Hercules.ai: Enterprise AI Automation, LLMs & Fortune 500 Innovation | Deconstructed VC

wear in the air plane is made out of dried paper and NCO is like hm what are we doing about AI wrong here wrong here wrong here why you doing this as Elon Musk said it's like staring into abis and chewing glass instead of year it can be done in four weeks to do crazy the next 3 years everything is going to change it's it's it's amazing time we're living in hey guys welcome to another episode of deconstructed VC Nick Marina we are co-founders of David of venture capital a community of 100 engineers and Founders investing in early stage startups and this show is about deconstructing businesses trying to understand how they built what they consist of and how Founders build them we today we have Alex the founder of zero system we were lucky to become investors recently that's not the first round of funding this is third round of funding for the company and the company is doing pretty well guys so today we want to learn how can we learn from you we want to first of all ask you to tell the story how did you come up with an idea how did you decide to start this business first of all thank you for having me here it's pleasure and honor happy to tell the story of zero system how we started and where we are right now and where we going and there are a lot of pitfalls and ups and downs like every startup has it's normal so people should know it's no bad of roses it's always as Elon Musk said it's like staring into abis and chewing glass people have to be comfortable chewing glass in the morning instead of a breakfast so that's that's how

it feels zero started about actually about eight years ago with a with an idea my co-founder and I had anide idea of summarizing long texts into short consumable form it sounds easy right now post llm era right like what what's what's can what can be easier is throw in text into llm and make a summary or extract the actionable items but eight years ago pre- llm it was a lot of work done and we we've been a very small team for pretty long time just doing that research and trying to figure out how to build something that never existed before because there were so many teams in a big institutions and in universities and in big companies as well like Microsoft and Google trying to address the same problem and they failed like how can you guys in a garage like four of you address the same problem and we decided to do it from a different angle instead of doing the generalization of like we can summarize everything for everyone we decided to go vertical by vertical and summ try to summarize and build models to summarize specific parts of the the text because context is important to make sure that you provide them highest possible quality and that's how we started building models of ours so by we who do you mean so myself and my co-founder gor and I keep saying he is the smart I'm the beautiful and that's how we work and gorg's previous experience was actually in a cancer image detection skin cancer image detection a lot of experience in that and it was pre again pre- llm and a lot of stuff has been done on medical devices and when we started the company he said you know

what we're not going to be using Cloud it was very untraditional at that moment everyone was in the cloud and still are like we don't want to spend a lot of money on cloud processing let's do it on device let's use clients and users own devices Edge Computing basically to to build them up models it was basically adding a next level of complexity because it was hard to build a model to summarize thing and extract actionable items from long form but building it so it can run on devices on H Computing devices was next level of nuts as some of the investors were telling us when we're explaining what we're doing but actually paid off amazingly well so gorg kitan my co-founder and CTO and myself and a small group of people we embarked on that journey and and during those years the idea stayed the same but the product line and the technology around it evolved into much much more so going back so you say a small group of people when did you hire your first employee I think it was probably around three months into the journey when we realized so what was the first vertical you picked so actually first vertical was not the industry vertical it was vertical of content so we decided to summarize long forms of email okay and we needed user rein like reinforcement learning people voting up and down that's that's the big problem for every Ian I don't think that that was how you solved the problem I think you just kill people who write long emails that was there was a plan B for us luckily we didn't didn't go that path but initially we wanted to make sure that the email is summarized

and people either satisfied with a with the form or not and the more than them up V up or down and we couldn't figure out how to do that and then was an idea how about we build a small AI application like triage type of application so all the emails coming in we summarize it extract the actionable items so people can understand and they swipe left or right like Tinder style but if you they swipe right the email is deleted if they swipe left the email is kept in the inbox for it extra extra a pretty cool idea and paid off nicely so people loved it we got got in a matter of probably 3 4 months we got to 100,000 users active users people went crazy about it it was a mobile email application and remember those models were working on a mobile device and it was still pre Apple's chips that were able to process that's like 2017 yeah it was 2016 17 so did they pay for it no it was free because we needed this community to actually reinforce the the models to to train the models but at some point we realized that the model model Works and works pretty nicely and at that moment we realized now we need to monetize it so we had a minimal Bible model not the product yet but the model it worked and we started looking into Industries and not just looking into Industries what industry would buy a product like this or will embed it into their process but what industry will serve as a beach Head Market because you don't want to stick with us it's a dilemma you want to focus on a small enough industry like or Head Market

so you can win it all and you don't have competitors to rub shoulders with but on the on the other hand you want to have that head as a jump start Leap Frog to much bigger Market later that's the probably the biggest dilemma every startup can can can face and we found that industry actually was again very counterintuitive remember I was telling you before VCS were like oh that's that's not and when we're coming back to them and saying hey we want to do that inside legal market like for largest law firms in the world they were paus and then saying that's the next level of nuts so it was literally counterintuitive because law firms they are very conservative they don't want to buy anything that is not close to 100% polished try to sell something speaking with the Russian accent in 20189 especially AI related email management system good luck but there were some we need your data it is going to be safe with us right here in my pocket almost this way but there were also incredible parameters of that market that were overshadowing everything else so number one no one was stupid and brave enough to actually even try so there were no competitors literally you just go in and everyone was scared of the market so it was blue ocean full of sharks of course because that's that's that's what it is but it was a blue ocean so you can focus on making the product better serving your clients better instead of competition and lowering the price so we immediately put the price like probably five times that we can do with any other market and it paid off that is super interesting but there is interesting component the

final one data because we realized that to go to other verticals we needed data and legal it's basically it's a serviceable market they work with contracts emails everything Communications and they serve every other industry and that's what the typically is being talked about right now is a data mode you can get access to the beautiful data you can train better models and it's like a flywheel you get better products better models and it goes up that was the reason we went there and it paid off we're how did you decide on the pricing model like so you it's I think it's a very interesting detail that you were able to charge more how did you get brave enough to you know suggest bigger prices and how did you set on a number there is so many techniques how to do that we found the wrong one but it worked you basically set up the price as high as possible and see if the client's kicking you out of the boardroom if they are you're too too high and you just go down until they're like okay let's talk about it so then you see like what The Sweet Spot is but also it's about the value of of course the ROI a lot of products and I've seen that in in in other Industries as well a lot of products they undervalue themselves by not delivering like not explaining what the ri is and not helping clients to see the ROI in the first day so for three years our main slogan like main Moto when we were talking to clients was not about the product was not about the features was not about what we're doing you know what it was Roi on the

next day mhm if you are an Enterprise client and someone like a startup comes to you and saying we'll talk about the product we'll talk about everything want to show you that there is an Roi next day since you buy the prodct once you buy the product that immediately Sparks your interest and they're like okay how you do that because Enterprise clients they know that it takes six months to deploy and like four months to buy mightbe eight months to buy and it's like RI somewhere down the road but if you tell them it is here it's right away like here that that makes them interesting how much did you charge on average per contract if you remember oh of course so and the pricing didn't change since then it's basically the same we typically charge and we if we talk about the large watch Enterprises we charge per application right now we started with one now we have eight it's big basically an App Store App Center of generative AI applications it's a different story we'll talk about that but it's about $300,000 a year but here is interesting thing because unlike other AI companies that has to run their own servers or pay arm and lag to Microsoft AWS Nvidia remember what we talk about our Solutions work inside security periment of the clients we don't have any cost to support it they pay their own Electric bills exactly our margins are 95% plus and scalable because that's on the client side how fine are you talking about 95% profit margin while your clients are listening to this absolutely fine because remember it's AI on the second day that is true how much money do they save so we have one client in

the insurance industry they actually saving $3 billion do a year do you care you're paying us $300,000 if you save billion I think we can like I would I would pay you 30 mil for three Bill no you will not actually you know why because that type of clients they have their own it they have their own Engineers their it budget is a billion dollars a year and they have 5,000 people in it if the price is more than they feel reasonable to build themselves trust me they'll buy a company or they build it themselves it's important to actually scale internally rather than try to charge for one thing but on enormous amount of price tag it doesn't work this way it's not flexible we always were we were always building a pretty long-term strategy like literally long term we knew that something new might might appear but we needed to have like five years plan in where we want to go it was not very detailed you can't look Beyond the Horizon five years for startups it's eternity right but we were building big like large brush Strokes where we want to get and at some point we realized that if we want to have a second application and the third and the fourth and we wanted to be basically an app center there was no generate AI name for that yet but our models were basically large language models starting 201920 but we realized we need to build our own tools to actually make those applications faster and better and easier and we preemptively started building the engine we called it Hercules and the engine basically is a set of building blocks like Legos like the Legos you have here in front big

big Legos based on which you can assemble a new AI application quickly and in order to explain how it works I want to use the historical analogy if you ask people what sparked the peak of the Industrial Revolution 120 years ago people will be saying hey internal combustion engine electricity Power steam engine no it was precursors that was a beginning invention of assembly line changed everything products became cheaper reliable quality and reliable quantities if we look backwards on how software is being built for the last 40 years you have an idea you got 50 engineers and designers and product managers and for the next two years they're building a product well you're lucky enough you get product Market fit and you get the company billion dollar company right not in many cases but that's that's the process and everyone is taking it for granted that's how it is but if we look back 120 years ago how cars for example been built they were all custom very expensive unreliable because every car was a custom build then Henry Ford invented assembly line and basically what they took an engine engine was going on assembly line and the special people people trained people were taking pre-manufactured components every component was a known quality no custom build and assembling the car in four hours you can have a car with a reliable quality and reliable quantity and the price went down 10 times and people could afford it that made an impact to everything else roads people Transportation gas everything everything changed so right now we're in a point where industry the software industry is going to be changed we don't know it yet but we were preparing for that we didn't know what the engine would

be we had we had an idea that that would be a model that's why we're building our models the beginning but then this explosion of llms gave us so much more powerful tools and we already had all the components buildt around it so think of this we can build a generative AI application for Enterprise like Fortune 1,000 clients in under eight weeks a new one because we take l M and move it as an engine or set of llms some hours some open source moving along that assembly line and our people assembling the application along the way in under eight weeks you get an application that in old world would be a billion dollar company on its own it would take two years to build and $5 million to be invested MH so that's how we in a matter of six months last six months Grew From three applications to eight and we're adding more and more and each application is reusable across every other client and that's the concept that everyone is going to be using from now like probably year or two away but that's going to be how the software is being built from now on long time time gone where it's custom expensive and really hard to I think it's super powerful concept of a assembly line for software and I think it llms made it possible because it now algorithms understand language and language is a universal connector to anything human built so you're writing a really good wave I've just read that like 90% of the companies software companies planning to release new AI feature mhm like this year maximum next year but only part of them applying to hire new people because they don't need any new people

with all other LS yeah so what are the eight apps that you currently sell so we focus on let's go top down so we focus on two specific parameters for whatever we are addressing number one is high Precision AI applications I'll give an example what it is and second paramet like we have two circles overlapping one would be high Precision AI applications and the data heavy processes so high Precision means there is no room for hallucination literally zero tolerant and for example some of our applications are processing hundreds of Millions of dollars each week for our clients imagine there is a hallucination instead of million dollars you send $10 million to someone like it's not going to work so high precision and data heavy process is where a lot of data involved or a lot of manual work is involved if it's a very expensive process but happens once a year that's not us and or another example if you have an AI company that creates I know marketing emails which is great business but no one cares about the quality it's like 60% 70% 80% because people still going to be rewriting those it just enables something there is no need for 100% quality of marketing email or a blog post right so it's not us we focus on those particular processes that are require absolute quality of AI application and a lot of data involved Better Than People right better than people people are 87 90% at best We once received $20 million instead of $2 million in our startup. nice did you give it back we had to like we we asked like what should we do and they said like just give us more shares and we're were like

but that's 60% of our company was it on purpose they might be like okay let's send this guys 20 million and see if they accept it look like it was on purpose but it was so hard to send it back looking back we should have gave them more shares but yeah so it's I I understand that AI is held to a much higher standard Than People computers are held to much how do you achieve this so there is no Silver Bullet there is a set of silver bullets it's basically the whole machine gun of silver bullets you need to use because with the traditional large language models especially truly large models like 200 billion PRS plus they are trained on everything and they are wide as ocean deep as a paddle so they can take your p&l and turn it into Shakespeare style poem would you care about it probably not but they still can what we do we specialize our applications if it's a vertically oriented application B in class for example we have an applications for private Equity notices processing it does Capital calls distributions navs invoice processing better than humans and it's specifically trained on specific data it's connected to the data of the client it's inside security perimeter so it's a ground truth and it can be verified and checked we have special models that supervise other models so it's checks and balances and there are parameters of confidence and quality so if model that does the job says hey here's the quality I think it's 95% and the model that supervis that model says wait a second I have low confidence in what you just said because I checked it with the ground truth so and that's how you operate

with it and there is so many other problemis like thinking about what ensures the Safety and Security of a car just can't point one thing of course do brakes but also airbags and elect electric wires and Lars Radars and everything overall quality of the assembly exactly assembly yeah your your wheels can fall off right no brakes will help if the wheel so that's how you ensure that highest possible quality it's really really hard to achieve and there is a term for that it's called L mile of AI quality it's so easy to proc prototype with GPT 4 do anything you want at 65% quality I call this Twitter demos work amazing on Twitter exactly but I would compare it with a like what we right now with a dawn of Aviation see I have a lot of analogies with the manufacturing I don't know why but it's basically historically proven so we are at the dawn of Aviation of AI right now it's like brothers ride first flight we're in the air plane is made out of dried paper and duct tape 300 yards flying amazing Enterprise clients need boing trip 7 they need everything around infrastructure in the airport and the cocktails in the first class because they can't buy the paper plane just doesn't work this way and it also can be seen as like we've got as an industry we've got a magic wand in 50% of the time when you use the magic wand it can turn yourself into a frog so and there would be this steps taken by the industry of those components and assembly lines and companies would going to be adopting this approach because there is no other way they can build it scalably with predictable quality

and predictable price so in the next three years everything is going to change we just don't know how much it will impact the whole industry but it will so what you're saying basically translates for the new Founders listening us out there is we are So Sophisticated that there's no way you can do it now you got to you you had to do it for the last six years to be on the same page actually opposite how do you how do you get to this point if you're starting now opposite opposite actually I think right now is the right time to get things the right way because large organizations like Salesforce of the world Oracles of the world they have this old process so engraved into the the nature of this it's part of the DNA they would never be able to adopt this assembly line type of concept they still going to be doing things the old way because there are 100,000 people and you can't change the processes this way but assembly line actually it's a concept yes we've been building it for four years and build all 200 components so we can do things faster now and cheaper and better but we are serving specific industry there are so many other Industries what I think startups will do better and it's a beautiful moment for them to actually take on those large organizations adapt the new way of building things and build your own components it doesn't require building everything from scratch we did it because it didn't exist back then take L length chain take this take that put it together and build it the new way scalable predictable and reliable and that's it so it's it's it's amazing time we're living

in it's a whole Paradigm of how software is being built is going to be changed so the eight apps you've you've designed for your clients so what do they do other than summarizing emails no they're not summarizing emails anymore it's it's probably one of the feature in one of the applications somewhere off the shelf right so every application of ours is focused on three steps and those steps they are engraved into every business process of every organization in the world starting from Healthcare and going to I don't know Banks we focus on Professional Services like Financial Services legal accounting and Consulting that's our area of expertise but those three steps are data verification I'll explain what it is then data data extraction data verification and then data transformation so those are three main components and I'll bring example of one of the applications how it does that so let's say you have an invoice and or hundreds of invoices or sometimes hundreds of thousands of invoices lot of organizations operate with a much bigger scale of data and then you have contracts and those invoices before you send it to a client they have to be verified according the contract and a contract is not a simple thing it might be 100 pages of a contract I've seen those and there might be 20 or 30 amendments sometimes amendments contradict itself there calculation how you calculate price discounts it's incredibly complicated no human actually can take an invoice or stream of invoices and check those invoic against those contracts it's impossible so companies being fined for doing sending something that shouldn't be sent they under underprice they overpay and so on so forth so one of our Solutions What It Takes it's called policy enforcement

one of the apps it consumes those all of those contracts like literally hundreds of thousands whatever you have of contracts understands them their special model for that extracts the rules mhm basically Pro digitize them basically make them programmable you have those rules and then another model inside the same application applies those rules to those Contra those invoices saying hey wait a second wrong here wrong here wrong here and then it transforms it the third step it says this is wrong but it should be this can I do that and in many cases it might be fully automatic no humans involved many cases just one operator just looking through that saying okay agree agree agree agree that also creates the reinforcement learning data which gets back to the model making it even better and faster and so on so forth that's an example this is a billion dollar problem for Fortune 1,000 each Fortune 1000 organization another example Healthcare you have medical bills and you have guidelines how to calculated medical bills everyone knows that if you get a bill from a hospital we know everyone like no need to in this case everything can be verified and fixed and that's where AI is the only way to do that because if you do it want to do it manually you have to employ Army of people and there will be still like 85% quality where I can achieve 95 plus qualified people too qualified yep and you have to train them then you have to retain them it's big problem yeah motivate so there are multiple applications eight but they all focus on these three things comparing something to something like data in big quantities identifying problems and then fixing those problems and if you

look at any business process it's three fundamental step of every business process for any organization in any industry and if it's large enough and then they can get enough Roi to pay you a 300K per year yes absolutely yeah I'm guessing there's a lot of companies that have a lot of large processes in this world one of our clients actually in Insurance sub broker they have a process that cost them $3 billion a year and one of our applications just solves the problem completely on a ground level so this problem doesn't exist anymore it's gone that's incredible how many competitors do they have that also have thison problem I would say there are companies that do parts of those three things some of them manually some of them half manually but no one still is doing end to end all three things from the beginning till that completely augmenting the process but there will be definitely will be it's such a big piece of a pie for every industry so there would be but right now no one is doing it completely end so speaking about your company now so what does it really consist of in terms of people we're about 110 about 60% engineers and the rest is GTM and GA so we have three main offices Silicon Valley here about 15 people now here Canada that's a customer office Toronto and second biggest office is Armenia Yan we also have people in Germany Poland and Ukraine cool so what do they do other than 60% of Engineers so right now starting December last year I'll say probably November it's such a big Tailwind especially with AI and it's so funny in every large organization what happens is that the board even even

in the fortune 1000 the board looks at CEO and the board meeting saying what are we doing about Ai and CEO is like hm what are we doing about AI okay I'll I'll I'll I'll reply next next board meeting so the CEO goes to the CTO or VP level saying guys what are we doing about AI these guys are hm what are we doing about Ai and they go down to level of director saying guys I don't care what you do but we need to do something about AI because in four months I need to report back what our initiatives are so and these guys actually who are in the trenches they go to try prototyping things you you you you've said that right Twitter prototypes like okay we have that problem we'll never be able to solve before let's use gp4 API and we have Smart Engineers they try and like damn 65% quality so they can actually go to CEO saying yeah we did it works because then the board says well let's implement it how it affects our bottom line and there's a lot of risks and there is a whole new problem right now the Board responsibility for AI like if the data leaks if AI does something wrong board members they have the fiary duty they have responsibilities and they're actually not comfortable using something that is 60 70 80% quality and in this case those clients actually what they do many of them come to us so just just in the last 6 months our Pipeline with a Fortune 1,000 companies grew probably th% probably even more and it's growing like crazy because they prototype something it's not yet there but the top is demanding something and they're trying

to find companies that are doing specific things and can ensure that the quality so you're mostly working with inbounds at this point yes how do they find you like ref by yes referred by clients they sometimes see podcasts like this industry related because they all like always industry focused also it's a like though it looks like insurance for example a big big industry actually there are probably 20 biggest players and they all talk between themselves right and they know about things and they share information so we are doing like we're growing in terms of inbound dramatically but it also import important to separate who you want to work with and who you don't want to work with and not because they are bad companies no they're all good clients potential clients but it's about use cases you want to narrow down you want to keep focused both on an industry you want to work with and with the products you either already can deliver off the shelf or it will require less time to to assemble so you're now past product Market fit and you know exactly what works so you're like laser focused and just doing the same thing over and over and over again until until what time like what's the limit when are you going to start adding stuff we we're adding stuff but we adding stuff very carefully because it's when you have something like Hercules our engine and you can easily add whatever you want you lose discipline it's like hey let's do this and it's very important to say no more than yes because the sales team comes in saying hey there is an amazing client came in they have a billion dollar problem let's do do it right

now like look at the pipeline you need to deliver to this client and this client and that client and this client they just been pitching to product engineering they are a different industry they are outside of our focus and the product well requires slight modifications and also will other clients use that that application it's always a balance if it's a custom build we would never touch it it should get back to the App Store of ours like app center and we can resell it to every other client of ours that's the flywheel how we grow inside with the same clients rather than getting getting new clients faster so it's much more important focus on that so your GTM team has people who are in charge of top of the funnel do you have sdrs or we don't actually need sdrs we have a yes we have AE but those more of a they they wear two hats they wear the traditional AE count executive hat but also Consulting head because they know the client comes in say hey we have that problem and raes need to dig in and help client on the first do they hire do do you hire them from Consulting Deo pmgs of the world I Love These Guys these guys are trained to spot the problem unfold it and present to a client the way the client can consume it saying oh my God why we were not doing anything about it right they're also good at Travel living off Hotel mini bars yeah no one no one is doing it this way right now it's all it's all remote but but it's it's a very important quality of helping clients not just get the ri delivered but in the

first step understand what the ri would be and how deep the rabbit hole is and once the client understand and also it affects the sales cycle which is incredible typical sales cycle in Enterprise is like what eight months 12 months Fortune 1,000 yeah how about two months two months that is that is what bought me honestly when we talked and we decided that we want to invest that was the two month cycle and you know why it happens this way because assembly line exactly client comes in and first meeting we basically deconstruct the problem they they have and typically it's a I call it X in the back problem X stuck in the back rather than patch on the like scratch on the knee it's bleeding it's painful what we need to deconstruct it to understand what client exactly needs then we tell clients hey guys give us examples like five documents five examples we don't need more that consist part of that process like might be contracts invoices whatever process you have it's always about documents give us the data five examples of course we sign NDA in two weeks we meet and we'll show you the working product we take that data basically put the llm on assembly line go it through replace some component on needed for these guys but the product is basically off the shelf 95% train the model on that particular five examples because we have a model that generates hundreds of thousands of synthetic data we fine-tune the model and next meeting in two weeks we show them working solution and we say now upload your documents it blows their mind away because they were expecting to keep talking and talking for months and kind of trying to

understand things and then they like okay send that's the same product that is going to be implemented yes exactly that is very impressive but it's on our side yet so we show them the product that's on in our in our on our site we just give them access to play with it but it's already working Pro product for them and then they the second question they ask how we can get it inside our security perimeter here's the contract sign it and in four weeks it's going to be there so it takes time to get through security checks but again because we're on Prem on inside the security perimeter C so and it loves us no data is being sent outside so we pass security audit in one day then the procurement three four days pricing is clear no volume based procurement doesn't have to scratch the head how much it will cost next year it's always $300,000 per application per year so procurement signs it quickly and we get into deployment with the clients and we build the flywheel to repeat that process there is still wrinkles of course some clients are different especially when you work with Fortune 1000 but in general 90% it's the same process that is very impressive when our startup we had a platform for Pilots that we would upload their video and produce results within one day our kpi was like one day but that was a different platform from our actual product because it had lower accuracy could not be deployed on Prem that is really impressive that you're like actually doing the product using the assembly line approach if you like with all your experience what you buil already if you look back you in 2017 yeah

so what what would you do differently such a loaded question there's so many so many things I would do differently but one I would actually accelerate the Hercules creation because it was been always done on a okay let's spend 30% of the time on building this fundamental things and 70% of the time on building like serving clients building bu building new features I would actually reverse that because now like we would probably get there two years earlier at the stage where Hercules as a as can function as a assembly line were you sure about your pmf back then to be able to investour absolutely not like I would say that time we were sure about the MF of the small Beach Head Market we worked on because once you get five largest law firms in the world as clients you know you're doing things right because if these guys buy you absolutely found the product Market fit what I was not sure about is how we can jump I knew that we will but I I was not sure about all the characteristics of how we will jump from this beach head basically from early majority of legal Market to the later majority or majority of bigger markets like Insurance Consulting financial services that was not clear and we managed to do that only because we had Hercules so it was like I wish we invested more in into it earlier but at the end it was right on time and interestingly actually it was right on time Hercules was ready probably mid last year but it lacked llms a variety of llms we had our models like our largest model was at that time close to a billion parameter based on BERT but we

lacked the variety the community of llms and then it's exploded and Hugging Face now you can go and take all the open source models and we got the variety of moders around which we can build this AI applications and move them faster on assembly line so I would say even if we build it earlier we'll probably grab more Market sooner but but overall this explosion of generative AI gave us enough fuel to build this new applications so you're now on the exponential curve and looks like it is it's probably you're going to be scaling everything hiring more people what is what are they going to do the next 100 people are going to hire well the same analogy there are two ways of doing things intensive and extensive so in in the case you have all those tools and all the pr requisites and components you don't actually have to hire that many people especially on assembly line because you already have everything automated so you automate your own processes and your own app creation so it definitely will hire next 100 people I just don't know what the time frame would be but those are mostly going to be customer support GTM salese marketing because that's where you skate and of course Engineers but Engineers more on a quality side because it's going to be more R&D what's next remember we always try to think five years ahead so what is going to be look like how it going to look like post llm or next level two years three years from now we need to start preparing right now so you're finally at the point where it pays off to start your own like R&D on we always been doing R&D but that

would be more of an R&D of futuristic like what next what next gonna be let's experiment let's start looking trying to look Beyond the Horizon and startups even at the very beginning of the life cycle should do that even if it's just in the head instead of building code because sometimes they can't afford spending money on that but it should be in the in the plans or trying to do this imaginary work like what if there's always this risk of being too early some say that being early is worse than being wrong yes of course it's it's better to be in the right place in the right time that's true but it's it you can only say that afterwards yes it's it's it's retrospective it's never goes into the future yeah I want to ask like when when did you start your fundraising process so what you you started your company how long we got the first check actually right probably couple of days after we started the company I was friends and family we got like $100,000 check we didn't even have a company yet at that time it was like blank check with 100,000 written on it like okay register the company put the name in so that's how it started and then of course we went through traditional process of seed rounds we closed a round last year we we started a new round recently so it's very traditional so another question that we we asked a few asked our audience on LinkedIn and Instagram to ask a few questions and one of the questions was do you really need domain expertise and it doesn't seem that you had any expertise in like how llms in this case or ilms worked international

law firms I lefts yes yeah I would say I will reply with a joke don't trust professional they build Titanic trust amateur they build the Noah's Arc yeah so I would say expertise is required the team has to have expertise absolutely you have to have people who understand needy greedy pieces of how industry specific industry works but having people on the team who actually open-minded and they put everything under question who do not have that engraved knowledge of the industry might be very very helpful because you've heard that many times it's been done this way for many years so no one is going to be changing it but if you say why why no one wants to change it is it the process problem is it a product problem is a technology problem or something else and if you start asking those questions and start drilling in Beyond this kind of a common truth like everyone is doing it this way everyone is driving horse to to work like should we have a car not a faster horse so that type of mentality there always should be people in the team and they have they have to have a voice they have to be able to articulate it and say hey everyone is doing it this way but why why not we try it differently and that's where the best products and Technologies are being born actually MH so designing this assembly line how did you come up with a concept of we got to have an engine and then like we have standardized building blocks for everything we started eating our own dog food it was hard to build new applications painful taking like a year to build a new application and we wanted

to optimize it's like wait a second we already have two applications we building third why don't we reuse some blocks from those here and we started doing it and then it grew into like wait a second we're like 50% are using reusable blocks of our own and there are new technologies appears oh how about we use this brought it in it's not all of of it built by us there like probably 30% are available on the market it's just the connect them the right way and then we realize like this is a better way because you see time building the application shrinks dramatically instead of year it can be done in four weeks eight weeks and it's incredible Roi because the same team can build 100 times more it was obvious at at some point but not at the very beginning yeah I also wanted to ask about the culture so you touched a little bit about it it's very important to give voice to people who are questioning and who are curious how are you doing this are you are this building your culture do you have like specific meetings Yes actually well I don't believe in culture that written on a postcard and put on the wall it's like right culture what about the conjunct triangles of success this is such a strong concept I think the the better the best way was explained it in the Ben hori's book what you you what you do is who you are right and it's basically part of how you do things and you know what are n the our value number one courage literally and we evaluate people by are they courageous enough to do crazy that no one believed possible if they are

if they question things and you can see it even an interview a person is not afraid like people could come to interview and they shy and they just like okay see always interviewing me if the person says why you doing this and why you doing that and ask questions and you see they are not shy they will do bold moves and that's the number one we're looking for because to build something like this that no one else built before and every startup does this this way and they looking typically for expertise and like ability to work like historical experience and working in a large companies I would say it's all so number two is attitude we never hire for experience alone we hire for attitude you can teach experience you can teach something you can cannot teach attitude if the person is hardworking putting everything under question and wants to drive things that's that's the person we're looking for and if the person I I made mistakes I hired people so Salesforce and Microsoft and oh my god there was like I still have scars on my back never again so value number one is courage ability to make decision make bold moves and see if it works thank you so much for coming and joining us and we hope that this podcast is going to become something more we're still early stage we're still looking for product Market fit hopefully this iteration is something that people won't find complete maybe will get some suggestions based on what we talked about today so thank you so much again for making this possible thank you my pleasure and we'll try to be courage and curious to make it Forward yeah