, right? It is a great company. I'm a Grammarly user. So you, grammar is an assistant writing assistant, and the day open AI dropped chat, G p t, it literally just killed at that moment the business because now you have actually more intelligent technology behind, and they caught up actually. they're pretty good at catching up with these technologies. they're still in good shape. just that, think about if you would be a company, Some other company who doesn't think what's gonna happen. And suddenly the marginal cost of assistant, writing assistance becomes 0 cent per 1000 tokens. That is the price of, open AI a p i and whereas, Grammarly charged, I don't know how much I'm paying, like maybe 15,$20. for, I don't even write maybe, I don't know, a couple thousand tokens. So suddenly it is like we can. Virtually assume that it is zero already. So I want every CEO to think whatever they're producing, whatever value they're providing, that the marginal cost of what they're doing can go to zero. Of course, if they're producing physical products, there is always some e extra cost. But if, when things digitize, like Peter Iman, this also talks about this digitization in the sixties. when things dig, digitize, they become exponential, but first they don't become exponential. There are a lot of solutions. They first actually charge.$10,000, then they charge starting, a thousand dollars, then it drops to$100. then it becomes a service. You pay only$10 a month, but we are in a time, even that$10 a month is turning into$0 right now, the marginal cost. So I want every co to ask themselves, okay, what if. The marginal cost will drop to zero. How am I gonna earn money with all that competitors in the same market? how will the new EUM of supply management look like? And that is where Blue Ocean strategy is really good. Blue ocean strategy will take you out of that competition where marginal cost is zero and people don't pay money anything anymore for the product and take you into a blue ocean they call it, where there is no competition because you just innovated. And you can still write that wave innovation for a while, let's say another five years. Of course after five years when that becomes zero marginal. Cause you again have to use ation strategy to jump to another way.

Isar Meitis:

Aan, I think these last two points are, Really amazing, and I think it's a great point to stop because you and I think can go back and forth on this for a very long time. But I think in general, if I wanna summarize the stuff we talked about, because I think we touched on many different points. One is the understanding that the future is very different and the present and it's happening very fast. So if we were used to, okay, the next business cycle that I gotta adapt to, the next innovation is in 10, 15 years, we're talking months. And that means you gotta react very quickly and you gotta build that muscle of being able to continuously reacting very quickly because this will just accelerate. And then like you said, what are the tough questions you gotta ask yourselves And what are the major changes you gotta make in your business on any of the layers that you mentioned that you gotta take action on right now in order not to put your business at risk? Adani, if people wanna follow you, learn more from you, connect with you, work with you, what's

Adnan Boz:

the best way to do that? Yeah, great summary. First of all, that is, there is even a four step framework how to act right now. I put it on my LinkedIn as well. I gave a training at, when I was still at Nvidia, about two half hour training to executives how to do what to do from now on, right? And ba basically, as you explained, you first have to be present. Didn't know these things are happening. You have accepted. Second, you apply a neutralization strategy where you fix your operating model in a couple of months timeframe, not quarters. then you actually go back, fix your infrastructure model and that is the optimization. And the last step is, doing actual differentiation, fix your business model. So those are the four steps you go through. yeah, you can find all information on my LinkedIn ad nabo or on Twitter ad nabo or ai product institute.com and you can download resources. I also write about these things on my block post. You can follow me there as well.

Isar Meitis:

Adan, thank you so much. This was really a incredibly valuable and b, really fascinating conversation. I appreciate the time you're sharing and the knowledge

Adnan Boz:

of sharing. Thank you, sir. Thank, thanks for having me. I really appreciate the time you spent here. Thank you.

Isar Meitis:

What an amazing conversation with Adnan. This is not like most of the episodes that I've shared so far that are very practical. This has profound implications on the future success of a business. And as I mentioned in the beginning, if you're in a leadership position, this has to be the future mindset of any person leadership position in the era that we're moving into. Now let's move to some exciting news in the AI world that happened in this past week. First of all, stability ai. The company behind Stable Diffusion announced their next version of the model. They call it Stable Diffusion XL 1.0 different than the previous one. That was stable diffusion XL 0.9, and it provides higher resolution, more accuracy, better colors, and overall better results. Than the previous model. It also supports in painting, which is reconstructing missing parts of an existing image and out painting, which is expanding an image that currently exists, and even image to image prompts, meaning you can upload an image and use it as a baseline to then go and make variations to that image. All of that is available in the new model I. Must admit that personally, I didn't get a chance to play with it and compare it to the current pack leader, which is Mid Journey. But from what I've seen, it's definitely a big improvement to the previous stable diffusion model. The biggest benefit over Mid Journey is that stable diffusion is it's completely free and you can get access to it either through their API or through their Dream Studio website. Another interesting news from the AI world that is not necessarily generative ai, but it's definitely really interesting to where this world is going, is that Google's DeepMind has unveiled what they call me, Palm M. It is a demonstration of a generalist multimodal biomedical AI system, which means it can get inputs and communicate across various aspects, meaning text, images, genomics, and. In other data as part of a single model architecture, it's geared specifically towards the medical world, and it's an amazing first step in the direction of having significantly better medical capabilities, while making it cheaper and faster and more accessible to the masses. And when we think about the dream future of ai, one of the biggest promises is obviously solving some big human problems, such as various medical issues like cancer and so on. Very interesting news in that aspect from Google's DeepMind Next piece of news comes from Stack Overflow. It's a company that, if you don't know, is probably the largest community of computer programmers and developers in the world. It's basically a huge community where people can come and ask for advice and get advice from other developers. It's been around for a long while, and it's definitely the largest and most active community of developers, and they just announced. Overflow ai. So they are taking the data of 58 million questions and answers that exist in the community and making it accessible through a chat interface. The cool thing about it beyond obviously, Fact that instead of searching through an endless number of posts and hoping to find the right ones, you can ask a question, get a summarized answer from all the previous content that exists with citations to where their sources are. You can ask follow-up questions. Then you can even have it help you draft additional follow-up questions to the community in order to get answers. Overall, it sounds like an amazing tool by Stack Overflow. and this obviously follows what we talked about many times in the show before. If you have a large amount of proprietary data, adding AI and L L M on top of that is gonna make your service or your product even more powerful than it was before. And this is just another example of that. From Stack Overflow to Amazon. Amazon is obviously one of the giants in the game. A while back, Amazon announced Amazon Bedrock. It's an infrastructure of artificial intelligence. I. That runs on top of AWS. So if you're deploying your software on Amazon's AWS servers, you can use Amazon Bedrock that basically brings the large language models infrastructure as a native capability into anything that runs on AWS and what they've announced this week that they are releasing agents for Amazon Bedrock. Those of you who don't know what agents are, it's an expansion of a large language model. It allows you to define a task that then the agents will create. Multiple sub steps for it, and will prompt itself and can take multiple different actions across things it has access to in order to complete the task. So this capability is now available natively for people who develop and deploy their solutions on Amazon's AWS and that are using Amazon Bedrock access to the different models. This enables close to magical capabilities to anybody who's using this platform. And I don't see Google or Microsoft staying behind, so I expect to see them releasing similar capabilities to their hosting platforms as well. And the last piece of news comes from Netflix and the backlash to a position that they opened. Netflix opened a position for an AI focused product manager, and the offered salary ranges between 300,000 to$900,000 a year for that role. The goal of that role is to increase the leverage of Netflix machine learning program. Now, while obviously if you're Netflix and you find ways to, as they said, increase the leverage of the machine learning program, it's a great return on investment to invest$900,000 a year in that program. cause the returns are gonna be, Significantly higher. That being said, there's a really bad taste in the community for a offering, a salary like that for a single person and especially in a time where a lot of people got released from large tech companies in a time where, Actors and writers are on strike because of changes that are happening in the industry that Netflix is driving. So you have a large amount of people who are afraid to lose their jobs, and at the same time, one of the companies that has pushed them into that situation is offering positions at$900,000. If I connect this back to the topic of this amazing episode with Adnan, you can understand the connection if you understand how to build a digital first and AI first company, and you learn how to leverage that. You can reach exponential scalability and you can make significantly more money without investing significantly more resources. And$900,000 in the scale of Netflix is negligible. And so I expect to see more of that, meaning companies who are willing to invest in the right people, in the right technology in order to capitalize on this amazing transformation that's happening right now and come up ahead faster and better than anybody in their field. Before you go, I have two small requests. One is, if you haven't done this so far, please write a review for leveraging AI on your favorite app, whether it's on Apple Podcasts or on Spotify. That helps more people find this podcast. And on that topic, the second request, if you know people that can benefit from this podcast and you know they can learn from it, please share it with them. I appreciate that very, very much. That's it for this week. Go and explore ai, play with it. Try different things, try different tools, and share what you learn with me and with the world. And until next time, have an amazing week.

People on this episode

Head Shot

Isar Meitis

Host