â Think AI Podcast
Loading summary
Transcript44 lines
- [00:00]
A
There are many founders, especially right now. They are experienced, they are experts, but they are trying to build startup in another niche, in another industry. Because startup building right now with AI is easy. But this is not the case. Building is easy, selling is hard, and it's getting harder and harder probably every month.
- [00:21]
B
Welcome to the Think AI podcast. Each week we talk about the most exciting AI research tools, case studies and more. I'm your host Dev Goyer and I've been working behind the scene in data and AI for over 30 years. Whether you are an AI expert, skeptic or something in between, this podcast is for you. So today I'm sitting down with Alex Gallard. He has been building companies since 2006 and he has a habit of reinventing his whole strategy every 10 years. He has BRNZ platform pronounced as brains since 2016. Now due to an operating system for agentec works since 2026. He builds out of Cologne, Germany and his eyes is on the US market. Here's what Alex worth next hour. His studio Brains builds complete AI powered companies for experienced operators and takes equity only by shipping. The founder keeps the CEO seat and 67% ownership floor in writing, no investors. The target he calls M12amillion in run rate by month 12, which is amazing, funded entirely by customers. But the line keeps coming back to me in this thesis. Speed is no longer a mode and I've been thinking a lot, so let's hold that thought about it and then accountability is. Alex, welcome to the Think podcast.
- [01:55]
A
Thanks for having me. Hey, awesome.
- [01:57]
B
So let's get on with it. So let's jump directly into my curiosity. You say speed is no longer a mode. Now I understand why you say that, but then that is the opposite of every AI pitch out there. Make the case why?
- [02:13]
A
Yeah, like you know, AI changed everything. So via all moving fast, our agents are building code. We can build software 247 without sleep. And so many, so many teams are doing it. But the main question is what are you building? And if everybody can build the software for your specific market, so what is your USP and how can you compete with all this different software providers and this is our experience from 20 years. And according our experience you need to find a founder which we iconic hero founder. Hero founder spent probably 10,000 hours in a specific industry. He knows exactly how to build this specific product for the industry. He has already built his customer base audience. And this is the main usp. This is how you getting into the market fast. But the main point is you have to have the right product and this is why it's not about the speed, it's about the building way. It's more about quality, it's about security, it's about everything. What is what you know from enterprise software development. And this is exactly the opposite of wipe colleague what everybody is doing now. This is why this year as I mentioned we've built a system what is called brains. This year we've built the system what is called youth AI. This is an agentic work os. And the main point of this system is to make the bots, the agents which are building software which are performing on other tasks accountable. I'm from Germany, we love rules and we see what agents need rules to run and build software the right way. And this is why the main point of toboot is how to how to build the software that we can prove even like 10 years later it was a quality build. We did this specific test scenarios, we performed on this acceptance criteria and our audit cannot be changed by humans and agents is saved forever in our specific database. So you can use this software also for enterprise customers, for evil IT and other industries. So it's all about evidence, it's all about accounting, acceptance. And in every loop we even integrated a senior developer who is actually bottleneck, actually de facto is slowing down the system. But according to our experience agentic team in combination with senior developer with weaving the code who understands exactly what was built and processed by the agentic team. This is the fastest way to build the product. Just one last startup of us what we've launched beginning of this year. It's called teletexting.com it's about sales recovery and Devop build was about two weeks and since then our founders scaling the startup no issues, no bugs. And this is why he can he can focus on sales marketing goals.
- [05:47]
B
That's amazing and I totally understand your point on speed is not the mode. And I think the key reason is more than ever I have build nine businesses burned five of them and but it used to take time to build something. Now due to all the agent tech hardness different models, you can build something fairly quickly. When I say fairly quickly, literally in a weekend. But then the importance of having your strategy having a correct design where your customer base is, what's the competition out there, the whole setting of our business and I've seen few clients of us failing miserably due to that. Now you don't need half a million dollars to fail. You only need probably a cloud code account to fail. But nonetheless you're going to fail if you don't pay attention to how you position yourself. What's your product market fit, where it goes in this situation. What's your experience looks like out there working with different vendors who just wants to build a product out there rather than seeing the demand, seeing the product market and other things.
- [07:00]
A
Yeah, very important question and more or less my first question about the project ideas startup founder is always validation guys, you want to build this specific product, what is your validation? What did you do? How many customer interviews did you have? Uh, and this is, this is interesting point Hero founder, this, this guy who spent like five years, $10,000 in specific industry, he can sell the product without having this product because he knows exactly what's going on. He knows exactly which customers would buy this product. And this is why this guy, he can just write like small one page. We know exactly what to do, how to build it and after this build is done, in most cases it never changes. We are running one company, what's called music to biz here in Germany. It's like sporty background music for businesses like for gyms, hotels and so on. And this founder, he was from the music industry so he could just build this document provided was and we knew exactly how to build the software. We built the software 2012 and it's still running. So nothing was changed there. It was for sure extended. But this here founder, he knows exactly how to build it, how to sell it. And this is why the validation phase is in most cases solved by real revenue by this form. And this AI. This validation is getting even faster and simpler. We have founder right now who is building AI assessment system and she's here from Cologne and her idea was a. I don't want to have like a real build. I will build it myself by using cloudable. And you guys, after the validation phase is over and customers ready to start the cooperation, you will rebuild this lovable build, this mvp. Not even MVP prototype. You will rebuild it and transform it in real product.
- [09:27]
B
That's amazing. And I'm also going deeper into one of the things you mentioned. So you talked about the Brains model and that definitely intrigues me in positive and both are negative, I'm not going to lie. So walk me through how Brains actually works. Business for equity. Interesting concept and I've seen a few people here in Southern California doing that. A67 founder floor in writing, right. And equity earned only by shipping. So the risk is yours because if they don't close the business, that's the negative thing I was thinking. But then you're saying 12 target of a million in run rate by month 12.
- [10:04]
A
Right.
- [10:05]
B
So I guess you're qualifying the companies who have a run rate by 12 months. Is this how it's worked? Break it down for me. Yeah.
- [10:15]
A
And we are building startups for equity, like business for equity since 2006. So this was from the beginning, our model. And the first 10 years we just built different software products and as you know, 90% of them were not successful. So it was also in our case so huge amount of waste, code time, resources. And this is why 2016 I decided to build one platform what can be the base for different companies. The only reason to build this company was to just save the time. And even if you have a startup without success, we can still like we use the code because the startup doesn't need this code because whatever it will stop the operations. And this was 2016. So we built Brains. And the first business case of brains was building of automated AI based e commerce brands. So we built our own warehouse like 20 different brands. And 2016 Viv built AI what was running on TensorFlow on trained model with this version, no OpenAI Cloud and anything. So yeah, you had to build scratch. And this is what we've seen. Okay, Brains make actually sense. So first in e commerce space we've built this brands when we extended the platform so we could build different fintech applications, different software service tools in the e commerce space and some others we've tried different systems. Our point was what is the maximum motivation of the founder? We've tried different equity splits, we've tried like 50, 50, 60, 40, 1090 and we've seen what like 67, 33. This is like the sweet sp part and this is why it's like fast. You probably can see it on our website, on our LinkedIn page and so on. And the model is actually pretty simple. We can build like any type of software in roundabout 1000 human hours. But it's not this time, it's just our orientation because this is time where you can build product that can make money in the market. And this 1000 hours human hours are super compressed by AI. This is what you can build very, very fast. Because we are not building from scratch. Brains is more or less like Amazon web services for building startups. You have your infrastructure, you have your security for sure. Everything about payments, integrations with blockchain, all the frontier models, they are just existing. You can just take them and use them. The only point we have to build for every startup what is unique is branding, ui, UX design and this is why we we can ship these startups very, very fast. And to accelerate this process we've built this new protocol. It's called dude AI. And dude AI knows exactly how brains is built and how brains must be extended if you have to build this specific case. And this is what is making developers even faster. So we are not building from scratch. We are using an existing core system and in most cases 60% of everything what this specific startup needs is already there. We can just use it, probably customize and integrate in the solution. And 40% are built by agents, verified by our security agents, approved by human senior developers. And then you're right, wrong. This whole sweat.
- [14:42]
B
Well that's pretty amazing. I'm still intrigued with that idea and concept. And then you know, my business leaders are saying so how do you manage the risk with the customers? Are you charging anything to them when you're building it for them?
- [14:58]
A
This is exactly this point and probably a USB of our system. We are not charging customers at all via covering development costs from our already existing companies and existing workflows. This is why this hero founder who is the industry and insider and industry expert knows exactly how to build the software, how to sell the software. He is not paying at all via completely on the equity strategy. And this equity is vested of months in most cases. So we are getting actually equity for shipping the product. So if something is not working, if we cannot provide this product and so on, we are not getting equity at all. This is the model.
- [15:48]
B
Yeah, no that's perfect actually come to think of it. And but I'm sure you are funneling them, you are filtering them based on seeing somebody serious and interested for long run because again your skin is in the game and they need to have something in the game to yeah, 100%
- [16:06]
A
and we win only if the company wins. And according our experience from probably like 100 projects in this space like with funding, without funding, this exits with success. We've seen what the founder might be. Full time founder. This is like the first point, this is why it's so is it what the data. Are you running this company full time or part time? Because you have your whatever full time job and you cannot focus on this.
- [16:42]
B
So this is point if you will.
- [16:44]
A
Yeah, yeah, exactly. This is why it never works wrong way. So you can say hey I'm running this company right now part time and I will switch to full time. Then I whatever raised money or generic revenues just not working. So yeah, full time is the first point. And the second point, as you know building startup is Slow. So the founder, he must have the ability to like at least like survive 12 months of operations and be focused not on making money like for living, but on building the company, building the product, marketing, sales actually grows. And this is why, this is, this is the second, the second theorem. What is very, very, very important for the startup building. And yeah, the circle points as I already mentioned, 10,000 hours in this specific industry of the startup. Because there are many founders, especially right now, they are experienced, they are experts, but they are trying to build startup in another niche, in another industry. Because Tata building right now with AI is easy, but this is not the case. Building is easy, selling is hard and it's getting harder and harder like probably every month. This is why my favorite book from zero to one from Peter Thieves shows exactly this. This first step from zero to one. This is, this is what I love. This is what is hard.
- [18:18]
B
This is.
- [18:19]
A
And scaling from 1 to N. This is totally possible by AI. This is what you can totally put on the agents and it'll just help you to run it.
- [18:30]
B
Yeah, well said. And you know one of my parallel story about jumping from jobs to business, I always have that mindset to create. I'm a founder, innovator and a builder just like you. And I was never good in selling but I knew how to build stuff. So now how about I build for others is what the thinking was when we created this business. My son was born, I was working with a large hair care company but I still decided to jump in. And if you're not committed, then you're not committed. If you put your foot in two boards, you're going to sink. It's basically as simple as that. And yeah, I think that's your criteria to bring people in. That's great. Let me switch gears. So I love everything about brands but then you also talked about the whole thesis about accountability and you know, you have talked about few points on it like identity, evidence review, audit metric and memory. Unpack a few of them. And how does this AI worker looks like inside dude AI in production going with that philosophy.
- [19:38]
A
Yeah, the main goal of duty is actually autonomous building of startups. So we defined five different levels of autonomy and as I mentioned before, our startup size, like MBP size is 1000 human hours for building this first version of the product. And the maximum autonomy of dude AI will be if we can just tell the system to our agentic product owner, hey, this is my idea. This is, this is what I want to build. And it will just build and provide us a ready to run company. So this is autonomy number level number five. And right now via the autonomy level number three, this is why humans are involved and they are, they are part of the loop. They are doing code reviews and everything. How we are building, dude, dude is building actually itself. Because this level 5 autonomy is defined as like main KPI, we are calling this agent in delivery intelligence. And every week we collect the data from the dialogues, like dialogues between the agents, between humans and agents. And when we collect source code and what we produce during this week and different other KPIs so Jude can understand okay, where are the bottlenecks, what I have to solve, how I have to improve these metrics. And it's even like forecasting when it will achieve this autonomy level five, how we are building it. We defined three level of these KPIs like main KPIs I mentioned is agentic delivery intelligence. And it's splitted to software code quality, security, stability and so on. So and so. And the system is doing following you have a startup what is in build right now. And during the daytime when people are working then people are providing tasks, the system is working on these tasks. And during the nighttime the system is doing self optimizing in closed loops to optimize this one specific startup KPI. So we have different KPIs and these KPIs can be defined even we can even define different KPIs for one specific starter because it has different workspaces. And our idea is or the main question is not to build a company but to build a closed loop as a company, self optimizing company what learns and understands exactly what's going on, collecting data, collecting in the best case already revenue data, customers dialogues and so on, so on. And based on this data it can optimize itself. So just as an example, so you understand Adi is for Duke, this is more or less clear. But if we would talk about teletexting. Teletexting is in the area of sales recovery. We are working with Roundabout 30e commerce brands and we are using imessaging for recovery. This is why we can collect from imessaging dialogues between our agents and users everything. Sales discounts and so on and so on. And then we have also financial data from Shopify shops and for sure we have different other KPIs. So on our side we can create or we've created new KPI, what is called revenue intelligence on our side and we can help these brands to increase their revenue intelligence because we are tracking all these thousands of dialogues of brands. We can say tell them the shop owner, hey, we know exactly because 50 people told us your subscription is not working. You have this specific box in your shop. We know exactly if specific influencer is profitable for you or not. We can create a huge amount of data, we can create a playbook for this specific shop, for this specific product owner, how he can increase his revenues even without spending more. And using these KPIs. Yeah, as I mentioned before, we are trying to create a closed loop self optimizing company in every space. And this is why we are talking about revenue intelligence, agentic delivery intelligence. New startup for the scamming here for cologne is doing pricing intelligence and so on. By using KPIs you are providing like the North Star, like the goal for the system, for the agentic system. And in the best case agentic team will achieve it itself without any human health intervention. Not possible for now.
- [24:40]
B
But yeah, our vision no very interesting. And I must say you've got the agent Ki done right on the accountability point. I have a peril story to tell. So I had a client who want fully autonomous document system, no human at all. And then we actually walked them to one edge case where the system flexed the wrong vendor payment and they decided then to keep a human at the 95% confidence threshold. And the technology was not the problem, it's the accountability framework was. And that needs to be think through more than ever in this era where you just cannot leave everything in the hands of AI. While AI can help you in thousand ways. I have a whole team of C suite for my own, what I call AI crews for my multiple businesses and you know they have daily standup and I see that in my notion and I see what they're talking about. I have this whole task database where they are creating different tasks and then they are also assigning it to human in making sure that that gets done and being executed upon. And then my focus in my second brain is just top three but then it makes mistakes and you gotta know and acknowledge that. And the accountability framework is meant to do that. I do want to switch gears. So we did talk about, you know how you are building startups, the profitability model, you know, shifting the risk on you so that they can freely build their idea and then you know everybody has their own benefits around it. Then you also talked about the accountability and self building AI using dude AI. But then all this brings with some dirty secrets on security. So you talk about the dirty secrets of AI generated code. Tell me about the real one. And the customer data hole you found and close in a partner's project and what security gates actually you apply?
- [26:52]
A
Yeah, actually very interesting, very interesting question. A couple of months ago we've seen what AI coding like wipe coding is not really working. And this is why we started to build our own even like security startup. What built but never publicly launched. So this is, this is why just our internal security system. But we've seen like 40 or 50% of all these wipe coded projects have security issues. So you can go to many of these new projects which are probably started this year or last year and run your security audit and you will see a huge amount of yeah security issues which can be used to steal the data and get an access and so on. And this is why as I mentioned we built system what is now part of Duty AI. What is integrating Duty AI software development process. And this is agentic security orchestrator running first the analysis of the code so it understands okay what this specific startup is doing, what is the product and so on. And then it decides which of 40 integrated open source security tools it should run. After this is defined, we are doing deep and deep scan using these tools and then we know exactly what's happening. There are the issues, what we have to fix before we go live, so on. And to do this we are collecting every four hours actually available CVE databases. We have already like 400,000 CVEs which are existing in the market in our database every of our process before we go live against this CV database and see okay there were no critical vulnerabilities. So we can launch the code and run it. And this is exactly the system what we actually just as a help run Ginset partner project and we found out what there was an access what was not somehow protected by a password. And without like any administrator administrator access you could just download all the customer's data users data. So yeah, we reached out to the partner and I helped him to close it and yeah it was removed. But this showing me what security security is one of the main point. This is why this agent, this security agent is part of every of our process in software development. And this is exactly one of the main companies right now also integrate senior developer who understands what software is built.
- [30:15]
B
That's really good and I'm sure you are paying a deeper attention to that. I have a little bit of different story on this topic. I worked with a client who discovered that they had about 40 AI tools running across the arc. Obviously there is no security audit and things that is out there and it only knew about three. The CFO was expensing a dozen, just a CFO on his personal card and the security exposure was real. So it's not just about the money by the way, as you can imagine. It's also about running in your organization what security holes it's placing. And so the first thing they did is they went backwards, you know, solidifying their infrastructure, finding out what security holes they already have in their organization, which is allowing all these AI tools to embrace, harness and grow in the organization. And the shift completely went towards it actually retracted them from AI solution, which is a little bit of a sad story. But at the same time it's the right thing to do at that point in time to bring some governance into it. So security compliance and AI works together and along with the data. But those four pillars is where it takes you to this new era of agent tech engineering, software development and things of that nature. I also want to switch gears on. So I am big on two things actually. Three leadership motivation and technology innovation. And motivation is. And also leadership is very well connected with failures. I just mentioned about nine businesses, me also being a disabled entrepreneur, being a dyslexic kid and some of those things. So I figured out my own way to overcome my own issues and at the same time how I can encourage and embrace others to do that. What failures you have actually seen, you know, your personal, professional and this tech life and how those failures have changed how you build.
- [32:31]
A
Okay, interesting question. And As I mentioned 20 years in building building startups like building companies Viet different, different stories. One story what had like bad and good sides was company in the area of project management. These huge customers here in Germany like Wundermann, what is part of WPP holding from New York, this young Rubicam and some others brands. We built project management system what was used by Microsoft, Mazda, Ford and so on for planning curve via ads campaigns. And this system was always built for equity. So we've been bought equity partner and after it was launched main partner like the hero founder who brought us this expertise. She. He actually died on cancer a couple of weeks later after the launch. And we've seen. What you have to do for the future to protect also your equity against in such cases. Because in his specific case the equity went to the government and we had to handle everything the government just to continue to run these operations, continue to run this company. And the main point this was what I mentioned the good side. The software was running for another like five to seven years in these different companies. We supported it but yeah, this was the point. So everything about, about like the bad side in the business. What happens if the partner dies? What happens if he wants to exit? What happens if he is not performing anymore? This is, this is what you have to manage before you start the company. This is very difficult to discuss after it's running, after you're generating revenues. This is why all of this is must be done before the start. And as I mentioned also during the whole way we've been trying to decrease wasting of time, energy, money for building software, for building startups and actually for instance you, this is the system where we can say we will probably even build more startups wasting less time because we can focus on choosing the right partners. Actually yeah, as you mentioned validation and the specific criteria, founder criteria and having these new tools, hopefully you are not building much more, but you are building the right things. This is, this is what you can really define right now. Before you start the coding, before you start the programming. So just small example, couple of years ago it was always a how, how I built mvp. But it changed now. First question is I want to start this company. What is the design system? What is the brand? What are the main colors? How can I define, define all these points so I don't have to rework my product so it looks from the beginning perfect, My landing page, my products, emails and so on, so on. And as you know we are in the attention economy. This is what allows you to get this attention. So design first, coding later. This is the main road.
- [36:37]
B
No, that's really amazing. And also a lot of times I have talked to my customers as well that when they are building their teams. I've spent my life building data teams and then do not hire the data scientist before the data pipeline and that thinking of, you know, fast cash instead of a strategy kind of failure in so many ways. And if you skip the foundation, you know, exciting part is the model. So people actually see, oh yeah, I will just go and build something. You know, I'm a builder but strategy comes first with the builder, what works, what doesn't work and then it always comes back around. And your stories makes perfect sense to the people who are seriously thinking about building but have no experience on building. And I think you have cracked the formula. So kudos to you on what you have been doing so far. I want to go back to, you know, the rejection a big chunk of audience that people I am connected is either solo entrepreneurs or jumping into the business of AI. They are either AI curious, AI skeptic, even Or AI enthusiast. So they may attract based on what you are doing on brains and dude AI but then you run tech for equity model company and obviously there's engage to get in. So which kind of founders do you reject and why?
- [38:10]
A
Yeah, actually we have this three criteria what I already told to you. So remember yeah your main point is full time, 10,000 hours in specific industry and in this specific industry should be the startup and Runway. Yeah like 12 in the best case. 18 months of Runway should be safe for the founder not to pay us or to pay whatever infrastructure. This is everything what we are providing, it's all about his living and and
- [38:43]
B
sorry to interject there. The reason why I push back onto that is so, so that's a good base level criteria. But then there is always this industry knowledge a particular line and segment if they don't. A lot of times people are just having an aspiration and they'll say oh yeah, I will do that. What experience you have? So I'm working with a founder right now or he wants to be a founder, he wants to build a tech in you know 3D imaging. But then so he's experienced there but there is no experience. This is for medical industry. That's a different ballgame altogether when it comes to USD to compliance. And if they don't bring the knowledge the risk is on you. So are you looking into those elements as well when you are eliminating them or funneling them?
- [39:32]
A
This is why we cannot do any type of startup as you know, you're completely right. But we cannot do at least right now like robotics and probably like deep tech systems platforms because we have to finance this first MVP and this first version of product. So this is why we are always talking about what is possible to build in like 1000 human hours as a product. But people would pay for this. So it's probably more as like mvp. But it's not, it's not like a huge, I don't know search engine like Google or whatever startup and in your case 3D or hardware startups, they take much more space. So we want always to understand what is needed for you found to validate your idea. Is it something what we can provide? Is it somehow already existing on our site? What we can just reprint, customize and give it to you. So you can provide probably even free or small amount of money to your existing audience, to your existing customer base. Because this first feedback, the product development starts. Everything before is just guesses. Everything before is like just an idea and probably hobby. But with the first customer feedback you started this real product development and we need to understand what is needed to get to get us to this first feedback. The interesting point about different countries in Europe, people cannot really think in MVPs. People are thinking, I want to change the world. I want to build this huge company with, I don't know, thousands of people and we will achieve something that was never done before and so on and so on. And US founders, especially hero founders, they always think in mvp, what is the minimal viable product, what I can ship in four weeks, in six weeks, in eight weeks to get to the customer, to get the first feedback, to optimize it, create the first revenues and scale our business in. This is more or less why our main focus is on the US market.
- [42:05]
B
No, that's amazing. And you just mentioned something which is a lot of philosophical value I have too, which is you build early and fail early rather than you keep building, building, building because your customer will tell you you fail and you succeed in the long run.
- [42:24]
A
So.
- [42:24]
B
So that tells you that's the philosophy US definitely runs on. I want to go back to that point because it's pretty interesting. You work or you build out of Cologne, Germany, and then your eyes is on the US market you just mentioned. And then, you know, our holy grail is Silicon Valley. Although now there are pockets in across United States, like New York, Silicon Valley, Denver, Nashville. So many places are evolving right now and have evolved. But Germany versus Silicon Valley, what does each side get wrong about AI? So good I understood, but what is wrong in each of the side? And I've also dealt with India, so I know a lot about what is good there and what is bad there. Why do you call colon a feature, not a bug? You being over there is a feature, not a bug. Yeah,
- [43:18]
A
we've seen what this agency can do, whatever is possible. Right now our understanding, like German understanding is everybody needs rules. And this is the main and the core idea of Dude. Every agent has his own like personal id. We know exactly what this agent is doing, which tasks it's working, which test criteria, acceptance criteria it has to perform, to process. And all of this is not really existing in the systems which are used by the most people like Claude, like Codex, because all of the Systems are getting PRDs and producing software at the end, agents telling you, yes, I'm ready, it's done. But there is no evidence, with no real evaluation, with no structured process, how you can prove what was done, how you can show your results which you achieved. And these are real results and not hallucinations and so on. And this is exactly what we see, okay? We take this power from Silicon Valley and bring it on. Our rules, on our framework, on our governance evidence. And when we can control this, when we can build closed loops, we can build self optimizing companies. But only if this is under control right now. Otherwise it will be just huge, probably explosion of actions, but no real results. And this is not something what I've like learned from the books. This is how we started also to build AI companies in SIEM and made this experience. And this was the beginning of Dude. So our main point of dude is all about control and governance of AI power. And as you know, frontier models are getting better and better every week. And this is why to update our models like four, six hours later and get this maximal power into the system. But yeah, without control it just cannot work because we are not building ideas, hobbies and so on, we are building companies. And these companies have completely other criteria as when you just want to build some small product for your personal use or for your, I don't know, for your team, but they'll never be exposed to hackers, cyber attacks and so on.
- [46:15]
B
Well, that's pretty good. And you just mentioned building companies, so let me put you on spot. Can a company run with no humans? And let me set the preface here, so give me the honest answer, not the pitch. How far a business can actually run today with no humans in it at all? And where is that line?
- [46:39]
A
Yes, this is an interesting question. And we are running experiments on our side. As I mentioned to you, it's all about autonomy, Autonomy, autonomy. Level 5 of dude would mean what we can just scale and build huge amount of companies, partly people involved. I mean on the technical side, yeah, everything about business process, business idea validation, this is still human. This is what we don't see, at least right now, what you can replace by agents. This is why it's not even like this small experience. We are talking always about 10,000 hours because this is where you are starting to understand exactly what's going on in this industry. You know exactly the insights, how to build the product, how to sell the product, how to scale it, which partners can help you to build, to build whatever partner network and other scaling options. And all this information is available, all this information is existing in the AI. And you can totally ask how can I scale this company? And probably will get the answer, but we don't see what it's working right now. So strategy and business model is on human. And this is exactly the point. From 0 to 1 and from 1 to N, this is Edge. It's agentic, AI based and fast scaling.
- [48:17]
B
Yeah, hold your thought on 0 to 1. I have a question. We want to geek out on that for, for a minute, but going back to the autonomous challenge here, do you think is it more technical, legal, so strategy I totally understand, but which areas can be fully autonomous, automated or fully autonomous? You know a good example, I drive Tesla. Being a disabled person, My wife and I both, we love it. We drove 3,000 miles across the country in Midwest this summer and we were very happy about it and had a lot of fun, pretty relaxed vacation. But then again it makes mistakes from time to time, so you have to be extra cautious over there. So where you see or where you see the potential of fully autonomous, there are certain areas and I have a philosophy there which is anything that is connected with emotion cannot be given to AI that is your own. If you want to go play golf, you can prepare through AI, the robots and things like that. You can improve your stance, you can improve your drives and everything of that. But the feeling you get, the emotion you get through playing rather than putting a human eye there is completely different. So anything connected to emotion is there, but in business it works differently. Is there any area you see potentially which can be completely autonomous?
- [49:44]
A
I'm talking right now about 2026 and the markets and technology is moving very, very fast. And according to our forecast, what I'm right now looking right now and you this up updating this forecast every week, full autonomy level 5 autonomy will be beginning of 2028. I would, I would really say it will be Pareto Distribution. So 20 80, 80% of the companies and businesses you probably can run then completely autonomously and 20% probably not. And this number will change with HEI and superintelligence, what is coming and so on. So
- [50:35]
B
no. That's interesting and thanks for answering that. I want to geek out on the book Zero to one and you already have briefed a little bit about it, but let's expand on it, let's build upon it. So your twist is that AI made one to end basically free as long as the value is in 2. Or I should say so the value move from 0 to 1. Walk me through what that means for anyone building right now.
- [51:05]
A
The first point is about probably love. Because this is exactly what I love from zero to one. This is what I'm doing for 20 years already. And this is excites me as most. Not the scaling the company, but from building from nothing, something what's working and what can code generate revenues and so on, so on. So this is my personal opinion and this is what makes me happier in my job and, and this is what I probably never stopped to do in my life. Yeah, so this is the point and the second about your question. All talking about pains and needs and validation. But every company, every serious company has an MTP like a massive transformative goal. And our transformative goal is to say they are these guys, operators which know exactly how to build a business, how to sell, how to grow. But most cases they are not on the cap table, most cases they are not getting all these rewards and especially exit they are probably not even included in the exit. This is the MTP of brains. We want to make operators to owners. We want to reach the goal what every operator, every hero founder who has this experience can just build his company. We will help him to build the right infrastructure, right product, scale the company and so on. And this is exactly the space from zero to one where this her founder and operator sees he can feel his industry, he sees this specific problem what is not solved and probably he's even solving it in this company as employee but has no resources to get out of the company and to start building his own stuff. This is our offer to the guys who want to stop building companies of other people and want to build their own.
- [53:36]
B
That's amazing and it should motivate a lot of young people who have ideas, who can imagine and have the ability to execute like you said, operate and building is one part they can forget about it connecting and partnering with your team and focus on what's the idea, how do they operate. And you know scaling will come free like you're saying because once you build that 0 to 1 the foundation of it using the govern AI the scaling will become automatic is what you're saying. And that's where you are moving the era on this agentic startup and product development. And kudos to you. I have a parallel story. So I worked a lot with Microsoft, represented them in my past life and every decision we made at the very start was zero to one moment. If you see Microsoft growth path it's always about that, right? They set up the office365, you know in the competitive world with Google or this setup. The I've done data and AI for 30 years. I have a patent in AI as well. But Microsoft came into the game pretty late. But then they did everything right, you know, targeted the mid size and small businesses. They can make use of data. A lot of foundation is being Built on that data platform. If you don't have it, AI doesn't exist for you. AI can only do the things that it can read from Internet, but what about your own data? And that's where Microsoft did things right. And that actually taught me something, that your protocol layer matters the most. You know, the foundation layer matters the most. Then you can build upon it as the foundation of the house, as we would say. And then you can build X number of N number of buildings on top. And that's something what I can relate to it. I'm coming towards the closer. I have one thing where the audience can make use of it. So last thing, five concrete moves a listener can make Monday morning to start building the accountable way. Go.
- [55:46]
A
First point you have really to define what is the final result for the agent. Because without this final result, without this acceptance criteria, you will never understand if it was built the right way or not. So this is, this is how we start building every whole project starting with acceptance criteria. And this is what, what must be done in every task just to start the next one is as I mentioned, decide design system for your whole product, for your whole startup is the must. You will just save huge amount of hours for your work. So build a design first before you start coding and it will save you time, money and energy. A third point is all about evidence. How can you prove what this specific task was finished by this agent five years later, ten years later, what is, what is the audit file where you can check what, how this software, this huge product was built, which test scenarios you have found on it and where it's saved like forever so nobody can change it. The next point is security. If you want to build AI based software, always integrate your security agent in the project. Because otherwise you are just one of this like 40% of startups which are building with AI and have security issues in production code and you are just on risk. And the last point, never trust the agent who is telling you it's done. You have to see the evidence. Our agents for example, they are saving screenshots of every functionality. What is done, what later is functionality is approved by our senior developers. And then we know exactly, okay, this is what was finished. Our guys approved this and this is what can launch live. Otherwise it will be just a fight against the bucks and great five pillars
- [58:06]
B
you've given to them to get started right this Monday and be motivated. Alex, thank you for this 60 minutes. The line I'm keep coming is that speed is no longer a vote. It entices me and accountability is. If you want to follow Alex's work, go to brnz.com Dude AI and also find him on LinkedIn. He has a couple of things for you there which we will put in the show notes as well. Subscribe to Think AI Podcast wherever you listen. We have real stories, real systems and real AI. I'm Dave Goyal. See you next time.
- [58:42]
A
Thank you Dave.
- [58:44]
B
You have been listening to Think AI Podcast with Dave. Take one idea from this episode and turn it into action.