The Best Way To Scale A Tech Company Is To Stop Selling Hours

The old way to grow an IT services firm is brutally straightforward: hire more people, ship more projects, track bench percentages, and hope margins hold. But that model starts to feel like a trap the moment “people” becomes your biggest constraint. I’m joined by tech industry veteran Igor Kruglyak to unpack the pivot from services to software products, and why the next wave of leverage is coming from platforms, reusable components, and AI-assisted delivery. 

We trace Igor’s path from the early commercial days of computing to web, mobile, and now agentic AI, and we get practical about what actually changes when you build a product. Igor explains how Prime blends the control of custom engineering with repeatable building blocks for mobile-first solutions, especially in document-heavy workflows where PDFs, metadata, and workflow state create the real complexity. That “context layer” becomes the foundation for agentic AI, because an assistant can’t answer safely without knowing whose data it is handling, what process it supports, and what constraints apply. 

Compliance runs through the entire conversation, from HIPAA-style expectations to audit-ready records and guardrails for AI. We also dig into how teams are already using tools like Microsoft Copilot and AWS Bedrock, why model swapping matters, and why “orchestration” may replace pure coding as the core skill. Subscribe, share this with a builder who’s rethinking services, and leave a review if you want more conversations like this. What part of your workflow needs a real context layer first?

Transcript

Alex Romanovich 0:26

Hi, this is Alex Romanovich and welcome to Global Edge Talk. Today, March 13th, 2026, and I'm joined by a very good friend, uh Igor Kruglyak. Hi, Igor.

Igor Kruglyak 0:38

Friday the thirty

Alex Romanovich 0:39

Yes, and it is Friday the 13th, so our discussions could be dangerous and at times spooky.

Igor Kruglyak 0:45

Exactly.

Alex Romanovich 0:46

Let me jump in and introduce you to our audience. Igor is an industry veteran in technology, in IT services, in software development, in product development. He has an an incredible experience also in taking those entities and conducting transactions with those entities. So he was involved with the MA transactions. He was involved with acquisitions. So he brings a lot of experience with companies that he's been involved in. Tremendous and a lot of experience in healthcare, health tech, medical devices. And obviously he's been involved with a number of other segments, but that one is very near and dear to him. We'll talk more about that. Igor, do you want to add something to your introduction?

Igor Kruglyak 1:43

Sure. And thank you, Alex, as always. And it's a pleasure joining you on this podcast. And I wish you a lot of luck and uh, you know, as as we as you build the audience for this. Yeah, I had a unique perspective and view on technology. The being kind of graduating high school where the computers were just being introduced, going through college where the alumni of Brooklyn College, one of the points of the DARPA system with the with the first internet was built, and then spending years at uh on Wall Street that really pioneered the commercial use of computers. I happened to be one of the guys who brought in the first PC onto the floor of New York Stock Exchange. And, you know, it's uh for a company which was very infamous at that time, Drexel Burnham Lambert, spent time in Merrill Lynch, managed the small to medium to large companies, then decided to go into technology and products, worked in a very interesting market of emulation of mainframe software for PCs, focused specifically on reducing the cost of engineering. I think that's a theme that we can focus on. At the time when I joined the dark side of products, the technology was much more expensive and physical hardware, the iron, than the people, right? And then as I transitioned through the client server, web, you know, mobile, supermobile, whatever else, and now into AI, you see completely different shift, right? You see a shift from hardware to people, right? Which is the where I did the bulk of my kind of business endeavor to now this new thing about hardware again, where hardware is seemed to be more interesting and expensive than the people. So I think it's a it's a rainbow. And I would love to talk about the lessons I think we'll learn from each of these sections. And maybe you and I can predict a little bit of the future.

Alex Romanovich 3:49

Absolutely. That's what we're here for. A little bit of a futurist moment, if you will. And I totally agree with you. It is such an incredible transformation within the industry to go from people to software to hardware from a standpoint of uh importance or focus areas. I remember, you know, uh selling hardware and software bundles for silicon graphics, a company that's no longer in existence. And actually at that time, nobody even realized that the graphics chips in silicon graphics and what we see now in NVIDIA would give birth to this uh incredible new movement of artificial intelligence and data center movement, a build-out movement and so forth, something you and I are very familiar with. But let me ask you, you have traversed a number of different areas. You have um, you know, from all those areas very easily and very frequently. And going from IT services and going from software development to product, when did you actually have that light bulb moment when you realize that continuing as a pure services firm or continuing as a pure software development firm is no longer an optimal path? And that's why you woke up one day and said, you know what, I'd like to actually develop a product. And that's how Prime came to life. And uh this is what you're actually promoting right now. But when was that sort of a light bulb uh realization, if you will?

Igor Kruglyak 5:16

Well, I think it's a great question, Alex. And actually, you and I met at the intersection of the ideas, right? I think back in 2013 became apparent that mobile is going to change the dynamic of the interaction in the business, in technology, and in human interaction, right? We didn't know how, but we knew exactly what. And um, you know, in the product journey, really, I could say traditionally, build it once, sell 250 times, get 20x valuation, right? This is all the dream of every entrepreneur. So entrepreneur in me always said, I'm gonna build something and I'm gonna be able to sell it many times, right? As a as a software delivery, right, as I as I went through, as you mentioned, number of projects in the pharmatech, number of projects in healthcare, number of projects in legal tech, right? You see patterns develop. And these patterns sometimes you we call them know-how, sometimes we call them experience, right? And uh, interestingly enough, our customers don't really even want to take the ownership of that, right? They are focused on making their dollar, right? Their value is delivering value to their customers, receiving. They're not a technology company, they are business delivery companies. We are the technology companies, and for us to be able to thrive, grow, right? We need to come up with something which is a nonlinear. The easiest way to grow is to hire more people. But to hire more people, it will be higher and more cost. When higher and more cost, it gives you more overhead, right? Now you need five HR, you need this much, you know, and you get into the situation, and then you get into an environment where me as an entrepreneur and the CEO is no longer interested in the business because it becomes management by numbers. It becomes we have 2% on the bench, which there's 4% on the bench, and 4% is unacceptable, but 2.5% is acceptable, where people capital becomes numbers and becomes cost, right? And that's that's I think is a dividing line, right? So if you can say, okay, I have a cost, but I it's it is my investment into function that will be the product, right? And it doesn't have to have exactly four corners. None of the products do, right? They all and when Benioff sat down with this his original team in Salesforce, which is one of the first products we actually utilized as part of our kind of a value plus chain, he did not see and understand where he was going to go, right? But they had an open mind and they understood that there was a niche, there was a need, and there was a way to solve that need using technology. Same thing we do right now. Right now we're looking at what we would call carnage, right? Of the SI model, right? The system integrator model is broken. It the new model hasn't been formed, and it's not clear to me when this new model or this new way of doing business is going to manifest itself, right? Somebody's, you and I were talking about anthropic or something like that. None of these companies are making a dollar, right? They're hundreds of millions of dollars in red right now, right? And and the and the dividing factor or guiding factor for them may not even be the skill set. It will be energy, it will be ability to cool the data center, or things that we absolutely have no, you know, no idea of foresight. So I think uh for me, in general, is travel the journey and make adjustments along the way makes it worthwhile and gives me the ability to do so. So I've transitioned from running a 750 people uh delivery team, right, delivering to 30 clients, uh generating $35, $40 million in annual revenue, to now focusing three to five clients that are on the platform. And they, by being on that platform, actually contributing to the next release, right? So and I can give you an example of that if you like.

Alex Romanovich 9:38

Yeah, I mean, but but what's interesting about it is I mean, the Entropics, the Chad GPTs, they are making money, they're making revenue. You know, whether they're profitable or not, it's a different story. But they certainly have not, I agree with you that they certainly have not tapped into a endless possibilities of monetization and how they will monetize those technologies, those models, and so forth. The other thing that's interesting is that if you were to build the IT services or software development company today, you now have options. You would not necessarily hire humans, as you said before, right? And the human costs would be now weighed against the costs of hiring agents or building agents to build certain functions, right? And even some of the models that are prevalent right now, whether they're Claude or Gemini or Anti-Gravity from Google or some of the other ones, they're now offering a variety of different uh virtual avatar-based uh roles, if you will, on the software development team. And how software development folks are using them is also and programming them and teaching them is also going to be very important. Now, let's talk about the product you've launched recently, or not so recently, if you will. I mean, I think you've been, you know, you've been uh iterating with that product, and the product is prime uh in the area of mobility, in the area of delivering content uh to, you know, within the mobile environment, and now obviously the product has evolved quite a bit. What do you see? Tell us a little bit more about the product, number one, and number two, how do you see the future of mobility? And how do you see the future of mobile content delivery, service delivery, whether it's autonomous or not autonomous, and so forth. Tell us more.

Igor Kruglyak 11:22

Well, this is a very wide topic, so let me pick some things. You know, let's talk about why and then the how and then when, right? So the why, the uh the idea that you have to build. So basically there are two options for any, you know, with the soon-to-be third. One is I can custom develop an application, whether it is an Android application, iOS application, you know, or cross-platform application. Hire a whole bunch of guys, give them a whole bunch of tools uh you uh using using different technologies. At the end of the day, I will own a pile of code, and this pile of code will compile, it will end up in the app store, and off I go. That's a traditional model. Different model, which is what we were touching on when we talked a little bit about Salesforce, is basically a domain model, right? You're basically no code, low-code, custom code structure, but you're domain bound by Salesforce, meaning they co-own your data, whatever the uh they co-own your processes, they co-they give you governors, they give you guidance, and you have to hire people who have the skill set to manage that environment, right? But they give you a lot of functionality. So that if you're gonna do a CRM, why do you need to write a new CRM? There's one right there, right? Um, the third one, which you talked about, which we really haven't seen the settled value yet, is this vibe programming, right? Or using the models, using the agents. And I have some thoughts about that. May discuss it later. So what I have decided, I wanted to take the best of two worlds, right? So in the custom engineering, I control the pace, I control the feature function, we do we do uh sprint planning, we do all kinds of stuff. We know what we're gonna show to the customer when, right? In the uh hosted model, or if you want to call it the uh no-code uh enterprise model, you basically are just you're iterating feature functions of that of that system until you get to the point where your customer says, this is great, this looks good. I want it to be in both worlds. So the first and foremost, I build certain things which I consider basic building blocks of applications, right? So we call them content components and application components. So if, for example, almost 80% of all my apps today, especially business focus, handle PDFs, right? So you need to uh send the PDF, you need to receive a PDF, you need to open PDF or whatever, upload, download, all kinds of stuff. Well, why do we need to write a whole bunch of code around it, right? We can create the attributes, we can create the metadata about the uh this PDF, and then what ends up happening is you're putting this PDF in context. And this is a very important aspect. Context of which we consume data is what enables us to manage the processes, and more importantly, now in this agentic AI technology, we can bring context to the agent. So if I have an agent that needs to answer questions about loan application or a final loan document, right, I need to understand the context in which that data is being presented. So it's presented for Alex Romanovich, it's presented for his house in Staten Island, right? You know, and so on and so forth. And all the other attributes are put together in one package. And believe it or not, as I look at all the applications being built today, 80% of the time is used to create that context, right? So why don't we build that context using the type of uh technology and labor that is not technical, that is not coders. Coders are terrible in this. Coders want to, you know, basically just add bits together, right? At the end of the day. Business analysts are the ones who should be handling that. So where is the difference from Salesforce? I do not want you to be my client for hosting. I want to be your vendor for your solution, right? So what I mean by that is we're gonna employ we we are providing tech transfer. Once we build the POC, MVP, V1, whatever, that goes over to your domain, if you will. And now you're in the complete ownership, and the technology that we use is so basically open source, meaning it's it's the best of technologies. It's offered by AWS, like OpenSearch, or or some of the other tech that we use, like Redis and um and other things, or like using authentication, the OpenAIGO or some other authentication engines, that allows us to really focus on these three things. And the most important thing, I think, from our from my perspective, is to condense the time to delivery. Right? From the time I sit down with you and talk about what you're trying to do to the time you see something, right? It should be in the small increments. I want to keep your attention, I want you to be focused. One of the things that I always told my prospects and customers that at the end of the day, a technology company like Core Value or Now Behold can produce a lot more tech than my customer can absorb, right? So at the end of the day, when I get into UAT, when I get into user acceptance testing, right, they're overwhelmed. You know, we'll be throwing releases and they're like, well, wait, I gotta run my regular business, I gotta be focused on this, I gotta be doing this. So we need to create that trade-off. And I think this is these are the things that Prime allows you to do. And it allows you to create this legacy. And then the last thing that I think that I wanted to focus on Prime is the areas of verticals that we talked about. Whenever you build something and you do not have compliance built in from day zero, right? Retrofitting compliance is one of the hardest aspects of any kind of critical development. And compliance is defined in many ways, but the way I define compliance is that I have to tell you how am I doing things. I have to teach somebody how on how I'm doing it. I want to make sure that they have confirmed that they understood what I've taught them. And then I have to have a record to show that they performed to the standards that I accepted. And if you don't have that, you're vulnerable to either be found to be not in compliance, in which case you can lose money, you can be fined, you can lose licenses, or you could be sued if you're not able to verify certain level of compliance. Compliance is the key, and this is an important aspect to to kind of communicate to somebody who is building a new insurance application or building a new patient management application, or is trying to build a device device uh something for a medical device industry. Yeah. Those are the type of items that I focused on, and that's what we build.

Alex Romanovich 18:45

Great. So uh within the uh within the uh mobile framework that you're building, which is obviously bringing compliance into it as well, when the content or information is being distributed, it is compliant within the um, let's say, HIPAA standard or something along those lines. And of course, the regulation is not decreasing, regulation is going to continue to increase, especially with the introduction of uh artificial intelligence into it and so forth. But let me ask you, speaking of products and speaking of functions and features, as you were developing the product over these years, over this time, you obviously had experience in bringing in humans and bringing in resources and managing those resources and so forth. Let's enter the world of AI. What are you going to do different, let's say, for the next version of the product? Are you going to actually use AI now in terms of a gentic implementation, or are you continuing to use the human component? I mean, I know the team is not that large, but still, right? There's certain things that you can say, hey, you know what, AI, help me optimize the software development environment. And let you this, I'll let you decide and recommend to me where we can take some of the remedial tasks and some of the tactical tasks and cons, you know, and outsource it to some of the agents and now, you know, concentrate on some of the more strategic tasks and integration tasks or legacy, you know, legacy data information that may be available at a customer site or something like that, and do it that way. Is there a, you know, is this the future of software development or product development for that matter?

Igor Kruglyak 20:20

So the short answer is yes. So and let me give you this one. So let me first tell you, I look at it as two sides of the coin, right? What our customers are going to be using and what we are using, right? What we are using is always impacted by time frames and the ability, right? And this is actually where AI is helping a great deal, right? The uh the technology stack that we have, for example, we use Aonic for certain things, required building a container for one of our key functions. And we realized about halfway through that this container was not sufficient for the functionality that we were implementing for our customer. So, what we had to do in the old days, we would have then gone out and hired two expensive consultants, one for Java and one for Objective-C or the other platform. What we ended up doing is the guy who has really understood the problem said with Copilot, I believe for Microsoft, this is the one we use internally, and was able to build Java code and the code on the uh the Apple to be able to manage that component. And uh one of the things that I also uh want to make sure is that you cannot just build mobile applications. We call it mobile first platform. But web interface continues to be an important aspect, right? These notebooks are not dying, iPads haven't replaced them, and there are still plenty of business to be done in kind of a hybrid environment where you have stuff on mobile and you mobile and you have more of a um uh kind of a desktop two-dimensional component. And, you know, so we we invested in and make sure that when we deploy a solution for a customer, it has all three components in it. So that was for on the engineering side. On the customer side, what we've what we were able to do is utilizing AWS bedrock, right, is to give the give our customer an opportunity to kind of manage and increment. Build on top of the Agentec solution. So we're using a model that is specifically focused on fintech, specifically optimized for loans, right? This is a great model. It's also inexpensive. Actually, it's free for AWS users at the moment, right? But it may change. And let's say two years from now, or 30 days from now, or 15 minutes from now, there'll be a new model. So if you don't build the info the technology in such a way that you can swap out this model and replace it with another model, and the name of it just escapes me at the moment. And we can we'll add that in the notes to our conversation. So you're not doing, you're not giving them any um uh you're not you're not helping them in that essence. The other thing is is that you have to create fluidity in terms of how your agenc solution is interacting with the customer. So the traditionally, right, these free five pre-build questions, adding semantic analysis, giving a user an ability to rate answers, but also constraining the model, right? We are not doing another clause, or we're not doing another ground, whatever the we're not doing another. Those are general general population models. We need to focus our our Agenc solution to be very specific, and that will also be the part of compliance. So the compliance angle that you're bringing to Agentec solution is structured use of the technology in such a way that there are guardrails around it based on what your client tells you. And we continue to enhance our capabilities is to take business requirements written in English, right, and produce what we call t-shirt sizing. And that's an important exercise, right? And the more we will be investing a lot more time and money into it, where we take these business requirements, we throw them into Confluence, we have epics, we have user stories, but then we can run an analysis, you know, have the model tell us this is no code, this is low code, and it's simple, this is low code, but maybe it's a little bit more complex, and this is a custom code and it will require engineering. Now, whether we can use the custom code and some copilot, and again, I don't know yet about how you hire these agents. You know, obviously they don't have any headaches, their daughters and don't need to be taken to the doctor and uh, you know, other things, but the quality of the solution as well as the kind of, I don't know, we'll think of it as a pedigree, right? You know, when I select a senior engineer, I look at his past experiences, right? It will be difficult for us as managers of people and technology to be able to rate these agents, right? Their biases, every model has a bias, we have a bias, right? You know, if you if you spend um 90% of your time in Microsoft world, you have.NET bias, you have SQL Server bias, you have, you know, MongoDB bias. So these things will work themselves out. But I am still, my my only and one only concern in this whole business is that whether or not government or some other entities are going to start to ration AI, right? Because at the end of the day, if I have a token of AI and I can decide I can cure cancer or I can write some code, right, we we're getting some artifacts. It the um is a not only a moral dilemma, but it's a social dilemma, right? Because we these data centers will be using energy. Stuff is finite. So, and of course, being in the open market that we are, I think the price and the cost will deliver the right components. So at some point, it may become too expensive to use agent, and it's better to use an engineer with co-pilot. So we need to look at this equation from many unknowns.

Alex Romanovich 26:38

Right. And that's a and that's a yeah, it's a very good point in terms of how AI is going to be monetized in the future. But I think what you're describing is really not development anymore. It's uh what we would call, you and I use that term uh quite a bit in the past. It's orchestration. Yes. And um uh is this the um is this the future of the AI first uh service model that uh software development companies are going to now deliver in terms of how they're gonna, you know, um uh handle the bulk of coding and the uh and the new um requirements and uh specs and you know the BRDs and PRDs and so forth and so on.

Igor Kruglyak 27:18

Well, this is an inevitable result, right, of introducing of technology. The uh the uh the aspect, if you remember, you and I were young weapons around 2000, right? When uh we went through this whole web bubble and you know, and pets.com work made it, but pets, I love pets.com did not make it. So the market will find the winner, and the market will dictate how this is gonna work. Variables that we are still not aware of. Right now, I think we're in a zero phase of this, right? The uh there the infrastructure is being built out. This is not the same as people running miles and miles of fiber, right, in hopes that somebody's gonna use that fiber, right? I mean, we know, but you know, that that process took years to pay back, right? Because, but now every mile of fiber that they've ever laid is being used because of the bandwidth, because of uh of the the way we consume uh news and video. We don't use the uh regular linear TV anymore, right? So the capacity, right now, the capacity to deliver is considerably less than our desires, right? So at some point we will we will have to see that curve change. And um, you know, we will have to adopt, we will have to teach our engineers. Um, you know, we may not need engineers in five years, but we will need business analysts. The last thing I think we we probably, you know, as we wrap up this segment, um, I used to say this, years and years ago I taught uh programming in COBOL. And people used to come to me and say, Well, Igor, what do you suggest? Should I learn CICS, VSum, DB2, you know, something else? And I always told them learn business analytics, learn to speak to business user customers. Because at the end of the day, they will change the technology, they will get rid of stuff, but at some point, they need to have that. I believe that would be the last mile, that will be the last leg. Because, you know, it's still in the our business users don't think technology, they again, going back to the previous example that I gave you, they think how to solve customer problems. They think about what stock to buy, they think about what what disease to cure, right? They're not going to be able to program the solutions no matter what tools we give them. We need to be able to translate what they need into some level of componentry and then connect that componentry together.

Alex Romanovich 29:59

I I totally agree with you. And as we wrap up, and I have a couple of more questions for you. I just wanted to make a comment that uh the the concept of uh building or using a method and then turning a method to the product, or or take take taking a business process process, excuse me, and turning that into a product, that is becoming almost the accepted path, if you will, in terms of how the products are going to be made, how the agents are going to be developed, right? As a matter of fact, you know, it's kind of interesting. I um I was playing with uh Claude 4.6 recently. I saw you. And um Yeah, and uh another product that works with Claude called Lovable. And I actually even built a small application called ReadinessIQ. Yes, I saw that. And the small application, and you know, a small application that I built was basically a method that I've used for the past 10, 15 years working with early stage companies. And now I had an opportunity to take that method and turn that into a product, which was amazing, right? And um the the I think that's what's going to happen. And that's what's going to happen with the IT services or any professional services organizations. That if they have the ability to capture the method, if they have the knowledge, they can now turn the method and knowledge into a product.

Igor Kruglyak 31:13

Well, there are a couple of observations in that, and I do not disagree with anything you said. The idea of a product, right, and this is what I used to again teach my young whippersnapper engineers, is that the product is all about discipline, right? And you know, and you in when when somebody thinks of a product, right, the first thing people do is put together a roadmap. The roadmap, right? And then because that defines the ability versus need versus delivery versus, again, ability to absorb what resources you need, when do you need them? How do you put them together? The other thing about is the 80-20 rule, right? And I always tell this, so basically, you need to get right 80% of the critical requirements and make sure that those requirements are done. And there is an there there is a thing in the industry that people uh kind of don't even want to talk about, the 20% never gets billed. Because by the time you get to finish with the 80%, there is a new 80%, right? And a lot of the things that were put into 20% never gets there, right? So the um when when the when we stopped making CDs, puppies, or some other delivering media, right, we created the fluidity of the product. I generally always been very wary of um CICD and uh what I would call DevOps, right? Because that's a very important aspect for mission-critical applications, right? But they don't really define the product structure. In the product structure, you have a major release, minor release, you have a certain level of discipline, right? And just because Alex is calling me and saying, oh my God, I can't press this button and get this, doesn't mean that I have to drop everything and go fix it. Now, obviously, if if that's button lands an airplane on the around way, of course, right? But the level of testing and the level of requirements management that we put in into the discipline really defines that. So in all the years that I've been involved in product, and I've been fortunate to deliver several, both as in uh kind of working in the product company, but what we did at Core Value, which later we kind of migrated to behold tech, is really outsource product development, right? Outsource product development is really bringing in the discipline and the structure, right? And eventually that too will be replaced by AI, right? You will go in and say, tell me how to build this product, and it or build me the product, right? You know, but until we get there, I think the us old timers, right, will still have a value in you know, we're we may be older, but we're uh, you know, we're very sharp, I think. One of the things I'm working on that I'm hoping, you know, so my birthday's coming up in two weeks, right? Is I am gonna be uh I was born in 1963 and I'm gonna be 63 years old. You were born in in the same age. So what I wanted to so I I I calling this the virtual millennia of my journey, right? And that spans everything from playing chess, you know, DESA, to delivering multimillion dollar applications for IQ via you know in the 90s. But there is a there is an interesting thing, and we were so f lucky that we actually lived this journey and went through this stuff.

Alex Romanovich 34:44

So I agree with you. That's the ability to compare, right, this movement from you know, it's it's almost a um a certain level of appreciation when we compare ourselves the Gen Xers and the baby boomers to uh some of the younger generations that did not have the appreciation of, you know, the uh the circular phones and the uh huge computer rooms, you know, data centers that would, you know, blocks, not uh, you know, smaller rooms and so forth.

Igor Kruglyak 35:15

I I'm looking to see if I have uh uh you know the uh paperclip. The uh paperclip and pencil was the best debugging tools I I was using when I started in IT, right?

Alex Romanovich 35:26

Well in sales they still use the analogy of sell me the pencil, right?

Igor Kruglyak 35:29

Right, exactly. Or have a Rolodex, right?

Alex Romanovich 35:32

How many kids now know what Rolodex is what what Rolodex is, Igor. As a big takeaway, as the um sort of as we wrap up this conversation, and we're definitely going to continue with other aspects of it. What is one truth? What is the one sort of prediction if you want to give us about the future of the IT industry, a future of the software development, uh that uh our listeners can take away from this conversation?

Igor Kruglyak 35:59

Well, if you don't like what you see, wait five minutes. You know, I mean that's uh it's like a weather predicting the future of IT is predicted the weather. I, you know, I I certainly, I don't know about you, I, you know, I think the uh LLM and all the other pieces of this kind of generative AI caught me a little bit by surprise. Right? I I I spent some time in NLM and my experience was not that great, right? You know, slow absorption, a lot of things. So it is hard to predict. But, you know, the only thing that I always say, if you can figure out how to build something faster, cheaper, better, right? That will be the way, right? And uh generally you can do two out of three, right? So you can do it faster and cheaper, but no, it won't be better. But if if if LLM or if AI or Agentec AI or all of these pieces allow us to get that equation, right? Faster, cheaper, better, it will win. All right. And uh we'll see, we'll we'll see how things go. The only other thing is is that I only have a couple of more years to worry about, and then after that I'll definitely be retired and playing golf.

Alex Romanovich 37:12

Great, great. Well, Igor, thank you for being with us today. Let's continue this conversation and we wish you a lot of luck with uh Prime and other products as you're developing them, taking them to market, and of course not forgetting to play golf uh in the middle of it. So thank you for being with us and thank you. Best of luck.

Igor Kruglyak 37:29

Alex, thank you so much.