Ukraine's Best Kept AI Secret: Building World-Class AI From a War Zone

Can see a trend today that a lot of people thinks that can just throw anything into the LLM and that will do the job. It's no always the case. >> This is a field so full of potential for innovation. And so tell me what you're doing in in AI. >> Yeah. So I was just going to say if you could give some advice to a a soloreneur or someone starting out, would you advise them to pick a technology or a tech stack to work in? You no longer need so much developers to do the same job as before. Uh and one proper developer who uses AI uh in a good way can accelerate so much that he or she can replace >> >> uh a couple of people. AI is it hype or help? Good day and welcome to the show.

I'm your host Pete Cooper and our topic is AI hype or help. Today's guest is Alexander. He's based in Ukraine with his company Crashoft. They're developing software using AI and helping companies get a real return on investment. So, let's get into it. Welcome, Alexander. Welcome to the show. How you doing today? >> Hi, Petia. Thank you very much. Thank you for having >> Yes. And you're coming in from Ukraine. >> Yes, that's right. That's right. >> Yeah. What's it like living there? I mean, I have to get this one out of the way. What's it like being there over the last few years? I'm assuming you've always been there. And has it been big changes for you? What's happened?

>> Yeah, I lived my whole life uh here in Ukraine. It's uh yeah, like my home. I'm feel pretty comfortable here. I love living here. Uh it is challenging. Uh yeah, obviously uh how it feels. Uh it's like a normal life with normal I mean challenges like health issues uh like uh money problems, work related stuff but on top of that you have a war with its own challenges uh like this constant uh let's say danger for your family for yourself uh yeah yeah this is how how I can describe it >> and whereabouts are you located east west north south >> I'm located on the east. I'm in Hark. Yeah, it's my uh native city. So, I'm living here right now. >> And that's quite close to the action.

HKE is is a town that often comes up in the news. Is that right? >> Uh unfortunately. Uh yeah, it is. Uh I mean it's it's fine. Uh here obviously on the western part it's more more or less uh quiet. Here. It's a bit dangerous in such cities as Harkfe in KE specifically, but uh I don't know, probably I get used to it, so no complaints. >> Right. Well, you and your people are very resilient. I'm a big champion of Ukraine's um efforts and I have known of Ukraine a little bit. I knew per people from Ukraine when I was in Australia. Very nice people um and very hardworking, very smart >> and also um a lot of people use um software developers from Ukraine um >> since I've been here.

I haven't um but I have many of my colleagues have >> um and so they're considered to be in high regard um software developers from Ukraine. Um, so do you think that's when we talk about software development, I guess we're talking about, you know, the job front. Do you feel that what's happened in Ukraine over the last four years has affected the job prospects? Um, has it made people less likely to use people from [clears throat] Ukraine? Affecting definitely uh it's considered uh some clients consider it as like a risk uh zone because obviously your employee can get uh killed uh or something some bad things could happen to them.

So yeah some companies uh try to avoid uh working with uh Ukraine. I remember when the war started uh a lot of uh clients have migrated from uh Ukraine to some Latin uh lat m uh countries such as Argentina, Brazil because they have um similar uh salary level. >> Uh yeah, and also for the US it was more beneficial since they have same or very close time zones uh to theirs. So yeah, we we saw some migrations but uh still uh a lot of uh companies, a lot of clients would love to work with people from Ukraine as they they have a good good reputation and >> yeah despite the the the war uh people still work and they do their uh their best to to their to do their job properly.

U so it's still in my opinion and I can see it a big demand on uh professionals from this area. >> Yeah. And Alexander, you're not the first uh podcast guest that we've had from Ukra from Ukraine and your country. Um and I I get the same impression from him as I do from you. Um you guys are definitely in the forefront of your field. Um highquality work is being created, innovation and um we only have to look at what's happening with the war to to to know how in innovative Ukraine has been under such difficult circumstances. >> And I think uh as we go into the AI, so let's start talking about AI. Um this is a field so full of potential for innovation and so tell me what you're doing in in AI.

Um in AI. So we um what what my company does with so let me uh maybe uh tell a few words about what we do in general and what AI uh what part AI uh is our is in our business. So we uh build software we are custom software software development company we act as a partner uh not a vendor. So we embed into customers business and try to understand what they actually do uh in order to drive uh proper and meaningful outcomes outcomes with uh software. So not just build something but build a part of the business and understand what uh what what it does what what value it brings at what responsibility it has.

Uh uh so yeah this is what we do and uh AI part um we uh do some uh development on top of the AI uh with lang large language models uh yeah we mainly build internal tools. So we uh embed AI into uh internal workflows of companies. Uh we do some uh client-f facing applications for example software as a service projects but mainly we work with uh internal uh and business uh software. So most businesses would understand some pretty simple metrics like so you want to drive revenues up, you want to drive your costs down. Um you want to maybe mitigate some risks. Um if you can translate what you're doing in software into those simple metrics, then every business is willing to go down that path.

Is that right? Is that what you're is is are are you feeling that momentum come through from your clients that they want those metrics and that AI is a way to to make that possible? [snorts] >> Uh yeah, first of all they they want uh since we work closely with internal processes, they want them get get right because they feel the pain when something is inefficient. They feel uh they are losing money or they leaving money on the table. They are not earning enough. So the first thing yeah they want from us is to make this right. Yeah.

And then uh they also want to understand how AI can uh can help them to uh yeah be it increase their revenue uh or or bring the cost down or just uh make uh their people more efficient so they don't spend a lot of time on uh stupid stuff and uh yeah automate as much as possible. Yeah, automations.

That's ex exactly [clears throat] where I thought we would land in this conversation because there are a lot of I guess routine um you can't it's hard to completely replace a person and I don't think that's what we're advocating here but we certainly can find those areas of work that are somewhat repetitious and that couldn't have necessarily been done without AI but now with AI they can be done right do you have some cases or without telling us specifics, you know, we don't need to know customer names, but something you've worked on in automation, you give us a bit of an insight. >> Yeah, like the first uh the first thing that ca came into my mind, it was one of our first uh projects.

Uh so the customer was uh an e-commerce business. They have like lots and lots of uh stores on different location. So they sell one brand in different countries or stores for different type of uh products and they have like uh a big mess you know they have multiple Shopify stores, multiple Amazon stores. If we talk about uh u their US part of the business it was really really messy. Uh they have like um a table a Google uh spreadsheet where they were tracking all their incoming orders and they were managing it through the um spreadsheet uh like the status uh uh even uh some leaving some notes for customers to send back to customers.

Yeah, it was very very inefficient and they have a a um crew of four people managing this since they have like 11 or something stores. Uh yeah, we just it was one of our first projects as as I mentioned. We managed to to to unify this to bring all the orders into one system. Uh so they can easily track from uh track it from one place and yeah the work workload uh reduced from four four people uh into one person and the rest people they didn't fire it they just reallocated them to other parts of the business. Uh like uh customer support uh warehousing or stuff like this. Um yeah. >> Yeah, that's a great um example.

So um it's interesting just the fact that there was a company that has so many stores um and you mentioned Shopify and Amazon, maybe there's others uh maybe just native websites. Um but so that this just to understand that business, they they are e-commerce. So they're shipping product um >> out through through I guess multiple products out through multiple stores and so the the complexity can go up very quickly, right? It's not like I've got a coffee shop and I'm selling coffee. I've got one physical store. I get the the goods come in through the back and I deal with the customers through the front and I sit in the middle. Right. That >> is easy to understand.

But once you have so many products going into so many stores, you end up with this kind of mesh of >> complicated things. Um, and you've been able to use AI to solve that problem. Can you tell us a little bit more about how you used AI to solve that problem? >> Uh, yeah. F first of all this uh this pro project uh uh the the main I mean the main uh problem for this customer have been solved even without AI here. Uh so this uh managing part uh have been done uh like a traditional software uh they they stuff didn't uh they they no longer need to do it manually everything is automated.

If if if we talk about AI part in this particular uh project uh it came later in the form of a uh of a chatbot uh pretty straightforward a chatbot for staff members where they can ask specific question about uh their order databases statuses and so on. So basically it was a uh it was um um an easy interface for them to quer their their uh pretty large they had like hundreds of thousands of orders database uh in neutral uh language in neutral conversation. Yeah, this part came later if you want. >> Did I not sure I got that? So was the chatbot something the customer interacts with or was it the staff within the company?

No, it was a stuff where they could uh ask a specific questions about their orders stuff uh like bring a history uh bring some uh for example conversation additional context like conversation with a customer in WhatsApp for example to to see how things are going why for example this um this order has been cancelled uh yeah stuff like this we have also a similar uh down the road we have a pretty similar uh project.

Um but for vet clinics we help them automate uh purchasing of their inventory because uh in vet you you need to purchase stuff uh like on weekly basis like every week a specific person supposed to buy certain stuff like medicine tools uh office supply and so a very repetitive uh let's say process. You buy stuff uh you you make [clears throat] uh almost identical purchases every single um single uh week.

You there is some pattern and we have to we we help them to develop we use some machine learning to develop uh a predictable model that will predict what stuff they might need to buy in the specific week uh based on the his purchase history of the previous like uh two years as far as I remember. Mhm. Okay. Interesting. So, on to the on the chatbot side of things. Um, we know chat bots. Most people know chat bots by chat GPT. You're typing something in and it's giving you a response. In this case, the chatbot is again, you can type or you I guess you can talk into it and then it's going and searching all of the databases to try and as best it can answer the question that you presented.

Um in the traditional way you would have to go and search each of the different uh databases, stores, products until you find the piece of information that you were added to and then synthesize up the solution yourself. So very very powerful in the sense that it can go out and drag right a lot of information in. Um and if you're using voice it's even quicker. So I can see how you can get the leverage there on that one. >> Yes. I I did I describe it well. Is that is that what you've done? >> Yeah. Uh this is a chatbot that help query I can if needed I can get some technical details how it was working under the hood but yeah essentially it was doing exactly what you have described. >> Yeah.

Okay. Yeah. I bring it up because it's very relevant to me. We did a project uh earlier this year which was very similar, very complex in terms of retrieving the information from different portals and different places and different databases and then presenting it in a way that was easy for us for the staff to manage. And that saved >> um one full-time resource thous $5, 000 a month. Um so we know there's real impact to doing that automation and that complex search and presentation of the data. We didn't use a chatbot in that case. We did it um using other tools, but a chatbot is a really nice interface, too. And I think we're all very familiar with that.

And I like chat bots because they meet me where I'm at. Like if I have a question, I present it the question to that. It doesn't it doesn't demand of me that that I follow a set procedure. It just says, "Okay, that's your question. I'll do my best to answer it. " Um and I find that quite appealing and very satisfying to work with. Very quick as well. >> Yeah, I agree with you on >> Yeah. And then the other side was the predictive side.

Um Y >> which is also I mean >> this one I think could be done more I guess mathematically right like you can just look back at the statistics and just uh you know say statistically you're more likely to need more >> Christmas cards around December you know because that's what happens. Um but AI can go further. How does AI make that work better than just logic? >> It was actually uh uh the predict uh the level of prediction was built as you mentioned uh on on statistics. There are like specific let's say Python libraries that can be used to to build machine learning models that will predict uh certain things based on uh previous uh events.

How AI uh uh helped here is to make this more uh uh like human readable. So the machine learning model was returning some uh let's say JSON or whatever else format that is not really you know human readable and you need to be technical and you need to understand how to read it properly. AI help to pre uh present this information to uh users in a proper manner. So for example uh clinic staff they don't have to they have a clear picture of why this uh system suggests certain things to buy. >> Uh so yeah AI was like a layer of uh uh trans transcripting it into yeah more human readable manner. So nontechnical people can easily and quickly understand what is suggested, why? What is the reason?

Because uh we are dealing with uh purchases. It's first of all uh money. If we buy redundant stuff, we'll lose money. If we don't buy enough stuff, we might run into problems. Uh so yeah uh this layer of making uh data more human readable is very important. It might sounds like easy, simple, just like uh uh I don't know return some strings, but it's an essential part of of a business >> because people to to run it to read it really quick and understand what's going on and why it's suggesting certain things. >> I got a few examples how I have solved multiple new problems with AI which traditionally were a real real pain. Um so for example um I have to reconcile Quickbooks in my business.

So as Quickbooks is a soft financial software package. >> I have to reconcile the books regularly and it can be quite tricky to know exactly what to do. You can make mistakes which can really make a big mess. And I've done that and I've had to employ people to come and fix up all these things and it's costing me hundreds if not thousands of dollars to make those changes and and there's a ripple effect that goes through. So, what I would what I'm doing now is I take screenshots of it as I work through the problem and I upload that into an LLM and I tell it this is what I'm seeing.

Help me do this and and and it's really really good at taking that and suggesting how you go next and how you go next. And I think nine 95% of the time it's taking me in the right direction. Occasionally it doesn't and I can catch it. So, you still need to keep your head on, but wow, was it ever powerful at looking at a a interface, for example, QuickBooks, and guiding me through it. And I'm going to broaden that problem because I would probably have a hundred interfaces to different software packages in my business. And I say this to my team. I say, I can't understand all those interfaces. They're notoriously they're changing. They're complex.

The setting buttons and the users and all [snorts] this stuff, you know, just goes on and on. Can't remember them all. I would have I would just I would be full-time keeping up to to speed with all the interfaces. Now, I get my team to do a lot of that. They help me and they say, "Okay, Pete, you just have to go this and this. " But also, I now have AI when I'm stuck with something like QuickBooks, which I don't want to reveal to everybody what I'm doing with the finances, then I can go to my LLM and I can get that to help me uh understand the interface.

Now why I think that's so relevant why I'm bringing I'm drilling into this is because AI is a fundamentally trained on all of the data that's in there on the internet and there is a lot of interfaces to different um uh software packages and tools and so on um which AI knows about right so it's a it's good application of AI because it's in AI's world right it's trained in that world um and I I think this is where I want to go with this conversation. There's also areas where it AI doesn't work. There's areas in the real world which it doesn't know anything about.

And so picking how AI has the biggest impact is also a part of what you do I think Alexander is you work with your clients to say well look I can solve that problem with AI or or maybe you go I can solve that problem with logic and software. Now with AI can I can solve these problems but there's still going to be those problems that you know are beyond what AI can effectively do >> then you would guide them to that and do is that part of your engagement with with your clients >> yeah definitely uh I can see a trend uh today that a lot of people thinks that thing that you can just throw anything into the uh LLM and it will do the job it's no uh always the case.

Uh and yeah, definitely we uh I'm a developer, yeah, myself and when I'm talking to a customer, I can u pretty much every time describe if this particular problem requires some LLM uh to to solve it or or not. Uh yeah, there is a I would say a pattern and I see to determine if the uh task requires some AI or it can be caught uh in a traditional way. >> Yeah. Yeah. Well, that's well said. And you you said you had a software background. I mean >> you must have seen what's happened. How did with AI coming on you've seen what's happened in the industry um like I have.

I think it was last year there was a huge layoff >> uh particularly in India they were hit really hard because it's kind of like a lot of software developers there. >> Have you seen a lot of people being laid off in as a software development world? >> Yeah. Yeah, I do. Uh we we also have a pretty in Ukraine I mean we also have a pretty large uh community of uh IT specialists and I can see the the market the number of uh job positions is getting uh lower. It's not dropping but it's lower and the number of applicants is is just huge.

Especially if we talk about some very common tech uh like ReactJS, there are like uh numerous uh like hundreds if not thousands probably thousands of developers uh available uh for the react and it's super competitive for this type of uh job positions. Yeah, I can see uh I can see the same thing you are seeing. >> Yeah. >> So why do you think that's happened? Um it's uh f first of all um in in general in world we see uh some sort of like a crisis. We have a war in Ukraine. We have a conflict on the Middle East. Uh we have a Donald Trump doing some crazy uh stuff every week. Uh so there is some unpredictability in the world.

Uh the dollar is the uh the currency the dollar is not really stable. Uh US and China have something to to uh to conflict about. Yeah. Argue about to conflict about and in general um yeah and >> companies are less uh willing to spend money on software. This is like the first uh uh the first uh reason I see why it's happening. And the second one is React. Yeah, obviously sorry React AI. You no longer need so much developers to do the same job as before. Uh and one proper developer who uses AI uh in a good way can accelerate so much that he or she can replace uh a couple of people.

This is like the second uh >> yep >> second thing that lead us there let's say >> yeah that that's exactly what I I I feel is happening um and particularly on the AI side of things I've witnessed it myself I have a a really good young developer he embraces AI uses it has been for a long time he's extremely productive and I don't need a team right because >> not only do you have to get multiple people back in before AI No, let's say I need three people, but then I need a lead for that. So, I need a fourth person, like a lead to to hold the whole team together and coordinate them and make sure that they're working together. So, I now I have all this overhead.

Now, I've got one guy and I work directly with him and then he's product very very productive with AI um because he knows how to use it and use it really well. So we've seen this is very very disruptive to as you say all you know people particularly in ReactJS because now AI is kind of take come along and taken that job. Um >> but is it swinging back do you think? I mean I heard something that said uh they're bringing back a lot of the people that they've fired. >> True.

Uh true because uh and I can see uh like last year uh I would say around October, November uh in my opinion the uh this um I would say the huge hype around uh AI that AI now can replace development developers started and it uh lasted up until I don't know mid uh spring this year. Uh and then people yeah re realize that uh wipe coding is not I mean solution to the to their problems. You still need to to know how to build software. You you you might not need to know how to write uh the code but you have to know how to build software which is different from the writing code. >> And yeah they realized and they realized that systems they have built are vulnerable. They're not reliable.

They have a lot of security issues, uh, breaches and so on so forth. And yeah, I also noticed that people started to companies started to hire back some people. >> Yeah. To to save them. >> I I and I agree with you.

I think um I mean we got this great tool called AI and I think it's it's only natural we listen to the hype sometimes and it becomes you know we expect a lot of it and look let's face it when there's big leaders who run these you know half a dozen companies are investing so much into it and they're telling us it's going to solve all of our problems you know it's hard not to be influenced by that you know the money's going that way the leaders are going that way okay I got to follow that because I've got my own business right >> um but I think we be a little smarter than that. And this is what I've learned. I want to test this with you, Alexander. Vibe coding is useful.

It's useful if you want to get a prototype out. Like if I want to experiment with a user interface and I want I've got an idea and I want to try something. I I think vibe coding is valid. Like you you get something prototyped, you can demonstrate it, you can move the buttons, you can scribe it and on you go and you can knock yourself out on that. I think the problem comes when you try to say, "Okay, I've got a product. " Now, I've had multiple clients say to me, "We've got the product. It's ready. " And they say, "You've got a prototype. " >> Yeah. >> You need to turn it into a product. And the gap between prototype and product actually hasn't changed that much. It's still there.

I'm think AI can help, but you still have to make that jump. >> Now, that prototyping stuff can be done inhouse. It can be done literally. You can do it on a weekend if you wanted to knock yourself out. A business owner could do it themselves and they can cut the costs and the time down for that. But to go from there to there, um, that hasn't changed much. And I think the the if somebody that's just new to this, not understanding software development, thinks that they've got the product when they've just vibe coded a prototype. >> Yeah. And uh I uh on you with this 100% agree.

Um it creates uh a problem uh that uh let's say a business or a person just uh built a prototype or as they think a almost completed app they purchase like I don't know a lovable subscription for $20 and they came to you and ask like hey help us bring this project over the finish line and get it live and you give it a proper uh quote uh and tell how much effort you will need, how much time, what the budget is. Uh they they are super confused. They uh like spend $20 on building everything as they think. And you are quoting like 1, 000 time more than this uh to to to finish last pieces.

And yeah, the it requires a lot of conversation to explain what uh to customers what they actually build and what they actually want and what the re real effort it is. So yeah, I'm I'm agree with you on this and I'm always say to those people that uh AI is just a tool. You you you're not developer. Having a tool doesn't make you a developer. If I buy, let's say, a Canon camera for like $2, 000, it won't make me a photographer. I still don't know how to to take pictures, even though I have a great tool for this. Same with your loable subscription or cloth code or whatever. >> So, yeah. >> So, I I I was successful in convincing a founder using this analogy.

I said to them, you might remember that Elon Musk said he had this Roadster that the new Roadster which had three engines would do 250 mph and he brought the prototype out onto the stage and it was beautiful. It was red and then he says, "We're going to build this thing. " That was six years ago and we still do not have a production version of that prototype. That's because making the prototype look pretty and drive onto the stage is very very easy compared to actually going into production with [gasps] >> and I think finally so they they get it. Okay, there's a lot more behind there that I don't understand. That's why I'm employing you guys. >> Um but I think you're right.

I think the price can be a bit of what we call a price a sticker sha shake a sticker shock you know where they see the sticker price and they go oh my goodness I wasn't expecting [clears throat] that I coded it up on the weekend I thought you were going to say a th00and you're saying 10, 000 or whatever you know >> y >> yeah that's that's a difficulty we have when we're developing things for our clients I agree with you there but um to switch gears a little bit how did you get into this you you were a a software devel developer, right? So, you came out of college as a qualified professional developer. Yep. >> Yeah. I I had a bachelor and master degree in computer engineering. Yeah.

I have been working uh I I believe since second year of college. Uh I really liked uh coding. Uh so I started to work earlier. Uh yeah. But eventually I always I realized that I always wanted to do some sort of my thing like my company and uh yeah like 3 years ago uh and opportunity came uh uh like I had some career switch and I decided not to find another job but I try to to build my own company. Yeah. And since then, yeah, I'm doing doing it and yeah, so far happy with the result and probably was one of the best decision of my my life to to go down this uh road. Yeah. >> And how did you build the business? How did you get the business? How did you win clients?

>> I believe that uh what helped us is this uh narrow um narrow focus. Uh we first start as uh um we mainly work with one particular technology uh a lot and uh yeah we were kind of known in this ecosystem and yeah we get a lot of uh clients uh who who were searching for professionals in this uh with this technologies.

It was pretty easy to to to win clients and to build the first client base let's say using this narrow focus and after this we we started to to expand and focus more on uh uh not technology because it's really hard to scale when you are working inside one technology but we started to work in specific industries like mainly healthcare and e-commerce and we try to work in these two indust industries. We we have some other we we work with but these two are our main and we we try to focus our effort on these two uh industries >> as we have experience we we can talk same language as our customers in this business and we yeah try to uh find clients there. >> Makes sense.

So what was the technology if you can share with that that that [clears throat] you you focused on? Uh yeah, it was uh retool. Uh it's a low code uh platform. It it allows uh to build it's a like a visual drag and drop builder where you drag some elements and you can write custom logic behind them. So you can uh write the whole um whole app uh yeah and simplify it by using some pre-built uh components. >> Uh yeah, it's called uh retool. It's used only for uh internal tools. That is why we that is how we gain our experience in internal tools specifically from from this tool. >> So you became known as the guys the go-to guys when it came to that retool. >> Yeah. Yeah.

Kind of this we we we are a partner of this tool. U they sometimes [clears throat] share a very good leads with us very qualified great customers. Uh yeah. >> Oh great. So actually the um that uh retool company that created that tool was actually lead generating for you and giving you referrals. Yeah. >> Yeah. They they gave us some good clients. Uh yeah uh since AI um got so much better with uh Korean I I have to admit uh it's no not much use for for this tool anymore. Uh that is why we're now focusing on on custom development. Yeah, we always had like uh custom development. We use uh Typescript stack and Python stack. Uh but it was like a minor part of the business.

Now it's like the major part. >> Yeah. So, I was just going to say if you could give some advice to a a soloreneur or someone starting out, would you advise them to pick a technology or a tech stack to work in and build those relationships with the supplier of that tool? Would you do that again? >> Probably.

Uh but uh I would uh say uh not uh don't um um particular text tech particular technologies if if you're starting just pick a narrow niche it can be a pretty let's say occupied service for example software development but if you can do it for a very narrow uh market uh I don't know >> like you a person lives in Brazil and you you are knowledgeable in you you you're good at writing quote for uh accountants in Brazil just work with them grow your customer base uh with this narrow market and then expand.

This is what I uh would suggest and not it hasn't be it don't necessarily need to be a specific tech but again a specific industry a specific type of clients yeah something narrow and then expand >> and maybe even a geography a specific >> geography for sure >> I mean I think there's certainly here in the US there's um a sense that people in the local town help each other with their local business, right? So, having that local element, um I know a guy that started a few years ago literally just set up business and said, "I'm just going to focus in this area and away he went. " He was doing um u marketing work for uh for for local businesses.

So, being local, I think, you know, if you say, you kind of said it in a way that said if you're in Brazil, then focus on Brazil, right? Mhm. >> Um, that can really help and I would definitely appreciate that. And then I think the other >> way you niched it is you said well pick an pick a say accountant. So you get to know their um vocabulary, their needs um and one accountant knows another accountant. So he's going to refer you this these for me. Do you want them to do the same? You know >> makes just don't try. When I started and I quickly realized it was the wrong strategy. Don't try to be the guy who does uh everything.

Uh be as much as specific as you as possible in your services and the value you provide. Yeah. >> In the beginning and then expand. >> Yeah. That's a very good advice. Don't try to be the guy that does everything because you master of >> jack of all trades, master of none is the press is the thing we say over here. So, how long have you been running your business now, Alexander? Yeah, it will be three years uh next month. >> Oh, so fairly fresh. And um in that time, how many people do you have working in your company now? >> Currently there are 19 people uh in the company we we had uh like u 25 or 26 at most. We also had some layoffs. Uh but yeah, now it's 18. >> Uh yeah.

>> So so let's say 20 20 odd people. Um, and they all work for you and your company in Ukraine. >> Yeah, they all Ukrainians. Some of them, a few of them actually are located abroad uh because of the war, but they are all Ukrainians. Yes. >> Okay. >> We had some uh foreign workers uh but uh I realized that it's much more easier to work with people from your area. You speak language, you share same values. Let's say you have same mentality. It's It's just easier to build the team in my opinion. >> Do you do you speak internally uh in Ukrainian or in English? >> Uh Ukrainian. Yeah. >> Mhm. Yeah. Well, that would definitely be an advantage to be Ukrainian if you speak Ukrainian in your company.

Mind you, your English is very good. >> Uh I think everybody I've met from Ukraine speaks very good English. >> Um no problem. >> All right. Well, I think we have explored some interesting stuff. Alexander, is any comments that you'd like to make because we're running out of time. Any questions you might have for me? For example, [snorts] >> in general, maybe you can uh I don't know, I shared what problems we try to solve to uh to our clients like what about you? Uh like >> so yeah, in a similar field to you, we we've focused on automations. Um and we did a we did we we wrote this uh product we call it reconnecter and what it does is it reconciles insurance uh for healthcare clinics.

So um they see their money coming in through financial package called zero and their customers come in through uh another package called clinico and the two don't meet and then the the way they get money in comes in through emails and so on.

So there was these three sources of information which was very difficult to try and reconcile uh and we used AI in a very similar way to you did and that is to go and search for all those things and bring them together and present that data and that cut the time down enormously um from as I said it before there was one full-time offshore person doing that was no longer needed and the bookkeeper was doing instead of five days down to three um and that saved that business $5, 000 a month every month month on month and month on month.

So >> um that reconnecter is a product we're looking at um actually rolling out um in Australia and here in the US um and we're currently seeking funding here in the US and we're seeking you know support in Australia as well because we see this problem um has been there. Real pain point and it's um if we have the solution people will use it. So we're going to see if we can do it and I think the relevance here in AI is that solving that problem was very difficult. I mean you can solve a problem without AI back then but it would be too expensive be too difficult.

>> So it's been pushed along you know and now we got to the point with AI it's made it cost effective to do that and that's the point. The point is that AI has made that solution possible economically. And I think that's where we are with AI to me. It's it's not about whether you can do it or not. It's whether you can do it economically. Okay. Do does the do the numbers work out in and give you a return on investment. >> Clear. And I remember when we first met, you mentioned that you are from Australia but live in the New York. >> So you know know the both markets. Uh >> well, I'm going to say the health sector is not my strength. Um I'm a real tech guy.

Um like like you, I mean, I was an engineer, not a software developer, but I've been software developing with my team for as long as I can remember. Um knowing the ins and outs of this as the of the health markets in Australia and here is has been a big learning curve and I'm nowhere near uh the top of that. [gasps] But yeah, knowing um enough to be dangerous in the two markets is a start. [gasps] [snorts] >> Cool. It's a good market to be in. Yeah. >> Yeah, that's right. Well, I got one more question for you, Alexander. >> AI, is it hype or help? >> Uh, it depends how to uh how you consider it and for what you're using it. Uh let's start maybe with hype.

Uh we we already uh touched this uh a little bit when we uh talked about people started to think that AI now can replace um developers for example which is not true because as I mentioned you you you can um use AI to write the code but you you still need to know how to write a software. So it's a big hype that AI can completely replace a specific uh person in your team. And I was guilty of this as well. And when I first uh work when I first touched the clo code, sorry, the clo design, I was really impressed by how much I can do with like one prompt. How result is awesome. It looks great. It's just exactly what I needed and even better.

But when I first uh worked with a design uh generated by clo design, I quickly realized that there are some things that designer wouldn't do wouldn't would do bare. And also what I realized that if I need to make changes, a specific change, it's really hard to to make it with prompt only. You still have you know you you still might need to do it manually via Figma for example. And yeah, this is where your knowledge of design uh is required. So yeah, it's it's a big hype when people uh think that they can replace someone with with an AI.

Where it's a real help is when a professional, a developer, a designer, an accountant, a uh I don't know a sales manager, when a specific professional use AI to accelerate uh the job. There are parts in any job that that are time consuming that people hate uh that can be easily done by AI and if a person who knows what he he's doing uses AI it's a huge help yeah no no secret you you can witness yourself how developers can now build uh code how quickly it is without sacrificing quality >> if person again know how to write the code and know how to build the software so it can guide the AI to do the uh proper things.

So yeah, AI is a big help if you consider it as your assistant that does exactly what you ask, knows a lot but needs proper instructions. >> Yeah, well said Alexander. Well said. It's it can be hype and it can be helped and so how you use it. Well, I think that's a really good point to end on. Thank you so much for being on the podcast and I think we could have you back. I mean you've got a lot of stuff. I'd like to know more about your cases and other things that you're working on. >> We are working in a similar area. We have facing the same challenges and I know a lot of people out there are in a similar boat as well. So thanks again Alexander. >> Yeah. Thank you. >> Yeah. See you.

Thank you. Bye. >> Well that was great talking to Alexander and finding out what it's like living in HEV on the eastern side of the Ukraine. We talked about a couple of his projects that give real return on investment to his clients. One of them was an e-commerce um company that had multiple sites, you know, Shopify website um Amazon sites and lots of product flowing through. So, you've got lots of complexity. He was able to solve a lot of that complexity um and make it much easier for them to search and figure out problems and deal with customer issues. In fact, it was so successful that he was able to redeploy three out of the four people that were doing that work.

Only one person was doing it. Now, good news is that they didn't lose their jobs. They were redeployed into other areas. But also, he was doing an automation that um made it easier to predict um using AI what needed to be ordered. Now, this could be done logically um but it can also be done with AI to enhance it. So, you can bring in more factors and more parameters more easily with AI.

We talked a little bit about how AI has really affected the software development industry and how many people have been pushed out of work, but surprisingly they're coming back because a lot of people are responding to the hype initially that AI is going to solve all of our software development needs, but it's not actually the case. It's good for prototyping. It's not good for creating a robust production product. So people have been, I guess, listening to the hype a bit too much. Now they're using AI a little bit more appropriately, especially getting the leverage out from where it really helps.

Now, if you want to know more about Alexander and his company, follow the links in the description. If you want to know more about Skillian AI Labs and what we do, please do the same. I hope that helps and thanks for watching.