Beyond the Agentic AI hype with Superbo AI

0:00Welcome back to The Serial Entrepreneur. I'm your host, Anna Wood. In today's episode, I'm joined by Demetri, the founder and CEO of Superbo AI. We discuss the importance of ignoring the hype and finding the real value in agentic AI, why sorting data is the first step in ensuring success in automation, and the hunger and drive that makes an entrepreneur. Let's dig in. Podcast today, how are you doing? same i'm fine i'm perfect thanks so much for having me yes thanks for coming on so we begin every episode with our signature icebreaker question which is what's your favorite breakfast cereal and why okay when i used to live in the u.s my favorite one was the um the the fruit loops with different colors.

0:53I like the taste. But now I don't do cereal anymore. So the last thing I remember about cereal was when I was in the U.S. But that's the best I can do when it comes to cereal. I think they are too sugarly for me and I'm trying to remain fit for some reason. And this is something that you think when you're in my age but when you are in your early 20s you don't think about it. I think a lot of people also tell us like, oh, we don't eat cereal anymore. But when I did, I'd always go for the really chocolatey or really sweet ones. So would you be able to introduce yourself and tell me a bit about your background?

1:33My name is Dimitri. I'm the co-founder of Superbow. But I've been around many more years before Superbow. So my background, well, I started as a professional musician. Yeah, I know it sounds very unorthodox. but I studied professional music composition and then down the road I also had the idea how I'm going to bridge my musical talent to business and the main reason was I was trying to make a living you know, when you're in the music you can create marvelous things maybe things that you love but only you or maybe somebody will love only if they have the chance to you know, you have the chance to expose them in front of the proper audience.

2:22So very early, I understood that I had to make a living. So I, you know, I went to business, I graduated, and then I started working around business with media tech companies. Eventually, after some years, I had my own company in 2011, which was a mobile advertising company, which I sold in 2019. And then I said, you know what? 2019, very young to retire. So I probably have to do something with my time and my money. And I bootstrapped Superbo with my co-founder. Here we are now, you know, five plus years. We have Superbo. We're doing pretty well. Pretty happy about it. That's my plan. Great. Could you talk a bit more about Superbo?

3:06What do you do there? All right. So we've been around, you know, Superbo has been around since 2020. It was before the AI trend and before the AI revolution. So traditionally approaching things around machine learning and natural language processing and understanding, NLU, NLP, all this kind of stuff. We always thought and we believed confidently that AI revolution would come. We didn't know when, but we knew it would come. Get what? Two years down the road, we were all introduced to OpenAI and the rest of the players. And so that was the tip-off, right? When everything started and we had already built an asset critical enough to allowing us to start harnessing the available technology.

3:55Of course, since then, many things happen and they keep happening on an exponential rate and we keep building. So we are asset builders. Today, Superbo is an agentic AI company by the full definition of agentic. The difference is that we don't do AI agents as a standalone task force, but we do the whole agentic system, which is comprised from many different AI agents. that communicate and talk with each other and they reason with each other and they observe and they plan and they design and they execute. So they are autonomous, autonomous for the workflows that you want them to embed, to be embedded in, right?

4:45And this is extraordinary because what we are achieving today, it is we get the work done in terms of, is it just another, you know, fluff of AI agent as a single task oriented tool? It's not a tool anymore. While we do, it's a brain that has the capability of executing all the intelligence within a workflow in diverse sectors, but also in diverse divisions inside an enterprise, starting all the way from human resources, which sounds simple. It's not so simple, but it is simple all the way to customer support, procurement, marketing. It's been there, done that. Now, it is also very important that we're one of those companies that along the way, we were lucky enough or we had the capability to put some of those into production.

5:41And demos and beautiful PowerPoints have nothing to do with production, right? So we have the production scale deployment when it comes to Agenda AI. So this is what we do. We keep building. We believe that. Now, it's also worth mentioning that we are all in a pre-inflection market. So it means that the Agenda AI adoption is not here yet. It will be. We also know that for sure. Once again, we're confident it's coming. We don't know when, but my wildest guess would be in the next 18 to 24 months. are going to have the first enterprises taking very seriously how they deploy and they adopt Agenda AI.

6:23So this is what we do. Great. So as you said, Agenda AI isn't being adopted on a large scale currently. So it still to some people sounds like an abstract concept. What does Agenda AI currently mean in a business context? Well, it has to do with the business. Agenda AI in any context should mean one very precise thing. Agenda AI is the squad of different agents communicating with each other, reasoning with each other, and they are goal-oriented, not task-oriented. So they have to achieve a goal. This is very important because when you build a business case, an agendic workflow, whatever, you're doing it for a purpose, to reach a goal, to resolve something, to accelerate something, to make sure that something works better, faster, safer within the enterprise, achieve AI sovereignty, and make sure you get that goal done.

7:29Now, this is also very important because most of the enterprises out there, they have no idea whatsoever, A, why they need authentic AI. B, if they don't know, they cannot really see the value. So it is important, first of all, for the enterprises who are the buyers of that to understand, clarify what they're trying to solve and what is the value they're trying today. Because I can tell you from the available technology, Superbo Headless today, we can bring value day one, measurable value. I mean, with a very solid return on investment. You pay X and you can have X multiple times back as a return because you're solving something that requires a lot of human capacity, effort, time, money.

8:20And don't take me wrong, we're not here to take people's jobs. On the contrary, we're here to make employees' jobs a lot more productive, faster, safer, and not so complicated. They have been wasting so much time in arguing, in fixing problems, in spending time counterproductive. You get tired at some point. So instead of having this, you know, there is a hook in the market saying, ah, we're going to be replaced by AI. No, you're not. You're not getting replaced by AI. What's going to happen, though, is that people who are going to use an enterprise, who are going to use your genic AI properly, are going to replace you.

9:06Not the technology itself. Definitely. I think that's a sentiment that's been echoed a lot around the tech industry. So there's currently a gap. Yep, there is a gap. There's currently a gap between experimentation and actually operational AI. Where are companies getting stuck? This is the$1 trillion question. So this is the curse and the blessing at the same time. Okay, let's go, let's, you know, let's see from a helicopter view why enterprises have been choosing to experiment more with a proof of concept and pilot projects instead of going with full deployment in production. Again, A, they have no idea why they need AI.

9:59All they know is that they need AI, right? So it is more like a marketing stipulating economy saying that, hey, you know, you're the CEO of XYZ company, you have a CIO or CTO or maybe chief AI officer or data officer, and you have no idea what to do with AI. All you need to know is that you need to make some announcement saying that you're harnessing AI power and now that you are AI powered. I have no idea why on earth you need that. So, and that creates the first level layer of noise. The second level layer of noise is that they invite multiple ventures and vendors for POCs. and at the end of the day they end up with 50, 60 different POCs and maybe a couple pilot projects and they consider these to be an experimental phase and they fail to understand how much money, energy, time, effort, work they spend on that without the value being on the table because value will never come from POCs.

11:15It's totally different. So instead of going out there and you know, nailing down with, come up with a case, a use case or a workflow that you really need to solve. You really need to boost in terms of production and accelerate. Instead of finding a vendor that has at least one Agending AI in production, this is the best you can do. Find someone like Superbo that has Agenda AI in production. Now, Superbo has done a little bit of exaggeration to tell you the truth. We have our own Agenda AI framework as well, which means that we don't use any hyperscalers for that. So instead of going to Google Vertex and say we're going to build our own AI agents there, which is fine, which is okay, we have our own framework because we always believe that the hyperscalers, as you know, they build horizontally.

12:16So eventually, when you build an AI agent on their agility AI framework, eventually you're going to hit a wall. Eventually, you're not going to be able to do a lot more of bespoke tailor-making customization. And it makes sense for them. And that's why they are hyperscalers. So we never wanted to depend on that. We invested really lots of money, time, and effort to achieve that. So we have our own framework, building our own Agenda AI systems and agents, of course. So the gap is here. The gap is the tick in the box that you need to make sure that you're still in the pre-inflection market. The moment the gap will start diluting, it means that you're entering the post-infaction market.

13:07What do you think are some of the best use cases of Agenda KI that can actually provide that value to a company? For sure, use cases like procurement, customer support, human resources, marketing, finance, would be lowest hanging fruit there. Today, you have fantastic RPA systems out there or workflow systems, and they do marvelous work. but they're not AI. So if you really need, first of all, you need to make sure that usually the gap in enterprise, the natural habitat of a gap is where AI can really come and provide value. Otherwise, it's just noise, right? So if you do your job with perfect RPA tool and you're happy, then you don't need Adjernic AI not yet right but if you want to scale faster if you're gonna be you know part of the Adjernic AI era if you wanna boost your employees morale and productivity and also understand that they're you know they're not gonna lose their job then you need to come up with very specific parts of your workflow that would need an Adjernic AI as a node many people they make a very common mistake.

14:32They say that, you know, AI will be replacing workflows. No, no, no, no. A general AI will be adding or replacing nodes in an existing workflow, not the whole workflow. You have fantastic, you know, software out there and companies that they do marvelous work when it comes to workflows. And we're not competing with those guys. We are just complimentary to them, maybe on top. It's adding value on top of them. So this is also one more part to be misinterpreted and misunderstood. And last but not least, there is a very big attention deficit. People don't listen. They just read a lot of marketing, jargon, passwords, LinkedIn.

15:18And they come out and they say, oh, you know what? What we need is this. And out of courtesy, you remain silent because you don't know how to say to those guys that, hey, you know what? This is not exactly what you need. This is what they have told you to say. This is what you need. So there is a way, there are so many sections and departments in an enterprise that could benefit from engineering AI today, like human resources, procurement, finance, customer support, marketing. And yet, who has the ownership? Now, this is the question. In my humble perspective, there should be not one person. It would be one person per department.

16:00Who is the CMO? Chief marketing officer. This is you. Okay. You're going to be responsible in knowing exactly what you want. And then you come to us. We are the experts and we can walk you through and tell you and recommend. The same goes for the CFO for the finance. And it goes on. And then all of these can be probably for bigger enterprises. that can be orchestrated by the chief AI officer who can understand tech. And he has a tech team or an AI team who can evaluate your technology at the end of the day. But when we talk to businesses, we don't sell technology because they don't understand technology.

16:43We sell the work. We sell what we can do for you and make your work simpler, faster, safer. Yeah, great. So when it comes to deployment of these agents, what do you think tends to break first when companies introduce autonomous agents? Security. Security boom, yeah. It's been heavily underserved and undermined most of the companies. Let's just see two steps back and see the reality. You have 95 % of the Avengers out there are LLM wrappers. and it feels simpler, faster, to go to market, they can grab 20 logos by that. Maybe, you know, with 20K or 30K or whatever, they can deploy something very fast.

17:32But that does not really mean enterprise production with all the security and the sovereignty that is needed those days. And you know very well that, you know, enterprise sovereignty has been in the top of the town at least throughout 2026, especially in Europe, and how we should not be dependent by non-European companies and vice versa. You cannot get this from an LLM wrapper. This is another job. You need to have all the layers of security, and you also need to have the foundational layer, your abstraction layer. How are you going to deal with unstructured data? I can tell you something. Even the humongous enterprises, Fortune 500 companies in the U.S.

18:18today, they have data structure deficit problems. Unstructured data, chaotic here and there, some on PDFs, some on CSVs, some on ERPs, CRPs, whatever here and there. And then probably a couple of years ago, they had a project to structure the data and they did it successfully. And then again, they let it go down. And it's a vicious cycle. So you also need to have a solution on how you're going to handle the infrastructure data. Because this is the beginning of no project. So instead of thinking, who's going to compete with me? I think your biggest competitor is a no project because they have this stumbling block, which is our structure data.

19:05And that's why in Superbo, we still invest a lot on our abstraction layer to make sure that we structure the data. We're making a favor to ourselves, right? We structure the data in order to make sure that our agenda AI systems can rely on those data and produce the best possible outcome. Yeah, and I was going to ask that, what does good AI architecture look like inside of a scaling company? And the second part of that is, if a company is building right now, what should be done from day one in order to be ready for autonomous agents? From an architectural perspective, you need to build layers.

19:50We always thought in four layers in Superbo. And the cleaner the layer, the better it is. So you need your foundation layer, your abstract layer, your knowledge layer, your agents. And all those should be independent from any LLM and from any cloud. It means that you should not rely to a model. Some models are better in reasoning. Some others are better in graphics. Some others in analytics. So, for example, we are LLM agnostic. We use different LLMs whenever we want to achieve something different. That's number one. Number two, we're cloud agnostic. So we can deploy on any cloud or our cloud or the enterprise cloud, or we can even deploy on premise.

20:43So this is also something important. Architecturally speaking, you should not be locked in with a hyperscaler. Because let's see the reality. some people they say okay well i can get some of the pre-built agents of i'm not gonna say a name but of xyz hyperscaler who is probably a leader in crm yeah okay but these agents are perfect to make sure they work and they boost their existing core offering how about your agents The whole difference, I think the simplest way to put it with definitions of 60 years ago would be if you're going to be renting your house or owning your house. So are you going to be renting your AI or owning your AI?

21:37What we are trying to solve here is that we offer enterprise or enterprise, large enterprises to own their AI. And from a security standpoint, this is very important. So yes, architecture has to do with security. And so bridging your previous question with this question. And your next question was, what would be the first thing to think for if you are building a new AI venture or if you are a potential buyer? Yeah, and how to be, if a company is building right now, what should be done from day one to be agent ready? Okay, that's a question. All right. you need your data foundation okay so you need to make sure you put some time and effort and invest in your data invest in your data structure before rushing to get on another experimental poc or eventually manage to give a one-year project to a successful vendor like superbo hey you know what take a step back and you have to tidy up your house, right?

22:50Your foundation needs to be there. Your foundation is your data. Otherwise, you know, in Superb, we have solved that because we have the abstraction layer. But then again, you need to make sure that your data, it's because you need to have your data structured. So take some time and you have to knit up your house and your data And then whoever you know, then it's easier to also choose your vendor. And also something very important. Go back and see how people, what the vendors and the AI vendors talk about and what they post about, right? They post. You see, they post something a couple of months ago and two months down the road forward, where they were going to post something totally different, just because it is a better fit for the current trend and narrative.

23:51But you need to have an opinion, and you need to stay solid. Otherwise, you're just going with the trends. And this is where things can really go south. So follow what the sea level says, what they post about. They're trying to, but I understand there's lots of noise. So, yeah. So fix your data first and then go shopping. So is there a way that companies can kind of try and move away from, I guess, the trends and really know what will provide value for them? I think, yeah, there is a way. A, fix your data. If you have your data structured, you know exactly what's going wrong. You have an idea.

24:44You're not speculating, right? You know exactly what, where, and what is going wrong. Before us, you don't need superb for that, right? You can do it by yourself. So this is 50 % of your success story. There is a 50 % since you know where, when, what is your biggest issue, challenge, or your biggest deficit in a specific department or division of your enterprise. Then you know exactly what you need to ask from this particular vendor and say, hey, you know what? We see that we have a humongous deficit on our customer support. They pick up the phone and they're 24-7. They have no idea what to say.

25:27They cannot resolve. people are not happy, your net promoter scores, our NPSs, we need to fix that. We understand, we have the data, right? Or, you know, we're having a huge trouble onboarding new employees, so our human resources is having trouble with X, Y, Z. We need to fix that. So with specific use cases, and they will also save you a lot of time back and forth. So when a potential prospect sits on the table with us and they have done their homework and they have structured the data and they know exactly what they want. This also enables us to unleash our firepower brain to upgrade the whole, not only the experience, but to upgrade their own goals and help them see.

26:21so it is a collaboration and they will never be out of you know out of the box or off the shelf it's not going to happen right so you need to work you know your enterprise so yes fix your data then you know what on earth is going on definitely so when it comes to industries that are quite regulated If there's a company who's looking towards wanting automation and agentic agents, are there any considerations that they have to take before they deploy these systems? Yeah, sure. Industry like BFSI or banking, healthcare, these are sensitive industries with very sensitive data. So first of all, they need to make sure that their data remains sovereign.

27:14It means there is no leak to any large language model, hyperscaler out there. So you need to make sure the vendor you choose is capable and able to make sure that it will work with the same results in terms of outcome and quality without leaking your data into the model. This is very important. The other thing is that they need to make sure they have the security infrastructure in place by themselves and then allow the company to come and fit on top of that and make sure that again there's no leak there's no um when it could let me give you an example it's if we get your data your name your last name your tax number your health file or your whatever and we look into the llm for a fast answer just because we cannot anonymize it and mask it, then you know that beforehand, the LLM has your data for training purposes.

28:20You know that. And you don't know, you really don't want that, right? So see what's happening right now with Universal Music and Suno, right? They're trying to find a way forward. What's going to happen with the training data? And you know what? Suno is making a fantastic work. I have lots of experience with Suno. They are phenomenal in what they do and what they offer. Now, the next iteration of music business, for example, I'm taking music because this is maybe sounds very familiar, or the movie industry, right? So you have many companies like Runway or MidJourney. They have been training on movies data and actors and voices.

29:11Somehow, this data got leaked out there, right? Nobody went out there to intentionally buy the data. There was no legal framework in place supporting that. Somehow, your data, one way or another, leaks. Now imagine if you are an insurance company and your vendor leaks the data of your clients because he needs the power he can get from the model, from OpenAI or Cloud or whatever. Then even unintentionally, your data is leaked. So you need to make sure that you can work with a vendor who can offer a sovereign solution to you. So a sovereign solution means it's exactly the solution I described of what we offer.

30:00This is the only solution, architecturally speaking, that you need in order to make sure that you're ticking the box when it comes to heavily regulated enterprises. Otherwise, it's not safe to do so. It's not even safe to have a POC. Mm-hmm. Yeah. So I think earlier on, you mentioned that it's going to be within the next 12 to 18 months that you will really start seeing companies make the most of agentic AI. What do you think the next couple of years will actually look like for the companies that get it right? The best question in the industry is this question. It's this question is, it requires a very simple answer for all the companies to listen.

30:47All right. So the trailblazers in terms of small, medium enterprises or large enterprises who are going to finally adopt and put into production mature agenting AI systems. Not just one AI agent, but a whole system of agents, okay? And they put it successfully. It requires many steps to be successful. Besides choosing the right vendor, as I said, your data is very critical. Your data structure. The ones who do that are going to be the ones outperforming with a massive difference from the ones that don't. the rest of the companies are going to be dead in the water. They're not going to be efficient at all.

31:41They will not be able to compete with the rest of the companies in terms of pricing, in terms of production efficiency, in terms of speed, in terms of cost, the cost elements. So the difference is going to be, in very simplistic terms, trying to travel from New York to London on a boat compared to a commercial airliner. That would be the difference. And it would be huge. And the more, because you said in the next couple of years, okay, I'm sure this year we're going to see many more surprises. I'm not going to be surprised if, you know, very big ventures in the U.S. achieve AGI probably by the end of the year or early 2027.

32:45So the available technology out there, it keeps becoming more powerful down the line, down the road, right? So, but the rest of the industry and humanity is not absorbing that at the same pace, which is understood, right? But the pace of the tech pace in terms of progress is multiple times faster than the adoption and the digestion, you know, pace of once this begins, then you're going to see. I'm sure we're going to see comments that we know today. They are probably leaders and they might be wiped out overnight. Right. So I think we've also spoken a lot about value and the value it can give to companies.

33:38And I think there's been a lot of talk about the time. Time is money and there's going to be the rise of the time economy. People are going to have more time on their hands. What do you think that looks like in practical terms for a founder that is running a business? As I said, if you want to save time, you need to focus on getting one thing right. Get your data right. Structure your data. This is 50 plus percent of your success story, I can tell you. Everything else is just jargon and passwords and very nice, fluffy, bleacher, LinkedIn posts and videos and beautiful saleswomen doing whatever.

34:25If you don't have your data factored, you're dead, my friend. This is the end of the story. Do something else with your life. This is time saving. This is cost saving. This is your insurance guarantee of achieving value day one very fast. I'm not saying it's easy. I'm saying it's absolutely necessary. You know, we have... I think this is the only year, talking to investors and to prospects, that people have started understanding, a glimpse of understanding, that, you know what? And I say, we're not ready for you guys. I say, oh, okay, please elaborate on that. And I say, we don't have a structure yet.