PurePerformance - The Pragmatic Approach to Becoming AI-Native with Pini Reznik

Episode Date: November 24, 2025

There is only one successful way to adopt new technology, and that is transformational! Sounds like a high-level consulting pitch but our industry has a track record to validate this statement. Just look at the recent web or cloud-native transformations!Pini Reznik has been helping organizations along the current AI-Native transformational journey. And what a timing: He just published his book on From Cloud Native to AI-Native where he provides a pragmatic approach to leveraging AI from Pioneering to Gradually Scaling!Tune in and hear from Pini why he thinks that AI projects are not failing because of bad AI, but because they approaching the problem the old and wrong way!And, stay until the end to hear how it was to write a book about AI using AI!Links we discussedPini's LinkedIn: https://www.linkedin.com/in/pinireznik/Link to Book: https://re-cinq.com/bookOur previous episode: https://www.spreaker.com/episode/ai-native-the-next-revolution-after-cloud-native-with-pini-reznik--67692567Prompt Engineering Conference Talk: https://www.youtube.com/watch?v=W7z5XMnvYt8

It's time for Pure Performance. Get your stopwatches ready. It's time for Pure Performance with Andy Grabner and Brian Wilson. Hello everybody and welcome to another episode of Pure Performance. My name is Brian Wilson and as always we have my co-host Andy Gravner

who's not making fun of me this time and this is our second of two back-to-back episodes we're recording today. Hi Andy.

Hi, how do you know that I'm not an AI bot responding to you now? Well, because an AI bot would probably be funnier.

Well, that's a good point yeah. Thank you so much for the compliment here. No, because we've been having you know,

the last episode was around the developer experience and the, in the era of AI and the different use cases of using AI. And now we have actually a repeat guest back. And we also talk about AI. Yes. Awesome. Yeah.

So without further ado, because I, unfortunately, I'm the real Andy and not the AI, Andy, that is funny. So I only have so much humor in me. And therefore, I would rather invite our guest, Pini Resnick. Hi, how are you? Welcome back. I'm good. And I guess how do you know that AI is not pretending to pretend being not funny, right? I don't know if it's advanced enough to do that, you know.

Right. So it's advanced enough to pretend to be, Andy, but not advanced enough to pretend to pretend being unfunny. Yes. Yes, yes. Yeah, I think we can comfortably say that at this point. It reminds me a little bit of the movie inception now, right? because you're having

a little bit. Thanks for invitation again and it's great to be here. Let's see what kind of interesting things we're going to talk about today. Yeah, I mean, Pini, I got to say last time

when we had you on the show and that was not too long ago. It was September 15th, folks, if you want to listen to that episode, the details and the link is in the description. It was called AI Native,

the next revolution of the cloud native with Pini. And I really, I've been using one of your quotes several times now since then. I'll tell you what I use. Because back then, you said right now in 2025, it feels like defining AI Native is like defining what Cloud Native was in the early days of Cloud Native when Kubernetes just came out. And everybody was basically just trying to, in the beginning, package their applications in a container deployed on Kubernetes and they thought that's Cloud Native.

obviously back then the ecosystem wasn't there many of the patterns the architectural patterns were not there the cloud services were not there that we have today so 10 years later we have a completely different understanding of what cloud native is as we had 10 years ago back to AI native we are still in the early stages but you wrote a book about AI native that is now being published and I think you have it in your hands somewhere at least it's on the yeah it's published with Michael Mueller who is co-founded and rethink to. So it's a, and it's, I think, a story that we are writing in the way for last 10 or 15 years or since the beginning of the career, which is, I guess, 25 years. Yeah. And so what is it, what is it in the book now that people need, that should people draw to actually go to Amazon or wherever they can buy your book to buy the book? So you can buy it on Amazon, and it's called from cloud native to AI native, catching the next wave of innovation.

So the general idea is that the innovation is not, it's not unique to AI or AI native. It comes in waves, and all the waves are different, but there is consistent behavior, how to adopt technology. And there are many, many models around that, innovator's dilemma, product life cycle, curve. and crossing the cars and variety of different models. They all generally talk about the same thing. There is the right way to adopt new technology, which is transformational. And transformational means the architecture,

all the team structure changed so dramatically or drastically that it will affect the other one. Due to Conway's law, the technical architecture and organizational structure should reflect each other. So when we move to CloudNay, we moved to microservices, which led to creation of new team structure, described in team topologies in the best way. But when we started 10 years ago going and doing containers, the microservices concept wasn't

clearly defined. The team topologies were not clearly defined. So we had to go through these difficulties of first using Kubernetes and containers, then trying to put smaller pieces in it, then adjusting our team structure. So by the end of it, we had a very clear idea what Coordnative is. And what's happening now with the AI Native is sort of similar, like you say. We're figuring out what's the right architecture, what's the correct team structure, and that's what we're trying to explain in a book, basically.

What is AI Native as well as we can describe it today, which is not very well yet? But more than anything, we are focusing on transformation itself. how you can actually start small experiment and then gradually increase investment and then gradually transition from old to new that is actually not very new concept so if you want to learn how to adapt new technologies the best way to do it is hopefully to buy our book

and we also have a bunch of patterns cards and stuff and we're doing a bunch of like this so the pattern cards are sort of the patterns that we identified over the years that help to actually go through transformation using the methods that are actually working well hey brian this just reminds me about just an hour ago when we had the conversation with laura tacco from dx because a penny for you you may know a lot of may know her. And we talked about developer experience.

We talked also about the role of AI. And we talked about that we should not make the mistake as an industry to think about AI in the context of developer experience just about, you know, generating code because they would limit the AI to just a very small part. She then actually said, we need to think broader. There must be a transformation where we use AI natively throughout the software delivery process giving designers AI tools so that they can not only create a nice mockup on on in a PowerPoint or in figma but also in you know a working prototype that is generated by the AI also doing

the first feedback loop on that prototype to then figure out is this really what we need is this the transformational thing that you're talking about i think yes but indirectly so i think this is exactly what's going to happen that I think the way we explain it in the book is there is a reversal of time consumption during the development process. If we think about

how it was done until today you have a bit of time you're spending on thinking about a thing right. So let's say 10% 15% you're solving a conceptual problem in your head. Maybe it was a bit of experimentation. And then you spend 90% of your

time actually coding, debugging, deploying to production. So really not very like cognitively difficult tasks. I mean, there's still cognitive tasks, right? But the point is that there is no much time that you are spending

on thinking and solving problems. And most of the time is basically telling the computer how I actually do it. What happens in AI eventually, maybe not right now, but gradually we're going that way, is that there is reversal of time. So basically,

you think for 10% of the time, so basically the same amount of time, let's say, five hours. You have a solution, then you tell the solution to the LEM or to AI system, and you get an answer within a minute. And the same would happen with designers and with others,

architects and other disciplines. So they think, they solve the problem, and then the AI can do the work very, very, very, quickly and then you need to continue thinking on the next problem. And that is transformational because that means that people are not going back to their desks working for a week and come back with the results, right?

But they need to work together in much more cross-functional teams that are not just covering dev and ops, but also designers, also architects, potentially all together in the same team solving difficult problems and then asking AI to actually execute on that. And I think that is transformational because that entirely changes the dynamics of work and organization. I wouldn't say we can do it already today, practically speaking, but that's the direction we're going to. Would that, you gather all the information, you feed you.

it to the AI now in your in your imagination of this is this we have the AI just go ahead and execute it or would you be also thinking of simulation phases

or is it just a matter of go ahead create what we talked about now let's run it through a ringer of tests and all that kind of stuff or do we let the AI do some sort of simulation of all that before it even comes back to us? I don't, frankly, I don't know.

But I can guess it's a combination of all of that. I can give you a real life, for example. Our co-founder, Christian, he's CFO. He's not a developer. He never did any development in no way related to, except working in technology companies for many years. He never did any actual work in development. So he decided to create a CFO.

agent, which is an agent that connects to the data in the company to all kind of internal, the variety of internal systems, and then can answer questions that typically CEO, board, or other people in the company asking CFO, like, what's revenue last month, what projections, what kind of, are we billing enough? All kind of questions the CFO would typically be responsible answering, but it would take CFO a lot of time to go and talk to five different people. that will look into 10 different tools and then they will combine it in some Excel sheet and then do something. So just today he was showing us the CFO agent that he built in about 20 hours with the front end,

which connects to, it has a SQL database for no good reason. It has all kind of internal things and it has a very nice dashboard and it asks all kind of questions. connects to the right systems in 20 hours without actually knowing development. So he used code code and then he took a bit of chat GPT for a deeper research and put it back in cloud code. So there is a bit of dynamic around that. But I think this is the achievement is very impressive. Now, it's not production ready in a way because he doesn't even, like he doesn't even know

how to ask production-ready questions. So this is something that a non-developer can do in days already. So what's going to happen in the next year or two? I got a question. I want to play a little bit devil's advocate here. So the scenario that they just explained, so as CFO, CFO, right? That's what he is. Yes.

He's now able to create. a software service or an application that basically solves a problem for him, right? So I think that's a great thing because he can then explain, this is really what I need to make my job easier. But the devil's advocate here is, if I think about 10,000 CFOs in the world,

would it make sense for 10,000 CFOs to do the same thing and basically build 10,000 times a similar solution because they all need to solve a certain problem? or wouldn't this be much better solved by a standard solution that doesn't necessarily need AI? It's basically just data gathering from five different systems and coming up with five charts.

I can answer it in two different ways. The answer is as usual, both is correct, right? But I can explain what it means. So first, if I go to your house and you go to mine, you will find two different houses. Why is that? Like, we can live in the same house. It will be cheaper and easier to build, right?

We don't want that, right? We want uniqueness. We're buying different clothes, especially now, but obviously the mass production reduce the cost of production of goods and services, and that allows us to become consumers, right, and leave the lives we live now. But we still now want to go to the next stage and sort of be unique in every way.

We want a unique car, a unique t-shirt. Like I have a Bat Katz Club, which is fairly unique. And so there is an element of that. We actually want different things. We're different people and we can achieve it. The reason we couldn't achieve it before, because we had to use mass production to generate enough of the same thing. It was impossible to create unique things.

The other side is that there is this sort of pyramid of three, year layers in every organization there is a basic functionality then there is industry specific functionality and unique business sort of secrets also or like a unique value proposition so basics is like email or networking or storage or stuff like that you definitely don't want to do anything unique on that layer it's something that you buy from Google the G Suite or from Microsoft 365 you just buy it right you just buy it Right. No one cares. It's cheap. It's commodity. It's easy to introduce. The second level, think of core banking system or fleet management systems, something that is specific for specific industry. It's configurable, but across the industry, it's more or less the same. It's not an advantage. It's disadvantage if you don't use it, but it's not an advantage. And on top of that, you have the unique value proposition. Like, we are really good in being flexible with or we deliver in that particular area.

You cannot get that. Only larger enterprises can build custom software to optimize the top of the pyramid. And what happens with AI is the smaller and medium businesses now can afford to do it not in human heads, but with automation by doing it themselves. I think this is the dramatic change that AI is introducing. It's allowing smaller businesses,

to automate the last mile and the last, like, the top of the pyramid of the value creation, which means that those businesses, let's say, take a typical non-a-T business. Their margin are often like supermarkets or transportation companies. Their margin is typically maybe five or even less percent. If you can automate the top of the pyramid and you can save 5% of the revenue on, optimization that's amazing achievement right so that's the two answers one is we do want unique and second there is actually something unique in every business that is actually required and currently is in the heads of people so that means if I kind of phrase this in

in a different way you're saying AI is basically leveling the playing field because also the SMBs can now use this technology to build their unique value proposition what makes them unique and it's

no longer just in the hands of the big organizations that actually have the money and capital to hire the engineers that can help them

but now with AI we have a better level set of this I mean we're saying for a decade now software is hitting the world

it means that the software company a company that is IT by definition and operating in whatever industry has an

With AI, this advantage is reduced because now non-softrow-driven company can do the same. I asked the same question to our previous guests that we had just an hour ago. Because if you think about, you know, we live in the AI native world. That means we have transformed. Does this mean we really see AI generated digital assets, whether it's code, tests data, but especially code, do we think this is really then, we're reaching a state with 90% of our code is AI generated, production ready, and then what happens in case something fails?

Is it the AI that manages the AI to fix the problem? Where is the human, whether you see the human, the expert in this? I think we actually, we had few wrong tables with quite senior people in last year. And this question comes every time, I think. I think it is totally obvious that we cannot understand everything AI does. I mean, it does it in different ways that we never imagined. So how can we assume that we can understand something we could never imagine? So there is an element of explainability, of course.

So we need to train AI in the way that it was. explain to us what it is doing. But then you are relying on it to explain to you, what if it lies, right? And there are all kinds of signs that it may lie or it hallucinates. So we cannot really control it. I think there are sort of two things that we need to accept. One is, yeah, we just need to accept it.

But the same way that we accept cars and other machinery, that we don't understand and we drive them and they do massive amount of work. And if it hits us, is like a thing, then we, you know, it harms us and we may die. So it's not that much different. It's just performing cognitive instead of physical tasks. So that's one thing. And second, we need to create sort of levels of obstruction. So I was talking a couple of weeks ago with an investor about patterns. So he was thinking we need to create sort of this layer of

different technical patterns that they would sort of, the code would be based on this sort of basic building blocks. So the question is, can we create this kind of basic building blocks that are understandable for us and then let AI to actually build and manipulate them? So then it would be sort of a puzzle that we will explain to AI how to, what to build, and it will use this building blocks to combine it and then we will understand it. So basically create sort of higher level of abstraction where

we stop talking about lines of code because frankly who can understand code how many people today even can understand code. And if you write a piece of code it's very unlikely I will understand what you meant. So

to be realistic to be like if you really look into like the current truth is that we already don't understand what computers are doing. So it's illusion of control that we are going to lose. And then we

will have to accept decentralization and delegation of functionality to AI in some ways. It's maybe not reassuring, but it's the only way.

I think, though, it came to mind when you made the car analogy and I don't know what the path forward is here but with the car analogy those are built and designed by well let's say partially built designed by humans right so there's a level of accountability right when things are being done by AI in that sense

then there's really no accountability because it's just well the AI did it wrong maybe we'll shut down the AI the accountability goes down to maybe a financial penalty against the company as opposed to a design flaw introduced by a designer whose reputation will be tarnished, they might even get sued, or even if you had like a pilot who came on the plane drunk, right? Chances are, yet either someone will notice or the pilot will realize, hey, I might die if I fly this plane, so maybe I'm not going to, whereas there would be no accountability for the AI. And, you know, this is much more of an esoteric conversation, but this goes, I think, into what we're discussing here, you know, in this component right here, is how do we account for the lack of accountability? So first, it is bigger. Like, those are ethical and legal problems, right?

And again, this is quite common problem that people are discussing regulation around insurance. Who is paying? When there is a car, self-driving car incident, accidents, who is paying for it? And of course, there is this element of uncertainty. So the car, let's say, Tesla, they built a self-driving car that performed everything correctly. So there was no bug and there is still an accident. And then you can say, okay, so we don't know whom to blame. because there is no driver involved, and Tesla company is fine, so somebody has to pay for it.

But in reality, those cars are much more secure, so we are saving potentially millions of lives by having those cars. We just cannot assign the blame in the same way. So I really think this is going to change, right? This is just a matter of regulation. We got used to being hit by cars, unfortunately, right? It's not a common problem. And in the past, we were never hit by cars before cars existed.

So somehow we created this regulation. I don't think it's a big problem. I think a much bigger problem is, so generally, AI is much more reliable than humans on micro scale. But on macro scale, it can have cascading failure that humans cannot have. Because every human has its own understanding. of the environment. We rely on this heroic action by a single person not to shoot nuclear missile from Russian submarine, right? Now, AI doesn't have those blockers. I think that's potentially much bigger problem because when AI fails, it fails all the way through without any

humans involved, right? So that's why I think the only way we can deal with it is by putting human in the loop everywhere. Even if it's not needed, this is the circuit breakers. that will prevent cascading failures. Okay. This is philosophical now, right? So I already know if we're going this way or not. I'm just hoping. I'm hoping that we're not going to the sky net and stuff like that.

But I think also in your book, right, and you sent us a little bit of a quick abstract or overview of the different parts that you have. And if I read this a little bit correctly, you also say, like what you mean? mentioned your little transition. You start small. You try to figure out, you know, AI can help you.

Then you really try to figure out what else can AI do. And then at some point you end up in a situation, what does AI do? And what's the human part? So I think that dial between, you know, how much percentage is AI and where does the human come in as a controlling mechanism, is a circuit break you mentioned. As somebody that confirms, that takes the risk maybe until we have also also. all the regulation, or until we get used to it, as you said, we've always been hit by cars. But back in the days, it maybe was a drunk person that drove the car, or they were tired,

they fell asleep and caused an accident. And now, hopefully, 50, 60, 70 percent less than before. We still get hit by a car. And now it's an AI driving, but we need to understand that in the end we saved. We're less risky on the street. So I think that's an important thing. something we control. Like talking about what's going to happen 10 or 20 or 50 years from now, will we have AGI or like, I don't know. I don't. I'm not like, I can try to predict it,

but I'm not better than others in this. I think what is more important and more relevant, especially for us as practitioners, is what you do tomorrow or today. And what you do is, in our opinion, that's what we're writing in the book, is that first, everyone should be investing in AI because this is obvious. the next thing. Regardless what people say, there might be a bit of burst of the bubble. Maybe not. I don't know. It's irrelevant. It's the same as internet. It's it, right? It's the next wave of innovation. It's clearly valuable. It's clearly going to change our lives. Now, you don't want to stay behind, but you also cannot jump all in because we don't yet know the problems. That's why the first steps is sort of, we call it pioneering mode when you have a small skunk work style team working in a sandbox, researching, building prototypes, doing things that are very low-cost, super innovative,

with the only purpose to learn an experiment and find a business case. And once you find a business case, and that's also something that people, why 95% of the AI initiatives didn't succeed, it's not that they failed. They just had too much expectations. So you need to find something small, simple, that will justify small investment, but not expect like doubling your revenue or saving 50% of your cost.

Something really small on the side. So you can start building infrastructure, platform, team structure, new culture. And then once you have that MVP, then you bring another team and another team and you scale down the all setup and scale up the new setup. And gradually, you get to the point where you build a new thing by replacing the old one. And this is exactly the same story as we recommend it with Cloud Native. So you start with this pioneering mall with independent small team, and then you gradually scale up. And again, if you look around in any business management or business books, you will find that this is the only way to transform your organization.

Yeah, and then the first step of that, just like with Cloud, is. finding out what project will benefit the most from moving into AI base, right? Just like with the cloud, right? Because there was always, as we saw,

you know, it's interesting you talked about the internet. I recently saw a similar article where there was the internet bubble, but it's not like the internet crashed. It was all these just superfluous things that grew up around it didn't survive. But then the

foundation was there. And I think, you know, bubble or not with AI, chances are there's going to be a bubble burst, but the foundations are there. But similar with cloud native, people were just taking and brushing to put everything. Let's lift and shift and put everything into the cloud, right? And it was a field day of just pushing everything. And then finally, reason kicked in. People started settling down and say, okay, let's do this the right way. Let's think about it a little more. So we're going to see a lot of mistakes. We're going to see a lot of very eager people looking to jump in and making those mistakes. But what I think is really important is the conversations and things like the books you're putting out

keep happening so that as people start saying, all right, let me slow down, let me look for guidance. There's guidance there and people can start making good decisions. My first proper job after the college was in Checkpoint, which is firewalls, air firewalls and VPNs. And I started in mid-99, so like peak of the bubble. And, yeah, I mean, checkpoint was affected to a certain extent because the whole bubble burst, right, and it was very dramatic. But people still needed firewalls. So maybe not everyone needed petshop.com, right?

But people needed firewalls. So in this sense, yeah, there was exactly like you say, companies that were over-invested, over-leveraged, over-leveraged, overcoat all kinds of things. They disappeared, right? But internet is a basic technology that we cannot imagine our lives without and all kinds of things like social media

appear that we never imagined in 99. And yeah, this wave of innovation is happening, for sure. And the question is how you're not getting into this sort of bubble style imaginary things that will never happen and never bring value, or you're more pragmatic and you're finding small inefficiencies in your company that will not bring, again, 50% of cost reduction of the cloud

or something like that. But just enough to increase productivity of two development teams in the next few months, right? But it's fairly easy to do because those teams are eager to start using technology. Their tools are relatively new, or the product that they are building is relatively new, so it's relatively easy to adopt new technologies.

So you find an easy way. It may not be even the best cost-saving exercise. It's just important that it brings value, and it takes you one step forward. And as you do that, you build a bit of platform, a bit of infrastructure, and then step by step, you move forward. This is the pragmatic way of adopting new technologies. And it's effective.

It's not dramatic like people wanted to be. Like, it's not spectacular, right? But after three years, the achievement is amazing. And I think the big challenge, too, is that everybody wants to race and be the first because they feel like they'll lose their competition if they're not. But just like with the dot-com bubble, if they rush in too fast and don't build that solid foundation, they're going to collapse in on themselves. So there's a definite balance of taking prescriptive steps, adopting, you know, somewhat aggressively, but with purpose and mindfulness.

It's very, very simple. Even now, more or less 10 years since Kubernetes exists, almost. Most companies don't know how to use it. So, or not using it effectively, right? So it means you have time. I think this sort of assumption that AI is so, so fast that in two years everyone is going to do. I'm actually talking to what real people are doing workshops and meetups and we talk to people who read the book. Most people don't touch AI.

And if they do, they just ask a couple of questions from chat GPT. You have time. Unless you build AI specific tools, like you are. literally open AI or something like that, then of course you are running against the time. But if you are a software company building e-commerce, then you have time, not like a lot of time, but you have time. You know what? There's another, it's a company, I can't name the name, but this is very common.

So we came there at some point to do front-end development, and they told us you cannot refactor or anything because it's really, we don't have time for that. And then you have to do it this way because that's how we're doing it. And you cannot touch anything. And it takes them two weeks to the tiny piece of functionality in front end. And in a few weeks, we can refactor it so you can do the same work in a day. Like, no, no, we don't have time for that. So that's the real problems.

And most of the companies are dealing with this kind of problems. and not like they're amazing with technology, and they're just waiting until Chad GPT will release the next version and then they will go to them on. Pini, do you think the problem and maybe the misperception that if we don't act, we will be overrun? It's also because there's a lot of startups right now in that space. They all need to create a lot of noise.

And obviously the echo chamber we are living in, right, the social media that is that we are only hearing AI, right? it feels like even on local news channels, where I never thought that this is a topic, all of a sudden, AI is everywhere. And I think the perception, obviously, if you only listen to this, it feels like we have to do somebody and we're falling behind. So it's like, to your point, right, we do have time, but currently we live in this loud, noisy environment where especially driven by some of the big players and also, I guess,

a lot of the startups that try to make money now. This is why we have this perception that you need to act now or you die. And that's not the truth. And that's why companies like Accentia and doing making millions of AI, right? And 95% of the POCs are not bringing value by MIT research, right? So this is the result of it. People are like jumping in too big and too fast. I'm not saying don't do it.

I mean, it's literally our business. We're basing our entire business on adoption of AI. And we believe this is the next thing, and you have to act now. But the question how, if you put now 10 million into some big project, it's very unlikely to work because 10 million means 100 people actually doing something. Now, you need to organize those people. You need to put them in the office. You need to create management structure.

You need to give them email addresses, laptops. Just doing that, it takes in an enterprise a year, right? Once we came to a customer, enterprise customer first auction. Our contract was six months, and it took four months to get laptops to actually access their system. So unless you're actually dealing with those problems, what's the point? Now, you can say they will over on me. With all the respect, if you're a bank and you have customers, they're not going away in a day or in the week or in the month.

If you compare traditional banks to all this new ones, Monzo and N26 and Starling, it took years, right? And it was in plain sight that they're going to grow and take the market. But it took years for them to do that. So the bigger banks had all the time in the world to catch up. The reason they didn't catch up, It's not because the technology wasn't available or they were not trying, but because they were doing it wrong, because they were doing it in the old way.

They were trying to use new technology in the old way. So they didn't get the value out of that technology. And I think that's what's happening now. That's why a lot of projects are failing in AI. It's not because AI is wrong. It's not because it's not useful. But because all those putting, pouring billions into using it don't know how to use it. And they are not investing in research and learning how to do it,

but they are spreading the money like crazy on consultants, consultants, which makes my wife very difficult because we go to a company and say, we are consultants doing AI. And they're like, I've already used 10 of them. They're all shit, right? Sorry for a... And they're right.

And how can I explain to them that we are not that? Right? This is a problem. The problem is not AI is not. ready. It doesn't matter. Technology is technology. Early technology is fine. The problem is that people don't know how to apply it and they need to learn and learning takes time. I mean, everybody that listens now has been listening so far in this episode. Hopefully understands that you obviously put a lot of thought into this and that you're definitely

a trustworthy source when it comes to this. Pini, if you can do me a favor, hopefully, I mean, you have your laptop in front of you, but you sent us an email in preparation of this podcast, kind of with the overview, and in the very end, you had some closing notes. And if you can do me the favor, because for me, I think you mentioned the word pragmatic earlier, right? You need to have a pragmatic approach to becoming AI native. And I think that's also what you're explaining in the book. Now, this is not going to be the end of the podcast, because I still want to talk about how

you wrote the book and how AI helped you. But I would like for you to read out the closing note, because for me, this is a perfect summary of what you're trying to tell us here. And I think also showing why you are different than many others out there that are just making noise to generate money without the right outcome. So, Pini, becoming AI native. What does this really mean for you? So this is the closing note from the summary. Right. Transformation isn't one-time project, it's a way of thinking.

AI doesn't replace people, it replaces friction. The rest meaning creativity, curiosity, remain entirely human. And I thought this is beautiful. And actually it also explains what you said earlier, right? Transformation doesn't happen overnight. Transformation is a bigger change. I really like the analogy that you also brought when we moved to micro- services that the transformation then also transformed our organizations, our structures,

and the way we think AI, native will bring the same transformation with it. It doesn't happen over time. I also like that you brought in AI doesn't replace people. It replaces friction. And this is, I think, similar, Brian, what we discussed in the previous session, right? It's funny you went there because that's exactly where I was thinking you were going. Yeah, yeah. It plays friction.

Yeah, and the rest and the meaning, the creativity, the curiosity, the innovation, the inventions, they all remain human. And that's exactly the point, is that when we have, let's say, five people doing cyber security in a company, right? And then we bring AI and AI, we say it will reduce the same functionality can be done or same work can be done with two people. What does it actually mean? Now you have three people. You can fire them. That is true. Or you can teach them or relocate them to other job.

Or you can do more cyber security. There are all kinds of options. I think what it actually means is that three out of 40%, 60% of your work was actually not very good work. It was toil. It was a waste of time. And now you remove that. And now two people with their creativity can do work

of five. Now the other three can use their creativity to do something better. And again, going back to 200 years back when machinery in agriculture replaced people. So yes, short term, it was problematic because there was job displacement. People moved from villages to cities and they moved to factories, but that actually allowed creation of factories and then later service economy. So, yes, there is displacement of jobs. Of course it will happen because you can do more with less people. But it actually not, we shouldn't think about this as people losing their jobs. We should think about this as people are free to do better work.

And we will always have more things to do. I think this sort of zero-sum game that people say that once I can't do what I'm used to do, like if I'm driving a car and now they are self-driving cars, now I'm useless, this is a wrong way to think about this. It's also an entirely wrong way to think that AI will replace your relationship with your customers. It's more like you still have people, but they use AI systems to interact with a variety of different systems to enable them to provide better service.

All the companies that will replace human interface working with customers with AI will suffer from it, the same way as they suffered when they moved the coal centers to India. Thank you so much, Pini, for that. Folks, again, check out this book. Links in the description now to close this episode, Pini, because there's a lot of food for thought that he already gave us. But in the preparation of this podcast, you said, you know, one of the things we could also talk aloud, a touch base upon is how you wrote this book.

that you actually used AI and how you used it, the lessons learn, and using AI to write a book. Now, I'm currently writing a book as well, trying not to use AI. But I would like to hear from you. What is your experience? Where did you use AI? Was it helpful? Where did it reduce friction? But where did it still allow you to keep your creativity, your curiosity, and your meaning to basically take the same words that you just mentioned earlier?

So I think, so first, I have written a book before, right? And about six years ago, which is that one from O'Reilly called Cloud Native Transformation. So in this sense, I can compare the differences. Because you can argue, like, how do you know how to write a book? Like, you know, it's a, and it's not a real book because it's a, yeah, it's not you. So I can compare it. And also, I use human editors and feedback from human people, so humans. And so I believe that this is a better book that the one I wrote before.

I think it was really difficult process, right? I think the first thing, and I did a talk twice about this and conference and meetup. And the first thing, the first slide was like asking AI, write me a book. And that doesn't work at all, right? Actually, worse, it does work. It does work, but it doesn't give you a good text, right? So you can ask AI to generate any amount of text, but it's terrible text. It's not original.

It's not good language. It's just useless. And you see a lot of small books coming out now that's written by AI and you don't want to read them. So there was a very long process of learning how to use it. So how to brainstorm with AI, how to, for example, using language. Like, English is not my native language. So editing is something that AI can do very easily.

At some point, I realized that I can use first book to teach LLM, my tone of voice and my language. That was like magical. You just drop entire PDF of the old book and then say, use that book to basically learn the style. And it just started writing like it's me. Yeah, and then there was a lot of things like you ask it to change something and it changes too much and you don't know what the diff is, so you cannot really check the differences. So you need to read it again and it's very frustrating. And there are so many problems with it. But I think eventually you realize that it's a tool that can help you in certain things, like editing, like writing a large number of words.

like creating patterns. It's amazing in creating patterns, actually. But you still need to give your ideas because otherwise it's not your book. And I think AI will allow a lot more people with great ideas to express themselves, but it will also generate a lot of noise because a lot of people will think that they can do it,

but they can't actually. So at the end, it was very, really difficult process. Time-wise, it took the same year as the first book. It was a lot of thinking. And eventually you sort of, you have to chop it in pieces and deal with each piece separately. But then you can go into much more depth on each one of them and use AI to sort of as a sparring partner and as an editor and as many other things.

So, Pini then, because it just said you had a lot of time for thinking, earlier in the very beginning of the podcast you mentioned it 10% thinking 90% coding that basically if I hear this correctly you could spend more time in thinking and creating better content and the writing is optimized

because you don't have to write it imagine research right like you have an idea and you know you need to do research what others say you'll you know you open 20 tabs and Google you search you go through different things and you read You compile it, like days and weeks later, you have a page of text that sort of combines all together.

Now, what you do now, you do deep research. It takes three minutes. It comes back with all that research. Then you take a walk for an hour. And then you crystallize what your opinion is about. And then you ask, what, you see that sort of. You write it in whatever language you like. You actually recently started talking to each other.

GPT and Gemini. So it doesn't have to be structured. It doesn't have to be short. It doesn't have to be anything. So this is the amazing part. You can do massive amount of research, condense it into short, consumable piece of information. Take a walk to think and then give your opinion

and let it actually formulate and that opinion and write it in your own language. Yeah, I especially like the part that you. said about the language, right? Because you remember looking at the notes, and one of the use cases was like, well, yeah, English is not my first language. And I was like, that's a phenomenal use for AI to help you with that, right?

And I also think, you know, obviously there's a lot of debate of using AI in the different various arts, right? And I think what people have to realize is there's different types of books, let's say, right? this is a little bit more it's not a fiction book right it's nonfiction

it's also not a history book but being written by a historical author who has a very very distinct voice right yes you do have your voice but it's a little bit more of technical and some other stuff so it's like for different levels of where you're going there are different

levels of let's call ethical AI to use right if I was going to write a fiction book and I told chat GPT you know I want to write a a book where Andy goes to his favorite fast food store. I forget the name of it Andy, right? And he gets... He gets held at gunpoint

and then saves the day. And it spits something out. It's like, that's complete cheating, right? But there are different levels, you know, on different things. So what people have to keep in mind is that, yes, there are use cases for AI and all different levels of these things. And I think

the more important piece is that we as creators, artists, whatever level you're putting this stuff out as gets ahead of how to use AI to help this, as opposed to having it forced down your throat later by the companies who start deciding, this is how are you going to use it, right? So as a writer, as a musician, whenever, if we're taking the helm and figuring out those creative ways to use AI or those very utilitary and helpful ways to use AI in this process, we can make it a good process for it to be used in the future.

I really think a lot is going to change I think I don't know if a book is the right format if people still buying the book it actually was saving quite well got to bestseller status congratulations so my feeling was writing a book

with or without AI it actually doesn't really matter it matters in a way like code written with or without AI it actually doesn't matter as long as it does functionality you want, and you decide on functionality. I think people that don't bring their unique value into the book or into their code or into their thing,

then they're missing the point. Then they're creating something that is commodity, that is useless, that is already there, essentially. Because AI is a statistical tool to shake the box and get out something that already exists in just slightly different way, right? You have to add your personality, your ideas. If you don't do that, that's why it will create a lot of noise, because a lot of people will create a lot of things, but without adding new ideas.

So writing a book is easier. Writing 60,000 words is fairly easy, but adding your ideas is still difficult. There are some people that they have amazing ideas, but don't know how to write 60,000 words in nice way, those people will be amazing addition to humanity in general because they will be able to express themselves using AI.

Yeah. This reminds me, and this might also be one of our closing statements. At Cloud Native Days, Austria, I had the pleasure to MC that event three weeks ago. We had one of our keynote speakers, Ali, he

talked also about the challenges of AI and like what this means for us as humans. And he's a, he is a UNESCO UN, youth ambassador. He's helping, especially the young generation, to learn about what the next job should be and give

them opportunities and perspective. But he had a very interesting quote. And I think that Pete fits, he said, you are unique, stay unique, don't become a copy of somebody else. And basically what you're saying, if I'm just using AI to create, I just become

copy and I create noise based on copy from somebody else. The magic thing is to take your uniqueness and put it in. I think that's a perfect, perfect statement is AI, that's where the AI is best, right? It helps you to emphasize or express yourself in your unique way. And it can help you to achieve so much more with less. But if you're not doing that, if you're just using AI to copy, the others, that's not adding anything to anything. Right.

And I think the responsibility comes down to the consumers, right? Because there are going to be tons of people who take advantage of copying. You know, we see it in music all the time. We see it everything. There's a hit. Everyone's going to do the same thing, right? So it comes down to us as consumers to embrace the people who are injecting their voice into it, who aren't doing that.

And hopefully keep labeling, you know, what levels of AI are used in the, creation, not using it as a turnoff, right? Because again, if it's like generated completely by AI, yeah, good. Right, but to have people, all right, if it's partially, if AI was used but wasn't the main thing, don't just dismiss it. Check it out, right? Be open to these new voices. But again, as consumers, we're going to get what we pick.

So that's where it really is going to come down to. Yeah. It's definitely, it helps creating a lot more noise. Yeah. I mean, I use it, you know, I know we're wrap it up, but I use it all the time. Like I do music production, and there's this process of mastering where you have to make it ready for like all the levels are the same. It does final tweaks to the EQ, the overall sound and everything. And it's a very, very, very fine art. I have an audience

of maybe like 200 people, right? I'm not going to pay a mastering engineer. So there's tools that you can pop your music into. It'll give you based on different algorithms that'll analyze your music, great, but it's not creating my melodies. You know what I mean? So, definitely use this. Anyhow.

Pini, it's been amazing how we can just keep going on and on, right? But I think we'll wear the audience out. Andy, did you, Andy or Pini, any last statements? Or we're all good at this point? It was a great conversation.

We went a bit philosophical or macroeconomics and stuff I mean it's fun to talk about these things but it's also it's very practical I think people should understand that this is this is a real thing regardless if there will be ups and downs there will be always ups and downs

on every technology it's a real thing and you can't miss it because if you will miss this wave it will be dramatic it will be dramatic in the impact on everything we do I think just maybe not as fast as

we think at all day well Pini I wish you all the best with your book we will make sure to mention it to our communities like through this podcast that this is definitely a book to check out for me a pragmatic approach to the next

AI native transformation and yeah as we're still learning what the I-Native really means, I would love to also invite you back to the podcast, maybe in a couple of months, to see the new learnings and the new evolution, how in which direction is moving. I'm always happy to share. Until then, we're working on a few projects, which seems to be very exciting, in very

practical ways. So hopefully we won't have more to share, or I will have more to share on behalf of our team. Cool. Thank you so much. Thank you, everybody. Thank you. Thank you.