In this episode, Michael Bernzweig talks with Demetri Papazissis, co-founder of Superbo AI, about the complexities of achieving sovereign AI deployment. Demetri shares his journey from the early days of AI development to navigating the current landscape where AI buzzwords are everywhere. He discusses the challenges of educating the market and the importance of building an infrastructure for accountable AI execution. The episode also includes a live Q&A panel from the Software Oasis AI Bootcamp, offering real-world insights into AI deployment.
**Introduction**
- The inception of Superbo AI before AI was trending
- Challenges of being a first mover in AI
- Superbo's journey to navigate AI buzzwords
**Interview**
- Demetri's career path leading to Superbo AI
- Evolution from mobile marketing to AI solutions
- The blessing and curse of being first in the market
**Presentation**
- Importance of sovereign AI deployment infrastructure
- Pre-inflection market challenges
- Superbo AI's bespoke AI agent framework
**Panel Q&A**
- Audience questions on AI deployment strategies
- Discussing AI execution and compliance
- Real-world applications and challenges in AI
**About Demetri Papazissis**
Demetri Papazissis is the co-founder and CEO of Superbo AI, a company focused on machine learning and AI infrastructure. With a background in mobile marketing and a vision for AI's potential, Demetri has navigated the complexities of early adoption in the AI field.
[MICHAEL]: So, imagine co-founding a company in a field that wasn't even trending yet... that's exactly what Demetri Papazissis did with Superbo AI. Back in 2020, AI wasn't in fashion, but Demetri and his team worked on machine learning and natural language processing when others weren't paying attention. Fast forward a couple of years, and suddenly AI is the talk of the town. But here's the catch—while everyone started throwing AI buzzwords around, Demetri found himself having to explain why Superbo's solutions were a blessing and a curse. By the end of this episode, you'll understand exactly why being a first mover in AI is both advantageous and challenging. I am your host Michael Bernzweig and this is B2B Automation Spotlight.
--- INTERVIEW ---
[DEMETRI]: I know for a lot of individuals that are listening in today, the journey to entrepreneurial success is never a straight path.
[DEMETRI]: Sometimes there are a lot of bumps along the way.
[DEMETRI]: So I was hoping that you could share a little bit of your journey prior to founding Superbowl, and then maybe we could chat a little bit about the AI solutions and clients you're working with today.
[DEMETRI]: Yeah, sure.
[DEMETRI]: Thanks for the opportunity.
[DEMETRI]: It gets interesting.
[DEMETRI]: I think I started working around 2001 on my early 20s, and, you know, in the, always other business and sales, started, you know, selling time-sharing with RCI, and then moved to construction companies in the Middle East.
[DEMETRI]: And then eventually I joined an American company doing mobile advertising until all the way to 2011.
[DEMETRI]: And that was a time that I think I convinced myself that maybe I should try something, build my own venture, and this is exactly what I did.
[DEMETRI]: So fast forward, I co-founded a venture in 2011 that it was around mobile marketing and value-added services for mobile carriers.
[DEMETRI]: It was very, you know, it was a very good time for something like that in the market.
[DEMETRI]: The market was very mature.
[DEMETRI]: It was past inflection.
[DEMETRI]: It was a cash cow sector, lots of competition, but also, you know, we had a lot of experience, a very good network.
[DEMETRI]: So we managed to exit the company in 2019, about eight years down the road, and we sold it to, as a portfolio company, to a German fund, and that was a good thing to do.
[DEMETRI]: And, you know, that was a time, we're talking about seven years ago, 2019, I thought I was too young to start buying fancy cars or houses, and I needed to do something meaningful.
[DEMETRI]: So I co-founded Superbo with my co-founder, who's the CFO also, Agis, and we co-founded Superbo, having as a vision to tackle anything around machine learning.
[DEMETRI]: Let me just put, refresh our memory.
[DEMETRI]: 2020, there was no hype.
[DEMETRI]: There was no trend.
[DEMETRI]: AI was not in fashion.
[DEMETRI]: It was not trending.
[MICHAEL]: No open AI, no cloud, nothing, right?
[DEMETRI]: So we were doing machine learning, NLU, NLP, for those that don't know, NLP is the Natural Language Processing, and NLU is the Natural Language Understanding.
[DEMETRI]: So we were trying to tackle with different kinds of solutions and R&D development, this kind of work.
[DEMETRI]: Now, we were lucky enough, because we had the network from the previous experience, to start, you know, proof of concepts and little pilot projects in large enterprises.
[DEMETRI]: And we were building something meaningful.
[DEMETRI]: And we always thought that, you know, AI will, at some point, kickstart as a revolution.
[DEMETRI]: We didn't know when.
[DEMETRI]: We just, you know, always knew it will happen.
[MICHAEL]: And actually, two years down the road, November, 2022, you know, open AI came in our lives, and people and individuals and big firms and small firms, enterprises, they started realizing that, hey, well, wait a minute, right?
[MICHAEL]: Something's going on here, right?
[MICHAEL]: And actually, something was going on.
[MICHAEL]: That actually, that was very true.
[MICHAEL]: And for Superbo, it was already a two-year company, right?
[DEMETRI]: So we had been, you know, already through growth pains and very difficult situations.
[DEMETRI]: And every time, I fine-tuned the strategy of how we move forward and how we can harness the technology.
[MICHAEL]: And I think the profound difficulty was always how we can educate the market in what we do compared to what was offered by, you know, large language models like OpenAI or Cloud or you name them, right?
[DEMETRI]: It was, at the beginning, we were trying to convince, you know, our clients that it's all the same thing.
[DEMETRI]: You know, these are not competitors to us.
[DEMETRI]: These are the best thing it could happen to a venture like Superbo.
[MICHAEL]: Because, you know, the first questions popping up of the late 2022 or very early 2023 was like, why on earth do we need Superbo?
[MICHAEL]: And we were like, what?
[MICHAEL]: I mean, you really don't grasp that?
[DEMETRI]: I mean, so, but then again, if you really put yourself in their shoes, it's something very important.
[DEMETRI]: It's very new.
[DEMETRI]: They understand the gravitas and the gravity of the new thing.
[DEMETRI]: And, you know, I cannot blame them.
[DEMETRI]: It was very hard to understand that.
[DEMETRI]: Now, that point created the first, let's just say, trend.
[DEMETRI]: So everybody started talking about AI without even understanding what they're talking about.
[DEMETRI]: They started throwing buzzwords here and there to make sure that they remain relevant, advertising companies or integrators, they started saying, hey, we offer AI solutions without even knowing what they are trying to frame, what they're selling.
[DEMETRI]: And so this, you know, it compiled to a super noise around the world, having different kinds of geographies, different kind of pace, different kind of organizations, not being able to understanding what they need or what they are talking about.
[DEMETRI]: And so it was, for us, Superbo has been a blessing and a curse.
[DEMETRI]: Blessing because we, you know, we were one of those first movers in the market.
[DEMETRI]: So we're here and we harness all the new technologies at the same time.
[DEMETRI]: We're building our asset.
[DEMETRI]: We keep building our asset on a weekly basis.
[DEMETRI]: But the curse is- There's a lot to be said for being first, you know, is a huge, huge advantage for your clients.
[MICHAEL]: Well, yes, for the ones who understand, but then again, as I said, it's also a curse, right?
[MICHAEL]: So why a curse?
[DEMETRI]: Because we are in a pre-inflection market.
[DEMETRI]: So imagine AI, you know, organizations, they're not really sure what to expect.
[DEMETRI]: And even the ones they do, they expect the wrong thing.
[DEMETRI]: So they expect that, you know, a vendor will walk in the room with a magic wand and make sure that everything's gonna work fine.
[DEMETRI]: And you're gonna feed your AI with crap and you're gonna get spectacular AI.
[MICHAEL]: Well, guess what?
[MICHAEL]: It's not gonna happen, right?
[MICHAEL]: So you feed it crap, you get crap.
[MICHAEL]: And it's not about the AI.
[MICHAEL]: AI, it's not stupid.
[MICHAEL]: It's not that we're lacking confidence or intelligence, that we're lacking understanding and ownership from the organizations.
[DEMETRI]: And at the same time, we also need infrastructure to make sure that we can put into play different vertical AI agents that can really work in very intense production environments.
[DEMETRI]: And, you know, we're one of those guys that we have deployed in real production environments.
[DEMETRI]: Today, we have around 6 million different unique users on our products out there.
[DEMETRI]: And, you know, they're battle proven, we understand that, and we know it's not easy.
[DEMETRI]: And, you know, Michael, we're still in the pre-inflection market.
[DEMETRI]: So at some point, one or two firms or enterprises from each different sector will decide to deploy vertical AI agents in the correct way.
[MICHAEL]: I don't know who's gonna do it first, but eventually someone's gonna wake up, and say, hold on a sec, right?
[MICHAEL]: We're doing this all wrong, right?
[MICHAEL]: So it's not about going live within a week.
[MICHAEL]: It's about doing it right.
[MICHAEL]: Take your time, make sure that you can, you know, use the available infrastructure and deploy safely in production.
[DEMETRI]: And also make sure that I work with a vendor or a venture that can really offer me proprietary technology, not offer from hyperscalers.
[DEMETRI]: I understand that venture like Superbo might choose to go and grab the small business or the small office, home office business, and all these guys that don't really need the bespoke AI agents.
[DEMETRI]: I understand that.
[DEMETRI]: So that, you know, it makes sense from a commercial standpoint, that they would need to use an agent AI framework from a hyperscaler.
[DEMETRI]: But when it comes to real enterprises, maybe your real, you know, workflows and your policies and what you need to achieve, then you need something for you, period, not something that is pre-built or out of the shelf.
[DEMETRI]: And so now we are, I think we're all experiencing this turbulence, which is just before the pre-inflection market.
[DEMETRI]: And so the, I'm sorry, before the inflection point.
[DEMETRI]: So I think if you speak even to the big guys in the U.S.
[DEMETRI]: right now, they're gonna tell you, you know, Michael, inflection point is not here yet, but it's coming very soon.
[DEMETRI]: We don't know exactly when, but it's coming.
[DEMETRI]: When the adoption will actually hit the inflection point, then, you know, I think, you know, not only Superbo, but a lot more ventures will start claiming their natural territory and position in the global playground.
[DEMETRI]: And then this will go on and on and on.
[MICHAEL]: You know, it's like the Oklahoma rush, right?
[MICHAEL]: The United States government, they gave a banner with a horse and they said, okay, guys, go on, claim your land, claim your land.
[MICHAEL]: This is exactly what is about to happen right now.
[MICHAEL]: And let's see.
[MICHAEL]: So here we are and we'll keep building and we keep selling and we'll keep piloting because right now the appetite is on the experimental level, right?
[MICHAEL]: So lots of POCs, lots of mini pilot projects, lots of experimentation, but not real production ones.
[DEMETRI]: We were one of the lucky ones that we do have real production projects.
[DEMETRI]: But then again, these are not enough.
[DEMETRI]: We need to hit the inflection point.
[DEMETRI]: Let's think about that.
[DEMETRI]: I'm very optimistic, by the way.
[DEMETRI]: I'm sure it's gonna happen.
[DEMETRI]: I'm not sure when, but I'm sure it's gonna happen, yeah.
[MICHAEL]: So Dimitri, let me ask you, when you originally brought the solution to market, what was the challenge that you were trying to solve for in the enterprise space that you felt had not been addressed?
[DEMETRI]: Well, I think we started in one of the most difficult divisions in the very big enterprise, which is the customer experience and the customer support.
[DEMETRI]: So we started working with tier one mobile carriers.
[DEMETRI]: They have millions and millions of subscribers.
[DEMETRI]: So what we tried to do, and we did it in the most successful way, was to make sure that we can offer a very smart 24-7 AI customer experience, a customer support agent, that it can actually reach the point that will resolve autonomously without any human in the loop.
[DEMETRI]: Although by regulation, you need to have a human in the loop.
[DEMETRI]: However, the desirable outcome is to maintain the end user inside your AI agent and try to resolve that.
[MICHAEL]: So we were trying to resolve different cases, right?
[MICHAEL]: So people who forgot their PIN, people who need to deactivate a kind of service or a subscription or things that usually will consume lots of physical agent call center time.
[MICHAEL]: And all these call centers, they cost millions of dollars on a monthly basis.
[DEMETRI]: So we try to make sure that we optimize and autonomously automate that part.
[DEMETRI]: So we started from the customer support.
[MICHAEL]: And I think it's one of the hardest because this is also client facing for the organization, right?
[DEMETRI]: I think it's a lot easier if you go in internally and say, okay, we're gonna do like a human resources kind of AI agent and see how we can apply your HR policies and help your employees grab their vacation time and interact without consuming the HR director's times, blah, blah, blah, or a procurement AI agent.
[DEMETRI]: I think these are a lot easier, not from a technological standpoint to deploy, but from the impact.
[DEMETRI]: The impact is internal.
[DEMETRI]: But when it comes to customer support, and I think it was a profound opportunity that we took, customer support, whether we like it or not, it is customer facing.
[MICHAEL]: So whatever you deploy as an organization has an impact real time on your end user, your user, regardless if you are a streaming company or if you are telco or you have end user or a bank, right?
[MICHAEL]: So all these guys that have millions of users, there's an impact for them.
[DEMETRI]: So we started from the very heart.
[MICHAEL]: Coincidentally, that was the opportunity, right?
[DEMETRI]: So we were off of the job.
[DEMETRI]: I love it, I love it.
[MICHAEL]: So can you share, and obviously you don't have to mention any specific clients, but some projects or successes over the years that were in the enterprise space that had surprising needs and surprising results when you incorporated your solution?
[DEMETRI]: Yeah, well, in most of the cases, we, as I said, we started with the customer support and the overall customer experience.
[MICHAEL]: So each and every deployment and each and every enterprise client is totally different and also in different jurisdictions, right?
[MICHAEL]: It is also worth mentioning.
[DEMETRI]: So I think one of the most successful cases we had so far was in an African country, one of the biggest ones, I think the biggest market in Africa right now, where we managed to absorb and complete autonomously without any human in the loop needed to intervene 84% of the cases.
[DEMETRI]: And now it sounds like a big nutter and some might say, okay, but it's not 94%.
[DEMETRI]: Then again, you need to see the 84% is humongous.
[DEMETRI]: I'm gonna be very honest with you, Michael.
[DEMETRI]: It's a very high number.
[DEMETRI]: And the reason is that you should expect your users to use the AI agent, not only because it is successful and easy to use, but also they need to get familiarized.
[DEMETRI]: One responsibility that we vendors have towards the market is educate our clients and prospective clients, but also we need to make sure that we have enough time and patience to educate the market of the end users.
[DEMETRI]: People, not all over the world are well familiarized or some of them are in denial phase.
[DEMETRI]: They say, oh, I'm not gonna talk with anyone.
[DEMETRI]: I'm gonna pick up the phone.
[MICHAEL]: Why?
[MICHAEL]: Because that's the way I have been operating for since forever, right?
[MICHAEL]: So it's also hard to push the end user to get that opportunity.
[MICHAEL]: So 84% for me is a very successful story.
[MICHAEL]: One, we're talking about millions of users.
[MICHAEL]: We're not talking about 150K, but millions of users that they keep adding up.
[MICHAEL]: Now, an interesting thing is that when we first started doing generative AI back in 2023, we were one of the few doing what we say, the rag deployment, which is now also a commodity, right?
[MICHAEL]: But back in 2023, it was a very big thing, right?
[MICHAEL]: At the very beginning, the first half of 2023.
[DEMETRI]: And then down the road, we said, okay, rag, it's very important how you retrieve the augmented generation using large language models, but everything's gonna be around AI agents.
[DEMETRI]: So let's start building that.
[DEMETRI]: And we pushed forward product-wise to have the first layer of this agentic framework, which was live in 2024.
[DEMETRI]: And we also had the opportunity and the luxury and the chance to deploy then AI agents within the same client.
[DEMETRI]: And what we saw a year after that, which was around Q4 in 2025, was that we saw that the end users started demonstrating a very specific behavior.
[DEMETRI]: They started choosing the agentic AI use cases where they can complete complex journeys, resolving different kinds of problems and trying to achieve the desired goal and over just having fun or having easy generated content.
[DEMETRI]: So they started from a behavioral perspective, choosing that.
[DEMETRI]: So now we put in a matrix, how the users have been behaving on the gen AI, like agentic rag that we do, and the agentic AI agents deployed to solve specific journeys.
[DEMETRI]: There is no comparison.
[DEMETRI]: The users go to the AI agents naturally.
[DEMETRI]: They have started feeling like home.
[DEMETRI]: So for me, this is also a signal that the market is about to hit the inflection point.
[DEMETRI]: Users understand that.
[MICHAEL]: They cannot really communicate that with their, you cannot call Verizon, right?
[DEMETRI]: And say, hey, I prefer talking to your AI agent because I can solve X, Y, Z number of problems within like a couple of minutes, very fast.
[DEMETRI]: And also understand why on earth you charged me 21.6 bucks more last month, which is something that they can do, our clients can do.
[DEMETRI]: But we can extract these results judging the behavior of the end user.
[DEMETRI]: So there is a clear, very clear indication that they have a very clear preference on using agentic AI and AI agents.
[DEMETRI]: Demetri, as we're wrapping up, let me ask you for any listeners that are listening in, they're working in an enterprise where they've probably rolled out all kinds of projects that have never seen the light of day over the last year or two.
[MICHAEL]: Can you share a little bit about your process working with a new enterprise client, starting with the early conversation all the way through to implementation, what that looks like?
[MICHAEL]: Well, we generally start off by inviting them to a meeting where we can demonstrate some of the simulations of use cases that we feel they bring value to the table, right?
[DEMETRI]: So if you are a leasing company, most likely we're gonna bring and we're gonna demonstrate what we can solve for you.
[DEMETRI]: It doesn't necessarily mean that we are touching the proper sensitive cords that you have in mind, but it's a hello meeting intro, and we can demonstrate telling them that, okay, listen, we're gonna work it through a couple of simulations to see how, and these are prepared like some days before the meeting, and let us demonstrate to you how we could solve some of your use cases by our AI agents.
[DEMETRI]: And this is how it looks like, because people need to see, they need to visualize exactly what's happening and how you're solving things.
[MICHAEL]: So we help them visualize from day one, the day that we say, hello, my name is X, Y, Z, right?
[MICHAEL]: This is the first meeting.
[MICHAEL]: So they understand the value.
[DEMETRI]: Now, if they can give us a lot more info of what their pain points, then we can have a follow-up meeting and demonstrate how we can solve the very specific pain points.
[DEMETRI]: So we discuss how the technology solves their pain points and bring value.
[DEMETRI]: We're not selling tech.
[DEMETRI]: We're solving problems.
[DEMETRI]: We're not selling prices.
[DEMETRI]: We are bringing value.
[DEMETRI]: Everything else, it can be so easily negotiable with a venture like Superbowl, because we're the owners of all the software from A to Z.
[MICHAEL]: So of course, there are third-party components like the large language model or the cloud, but these are standard for any deployment, right?
[MICHAEL]: So in most of cases, a large or small, large medium organizations, they have their own cloud instances and they also have their own subscriptions to specific models.
[MICHAEL]: It's not about the price.
[MICHAEL]: It's about, let me bring the value to you and solve something as fast as possible.
[DEMETRI]: So this is how we begin.
[DEMETRI]: And then we engage to a follow-up meeting.
[DEMETRI]: And if they see the value, usually we can start having a dedicated technical meeting where we can assess what is needed from us.
[DEMETRI]: But basically what the readiness level from the prospect client, which is not always the same, unfortunately.
[DEMETRI]: Yeah, well, I appreciate that journey.
[DEMETRI]: And I think at the end of the day, obviously I appreciate you're also sharing this conversation on the podcast this week, because I think so many organizations out there have been trying to figure it out.
[DEMETRI]: And obviously this is a good starting point to figure out what the next steps are.
[DEMETRI]: So very much looking forward to hearing your presentation at the live event.
[DEMETRI]: So we will look forward to that.
[MICHAEL]: Demetri just gave us a fascinating look at how being a first mover in AI can be both rewarding and complex. Next, he'll walk us through the infrastructure needed for sovereign AI deployment, highlighting why having control is crucial in today's market. This next part connects everything back to understanding the profound challenges in the AI landscape.
--- PRESENTATION ---
[DEMETRI]: Well, Michael, thank you so much for having me.
[DEMETRI]: It's a great honor.
[DEMETRI]: It's a great opportunity to present what we are building.
[DEMETRI]: We believe it's meaningful, and I'm going to just, you know, cut through the chase right now and walk you through on a very high level on what we have been building, how we continue building around trust AI execution infrastructure, as we call it, which is, we believe it's the next most important thing because you see the operational gap today.
[DEMETRI]: When it comes to every single AI deployment, it's that, you know, you have a bunch of very interesting and very intelligent models out there that can, you know, foundational models can generate phenomenal analysis and recommendations.
[DEMETRI]: And we believe they are commoditized right now.
[DEMETRI]: So by deploying AI agents, I mean, it's not as easy as it sounds.
[DEMETRI]: So if you want to deploy AI agents inside sovereign and operational environments, it does require a lot more than just the model capability.
[DEMETRI]: I think models can be plethora, and you see that almost on a monthly basis, they become smarter and, you know, a lot more intelligent, and that's why they are heavily commoditized.
[DEMETRI]: Now, that's not actually bad.
[DEMETRI]: That's actually very good news.
[DEMETRI]: But we all need to understand that it requires infrastructure that enforces the sovereignty, the provenance, and human accountability at every single step.
[DEMETRI]: And right now, I believe the top of the town has been the enterprise sovereignty and the AI sovereignty.
[DEMETRI]: So the barriers of the model capability anymore, we are talking about a new barrier, which is the sovereign control over any AI execution.
[DEMETRI]: So what we have envisioned and what we understand from the global market, and let me just stress out something which is very important, global markets, they are very different pacing right now.
[DEMETRI]: But I believe right now, which is the end of March 2026, we are all experiencing a pre-inflection market.
[DEMETRI]: So the agenting AI market is not here yet, in terms of adoption rate.
[DEMETRI]: Actually, I think we are experiencing the adoption death valley, which is, you know, good news and bad news at the same time.
[DEMETRI]: The good news is that, you know, the post-inflection market is about to come.
[DEMETRI]: We don't know exactly when, but we know it's coming for sure.
[DEMETRI]: And the bad news is that all the challenges that any venture has to go through now, we're taking this time to make best good use of this time and build meaningfully.
[DEMETRI]: So in order to achieve, what we say, sovereign AI deployment, organizations must ensure that they have sovereign control over execution, full evidence and provenance trails, we call it evidence-packed, end-to-end audibility, which is very important, human in the loop, which is the human accountability for every step of the way, and a safe failure under uncertainty.
[DEMETRI]: Now, this is, let's just say, the map of what any organization needs to have in place and ensure, in order to make sure that it achieves sovereign AI deployment.
[DEMETRI]: Now, AI systems must also safely interact with unstructured and siloed data.
[DEMETRI]: This is very important.
[DEMETRI]: We spend lots of time discussing with various clients and partners in diverse geographies.
[DEMETRI]: And one thing they have in common is that they have a chaotic, data, you know, there's no structure in terms of their data.
[DEMETRI]: So they have PDFs and CSVs and spreadsheets and ERPs and back-end systems and CRMs here and there.
[DEMETRI]: And also, most of the time, there's no clear owner.
[DEMETRI]: And many organizations don't even have a data store.
[DEMETRI]: And so we also need to come prepared and make sure that we can build an abstraction layer that we can safely interact with unstructured data.
[DEMETRI]: We also need to interact with fragmented operational systems, because this is the reality right now.
[DEMETRI]: We need to interact with air-gapped and sovereign environments.
[DEMETRI]: And last but not least, interact with very sensitive decision processes.
[DEMETRI]: So this is the reality now.
[DEMETRI]: And when I say reality, it's not only in small businesses, but it is the reality in small, medium enterprises, in some cases, even the very large enterprises.
[DEMETRI]: So we were working on our AI stack, as we say, which is why, you know, the Superbowl is a government orchestration infrastructure for AI agents operating across fragmented systems, definitely unstructured data and sovereign environments.
[DEMETRI]: So this AI stack is literally deployed inside sovereign boundaries.
[DEMETRI]: It can really be deployed on-premise or on any protected cloud, air-gapped when required, and provides the evidence back, as we say, or the provenance and the policy enforcement layers required for accountable AI execution.
[DEMETRI]: So we are, I think we are now, have really entered the era of how we're going from model output to governed action.
[DEMETRI]: I think this is very interesting for what is about to happen right next.
[DEMETRI]: And probably, or hopefully, this will also trigger the post-inflection market.
[DEMETRI]: When it comes to deployment architecture, you can see that the way we've been approaching that is that we built the bottom up and top down.
[DEMETRI]: So the resource abstraction, the knowledge foundation, the government orchestration, to the AI agents.
[DEMETRI]: We do have our own agenic AI framework that allows us to build AI agents without any limitation.
[DEMETRI]: Because many, many ventures out there, and it makes sense, they tend to use agenic AI frameworks that are provided by most of the hyperscalers.
[DEMETRI]: And, you know, don't take me wrong, they are very good, and actually they might have a very, very fast go-to-market.
[DEMETRI]: But when it comes to bespoke AI agents and, you know, achieve the maximum of customization, then, you know, there's no agenic AI framework offered by any hyperscalers that would actually offer that to you.
[DEMETRI]: Because these guys, they build horizontally and not vertically.
[DEMETRI]: So you need ventures like Superbo that we have our own agenic AI framework.
[DEMETRI]: Not only the framework, but all the other layers that we have ready or we keep building.
[DEMETRI]: So the air gap deployment, it operates without any external network dependencies.
[DEMETRI]: It's very important.
[DEMETRI]: It's one of the building blocks and tick of the box that you need to put in if you want to achieve your sovereign deployment.
[DEMETRI]: Number two, the sovereign data control.
[DEMETRI]: So anonymization before any data reaches the LLMs is very important for sensitive data, financial institutions, military, education, medical, it needs to be there.
[DEMETRI]: Govern model integration.
[DEMETRI]: So different model outputs govern and contain within the boundary.
[DEMETRI]: Right now, we don't use one model.
[DEMETRI]: So we're not dependent from one model right now.
[DEMETRI]: You know, we are agnostic and we can use different models under different circumstances and what we're trying to achieve.
[DEMETRI]: Next, policy enforcement.
[DEMETRI]: So governance rules enforced within the boundary and not externally.
[DEMETRI]: So it is very important to see, you know, when we talk about sovereignty, you know, the difference is not where the organization's data sits.
[DEMETRI]: The difference is that if the AI agent responds under your vendor's policies or your own policies, it's between owning or renting your AI agent.
[DEMETRI]: So you need to be the owner and this is what we believe.
[DEMETRI]: This is where we come in.
[DEMETRI]: Last but not least is the full provenance chain.
[DEMETRI]: We call it the evidence pack, right?
[DEMETRI]: Every action needs to be traceable to source, regardless if it, you know, it is successful or if it fails, you need to be able to provide evidence packs.
[DEMETRI]: So every single AI action within our stack in Superbowl produces an evidence pack containing the data source used, the reasoning trace, the applied policies, as I said, the organization's policies, not your vendor's policies, right?
[DEMETRI]: The received concepts, the action outcome, and the audit identifiers.
[DEMETRI]: Now, what we're trying to achieve is to be able to explain all the AI decisions because I think right now, what we see from the time that we spend with, you know, market shapers or small and medium enterprises or even the very large ones, they need to be able to have AI agents that can answer the why question.
[DEMETRI]: So one of the purposes is to explain the AI decisions, investigate the behavior, and demonstrate the compliance.
[DEMETRI]: And that's why the AI agent decision, it breaks down into two parts, two paths.
[DEMETRI]: The failure part, I'm going to go to the failure part.
[DEMETRI]: So it's confidence policy check.
[DEMETRI]: If it fails, then it means that it was low confidence, you have a policy conflict, there is a constraint in execution, and eventually it will come to human in the loop, which is we call it escalate to human.
[DEMETRI]: Or, the good news, it might be a pass, so very successful, because why?
[DEMETRI]: There was sufficient confidence and policy was satisfied, so the AI agent moved to execution.
[DEMETRI]: So both paths produce an evidence back.
[DEMETRI]: And this is very important because full provenance is preserved regardless of the outcome.
[DEMETRI]: In terms of how we have been operating today, we span in diverse geographies, in Superbowl, all the way from the US, to Africa, to Europe.
[DEMETRI]: And, you know, multilingual interactions, different languages, fragmented infrastructure, a plethora of regulatory variability depends on the, not only the country, but also the sector.
[DEMETRI]: We are in sectors like, you know, telco, or media, or medical.
[DEMETRI]: So each and every sector comes with its own peculiarities and regulatory policy enforcement.
[DEMETRI]: So then you also need to make sure that you adapt to different compliance requirements, not only per organization, but also per jurisdiction.
[DEMETRI]: You know, the complex also operational environments that we're in require full auditability and human accountability in different complex operational contexts.
[DEMETRI]: So this is more or less of what we have been building and what we keep building, because this is a never-ending story.
[DEMETRI]: This is an R&D ecosystem by nature.
[DEMETRI]: So I think the natural habitat of AI is R&D.
[DEMETRI]: So from model output to government action, you know, government AI execution today, you know, it's around infrastructure for sovereign, very well air-gapped, and operationally complex environments.
[DEMETRI]: And the objective is to make sure that you achieve sovereignty, provenance, and accountability.
[DEMETRI]: And all these should not be negotiable at any point.
[DEMETRI]: Otherwise, you're not actually having achieved any AI or enterprise sovereignty.
[DEMETRI]: So I think I'm in time.
[DEMETRI]: That's about a very high level about Superbo, Mike.
[DEMETRI]: Fantastic.
[DEMETRI]: I think that gives the audience a great view of the area.
[DEMETRI]: And coming up right at the top of the hour, I know you'll be on the Q&A panel.
[DEMETRI]: So we'll have the opportunity to take questions.
[DEMETRI]: So for everybody that's listening live, just type in your questions into the Q&A box that you see right in front of you.
[DEMETRI]: And we will get to as many of those questions as we can during the live Q&A session.
[MICHAEL]: Demetri's presentation highlighted the importance of sovereign AI infrastructure as a key to future-proofing businesses. Recorded live at our Software Oasis AI Bootcamp, this panel features audience questions that dig deeper into these insights. These real-world queries will illuminate how businesses can strategically approach AI deployment.
[MICHAEL — CHECKPOINT]: So, what Demetri made clear is that achieving sovereign AI deployment requires not just advanced models, but a robust infrastructure for accountability and control. This leaves us wondering how exactly businesses can ensure they're prepared for this shift and what pitfalls to avoid. And that's exactly what this audience brought to the table during the panel session.
--- Q&A PANEL ---
[DEMETRI]: I hope everyone is enjoying the event so far.
[DEMETRI]: For everyone that's been with us since first thing this morning at 8 Eastern time, we've come to our first Q&A panel, and we'll have the opportunity to ask a lot of the questions that have been submitted to each of the panelists.
[DEMETRI]: After this panel, we'll have a presentation from Lucas over at LL Informatics.
[MICHAEL]: So why don't we get things started?
[DEMETRI]: I'll introduce everyone on the panel.
[DEMETRI]: Joining us from NTT Data, we have Niraj Singhal.
[DEMETRI]: He's the group senior VP.
[DEMETRI]: Welcome, Niraj.
[DEMETRI]: Next we have joining us from Superbow AI, we have Dimitri Papazisis.
[DEMETRI]: He's the co-founder and CEO.
[DEMETRI]: And finally, joining us from Abbey, we have Maxime Vermeer, who is the VP of AI Strategy.
[DEMETRI]: Welcome, Max.
[DEMETRI]: Hey, everyone.
[DEMETRI]: So let's get started.
[DEMETRI]: I want to try to get to as many of these questions as we can.
[DEMETRI]: We had just a ton of questions come in from all around the world.
[DEMETRI]: So the first question that came in was for Niraj, and, Niraj, this question comes to us from right in this area where we are, just outside of Boston.
[DEMETRI]: This is from Ethan, who's, as I said, just outside of Boston in Massachusetts.
[MICHAEL]: And Ethan is asking, when you advise executives to move from discrete AI pilots to business model reinvention, what's the very first metric you ask them to change so everyone feels a shift towards outcomes and not experiments?
[MICHAEL]: Yeah, that's a very, very interesting question, Ethan, and thank you for that one.
[DEMETRI]: I think a lot of people confuse AI as any other technology.
[DEMETRI]: My feeling is AI is more a mindset than technology.
[DEMETRI]: So I do not believe that you can get the full power of AI unless you change perceptions, you change mindsets, you train people up, and you make them more outcome-oriented.
[DEMETRI]: So I think the starting point is really sensitization of the teams that are involved, the extended leadership within the client organization in terms of what needs to change from behaviors, attitudes, perceptions, before they really have the full impact of AI felt within the organization.
[DEMETRI]: So you start with people, and then ultimately you get into process and technology as you scale up and move towards business process reinvention.
[DEMETRI]: That's really the way we look at it.
[DEMETRI]: Makes sense.
[DEMETRI]: The next question came in for you from Harper, who is in, looks like just outside of Denver, Colorado.
[MICHAEL]: And Harper is asking, when a team shows you a successful LLM invoice pilot at 95% accuracy, what specific questions do you ask to stress test whether it will survive real production document chaos?
[MICHAEL]: Well, the very first question is, how many times did you run through the same invoice and did you get the same result every single time?
[DEMETRI]: I think the probabilistic nature of this technology really shines in the fact that it can create incredibly plausible results.
[DEMETRI]: And it's not that you should not use LLMs in any type of production pipeline when it comes to documents.
[DEMETRI]: It's actually the inverse, where you need to combine it with other types of technologies that truly ground it inside of the data that is truly inside of documents.
[DEMETRI]: You can have all of those references and you can trust the output that you're getting.
[DEMETRI]: Where we've seen much more success in leveraging that kind of capability is when you have that ground through data and you are then able to leverage a large language model, for example, or vision language model for a specific part of the task that you're trying to do in your full document pipeline.
[DEMETRI]: Fantastic.
[DEMETRI]: And Dimitri, this next question came in for you.
[DEMETRI]: This is still here in the U.S.
[DEMETRI]: We have from Avery, who is in, looks like just outside of Miami, Florida.
[DEMETRI]: So all the way in the southernmost point in the U.S.
[DEMETRI]: Not quite the most southernmost.
[MICHAEL]: Avery is asking, when a CIO tells you their agents are working fine behind an API, what red flags do you look for to judge whether they actually have the sovereign governed infrastructure you described?
[MICHAEL]: All right.
[MICHAEL]: So I think everything starts from a different standpoint, right?
[DEMETRI]: So first of all, the way we believe and approach is, first of all, we need to make sure that the enterprise has the proper data structure.
[DEMETRI]: Most enterprises, they are, you know, magic results and making sure that their data is very well structured.
[MICHAEL]: Most of the time, data is chaotic, scatters here and there, back and see your PRMs or different media, you name it, right?
[MICHAEL]: So for us, before you reach the AI agents, you need to make sure that your vendor has isolated systems and data.
[DEMETRI]: He has an abstraction layer to make sure that we vacuum the unstructured data and we structure the data in a way that at least our system will understand and make sure we deploy a system of agents, not one agent, a system of agents, agentic AI.
[DEMETRI]: Then we need a knowledge foundation layer, which is basically all the knowledge, sovereign data control and the governed orchestration layer.
[DEMETRI]: If you don't have those in place, in our opinion, your AI agents are not really working.
[DEMETRI]: And also we need to make sure that your AI agents co-work together.
[DEMETRI]: So it's different when a task-oriented standalone tool and still a different thing if you have an agentic AI, which is a system of agents, they communicate, they work together, they reason together, and they're goal-oriented.
[DEMETRI]: So these are very important.
[DEMETRI]: Fantastic.
[DEMETRI]: The next question came in from Neeraj, but before that, I will stand corrected.
[DEMETRI]: I see a lot of comments came in.
[DEMETRI]: Yes, Key West is the southernmost point in the US.
[DEMETRI]: So for a lot of our fans in that region.
[DEMETRI]: True, very true.
[DEMETRI]: Correct.
[DEMETRI]: Neeraj, the next question comes into us from Priya, who's in London, over in the UK.
[MICHAEL]: And Priya is asking, how do you keep business stakeholders focused on reimagining customer outcomes when the loudest internal conversations are still about model choice and vendor selection?
[MICHAEL]: So that's, again, a very, very interesting question.
[DEMETRI]: I think a lot of times you got to take a step back and really reflect on the reason why you're there.
[DEMETRI]: The reason why every business is there is to make it easier for our customers to get the outcomes that they desire.
[DEMETRI]: That's the sole purpose of any business.
[DEMETRI]: So it's very important to start reimagining your business with the customer in the center.
[DEMETRI]: That's the starting point.
[DEMETRI]: And I firmly believe that customer dissonance and customer fiction become the starting point of reinvention.
[DEMETRI]: So what AI and agentics and everything new that's coming in from a technology perspective really helps you do is reimagine the business with the customer at the center, which essentially means the way you were dealing with the customer and the way you had outlined your processes may not be relevant today, and there might be an easier way of making the customer do business with you.
[DEMETRI]: So I think it's true self-reflection of the reason why you're there, keeping the customer at the center, and then thinking about the business and reimagining the art of the possible is the approach that one needs to take.
[DEMETRI]: The models will flow.
[DEMETRI]: The models will be commodity.
[DEMETRI]: You know, I have this very interesting thing in my mind.
[DEMETRI]: If you look at Java as a technology that came in about 25, 20, 30 years back, everyone was talking about Java.
[DEMETRI]: But today, Java is pretty much involved in everything, but nobody's talking about Java.
[DEMETRI]: It's about the outcomes that you can achieve with Java as the underlying basis for that.
[DEMETRI]: So I think AI in the next five years is going to become commodity.
[DEMETRI]: What's going to be really differentiating is the outcome.
[DEMETRI]: AI is going to be native in everything that we do for our clients, everything that we do in our businesses.
[DEMETRI]: And I would agree with you that the models, even at this point, are getting so good that it's in many cases hard to differentiate.
[DEMETRI]: So this next question comes to us from Max, who's in, actually from Jackson, for Max, from Raleigh, North Carolina.
[DEMETRI]: And shout out to any NC State listeners in the Raleigh, North Carolina area.
[MICHAEL]: The question is, how do you decide which fields in a complex document should be handled by deterministic IDP versus left-to-gen AI so you get both reliability and flexibility?
[MICHAEL]: No, that's a great question.
[MICHAEL]: And that truly speaks to thinking about what the right approach is for the task at hand, right?
[DEMETRI]: So in essence, my recommendation is always that you go for the majority of all the data points that you want to extract from a document or basically any piece of content through deterministic capabilities, deterministic AI, machine learning models.
[DEMETRI]: And it's only when you actually can basically limit the amount of context that needs to be actually analyzed or processed by a large language model that you can actually then also minimize the amount of risk that you might have for hallucinations based on the original document that you're processing.
[DEMETRI]: In a great example, we've seen this a couple of times where one of our customers had to process lease agreements, massive documents, thousands of pages sometimes.
[DEMETRI]: And a specific use case where they did use a large language model was after we already had extracted the specific segment that contained the answer, but then it required interpretation.
[DEMETRI]: It required the deeper level of transformation that you can get with deterministic AI.
[DEMETRI]: And that's, for example, a perfect use case of where you want to use the tandem of the two and all other, actually, this was 350 data points were all extracted using deterministic capabilities because it gave them that assurance that if there was something wrong, we could actually be able to tell them, whereas the probabilistic will always give you an answer.
[DEMETRI]: It's just much harder to detect whether it's the right one.
[DEMETRI]: And so in this particular instance, that was also linked to compliance and governance that they had to provide to the regulators.
[DEMETRI]: So in this particular instance, the mobility of a wrong answer had a much higher impact in terms of down the line for the business than having to go through and do something manually.
[DEMETRI]: So that was actually their preference.
[DEMETRI]: So as always, there's more than just a technical aspect that makes the decision of what you're going to use and when.
[DEMETRI]: Dimitri, this next question comes to us from Oliver, who is in Paris, France.
[MICHAEL]: And Oliver is asking, when agents need to work across PDFs, spreadsheets, ticketing tools, and air gap systems, how do you design the orchestration so data never silently leaks outside sovereign boundaries?
[MICHAEL]: All right.
[DEMETRI]: So we have our own AI stack entirely within the sovereign boundary.
[DEMETRI]: We begin with a full governance chain, which is for every action, there's a traceable to source.
[DEMETRI]: These are our systems and data.
[DEMETRI]: Then we go to policy enforcement, governance rule, enforce it within the boundary and not the.
[DEMETRI]: Then we have the governed model integration.
[DEMETRI]: Actually, it's model outputs governed.
[DEMETRI]: Actually, it's model outputs governed.
[DEMETRI]: And we have for any data, it's covering data control.
[DEMETRI]: So we don't have any sensitive data at all on our language model.
[DEMETRI]: So we make sure that we are not in mind.
[DEMETRI]: And then when we are talking about deployment, we're talking about air gap deployment, which actually operates without external network dependencies.
[DEMETRI]: And these were our agents, AI agents really work.
[DEMETRI]: So the way we approach it, it's probably, it's not a fast way to go to market, but it's the only way to go to market.
[DEMETRI]: What I'm trying to say, it's a secure way to go to market.
[DEMETRI]: That is why we decided this is where we've been around since.
[DEMETRI]: So before AI evolution, we began from doing NLP.
[DEMETRI]: So we always took under very serious consideration to everything.
[DEMETRI]: So it's a replay here were actually layers.
[DEMETRI]: And that happened within layer.
[DEMETRI]: From isolated layer to the resource layer, to knowledge foundation layer, to government orchestration layer, and eventually to AI agents, which are going to be air-gapped, it's going to be the air-gapped deployment.
[DEMETRI]: Makes a lot of sense.
[DEMETRI]: So Niraj, this next question comes to us from Grace, who's in Austin, Texas.
[DEMETRI]: And Grace is wondering, and I gather that she's in the banking industry.
[MICHAEL]: For a bank that has already automated pieces of the mortgage workflow, how do you decide which end-to-end journeys are mature enough for agentic AI versus those still too broken in their underlying processes?
[MICHAEL]: Yeah, excellent question again.
[DEMETRI]: So this is the part that I think we should be moving away from.
[DEMETRI]: So if you really look at the initial adoption of AI and predominantly agentic AI, most of the adoption was for cost and efficiency.
[DEMETRI]: So if you look at an end-to-end process and you say, I'm going to streamline these processes, I'm going to deploy discrete agents across different aspects of this value chain.
[DEMETRI]: And I'm going to achieve a compression of about 30% to 40%, I'm going to reduce cost, and I'm going to improve efficiency.
[MICHAEL]: But I think the fundamental question that one needs to ask is, what's the outcome that the customer is looking for?
[MICHAEL]: Well, the customer is trying to buy the best home for himself or herself.
[MICHAEL]: So the outcome that the customer is looking for is, can you help me buy the best home possible?
[MICHAEL]: And if financial institutions start reimagining their businesses to make customers get the best possible home, rather than provide the customer the best possible mortgage, you start looking at the art of the possible.
[DEMETRI]: So I would say, look at what the customer's trying to do and reimagine your business end-to-end in terms of how you can cater to that and how you can help the customer, rather than cherry-picking pieces of the value chain where you can drive discrete efficiency.
[DEMETRI]: So it's a completely different paradigm.
[DEMETRI]: I don't think many financial institutions are there yet, but that's really the way to get the full power of agentic AI.
[DEMETRI]: It's not to look at discrete compression, but to look at the end-to-end lifecycle with the customer at the center and becoming more outcome-oriented in terms of what the customer is trying to do.
[DEMETRI]: The next question comes to us from Max, and it's from Chloe, who is just outside of Phoenix in Arizona.
[MICHAEL]: And Chloe's question is, for leaders under pressure to do agents now, how do you persuade them to invest first in document quality and structure instead of skipping straight to agent orchestration?
[MICHAEL]: Because I think the reason, or at least the persuasion, has already been kind of called out by the other participants in the Raj and the Mitra as well, right?
[MICHAEL]: You need to have that level of data and perception available for your agents to actually be successful.
[DEMETRI]: I often compare it to when they first started working on the self-driving car.
[MICHAEL]: Did you actually figure out first how to make it turn left or right?
[MICHAEL]: No, you actually needed to give it that perception layer to understand where the hell it's going on the road, what lanes are, if there's pedestrians.
[MICHAEL]: So if it doesn't understand the business world that it's operating in, how on earth can you actually have it do anything that's going to actually further your business, right?
[DEMETRI]: So I think that's the important thing to remember.
[DEMETRI]: And I think if you just give them that metaphor to your business leaders and you're trying to convince them, like, okay, we need to do the groundwork first, otherwise we're going to have a car crash.
[DEMETRI]: It's as simple as that.
[DEMETRI]: Makes a lot of sense.
[DEMETRI]: The next question came in for Dimitri, and this is from Jaina, who looks like in Warsaw, Poland.
[MICHAEL]: And Jaina's asking, as model providers ship new capabilities every few weeks, how do you maintain model agnostic governance so that policy and accountability don't have to be rebuilt with every upgrade?
[MICHAEL]: Well, first of all, yep, that's true.
[MICHAEL]: Models are getting smarter and this is this is exactly what makes them a commodity.
[MICHAEL]: All right.
[DEMETRI]: So and this is that's why we just use them today.
[DEMETRI]: We, as you know, in Superbo, we are LLM agnostic.
[DEMETRI]: So we use more than one LLM, depending on what we are trying to succeed.
[DEMETRI]: For example, a different LLM for reasoning or different LLM for analytics.
[DEMETRI]: Depends.
[DEMETRI]: Each and every model has its own unique superpower.
[DEMETRI]: Right.
[DEMETRI]: In terms of strengths and weaknesses.
[DEMETRI]: Now, the security does not, you know, the policies don't happen inside the model.
[DEMETRI]: They happen outside the model in your infrastructure.
[DEMETRI]: That's why you need to have your own AI stack.
[DEMETRI]: Otherwise, this is, you know, otherwise you're just an LLM robber and it's going to go south.
[DEMETRI]: So you need to make sure that you have your own AI stack, which is going to be a full layer above the model.
[DEMETRI]: What we do in Superbo, besides being LLM agnostic, we're also infrastructure agnostic.
[DEMETRI]: It means that we can deploy in any kind of cloud, private cloud or on premise without a cloud.
[DEMETRI]: So we don't depend on one.
[DEMETRI]: It's very important.
[DEMETRI]: We cannot go out there and claim sovereignty if we depend on one model or one cloud infrastructure.
[DEMETRI]: Not only that, but I'm going to take it a step further.
[DEMETRI]: Even the framework that we build our agents is something that we've built from scratch.
[DEMETRI]: So we don't use hyperscalers to build our own AI agents.
[MICHAEL]: What I'm trying to say is that, for example, it would be very easy and very fast, actually, to build AI agents on Google Vertex, right?
[MICHAEL]: Go there and it would be very fast.
[MICHAEL]: Go to market.
[DEMETRI]: The reason that we chose the hard way was we know very well the hyperscalers are built horizontally.
[DEMETRI]: It means that we're going to hit a wall at some point when it comes to bespoke tailor making and customization.
[DEMETRI]: That's why we build our own agendic framework so we can build our own AI agents in there.
[DEMETRI]: We had to make the whole framework to be infrastructure agnostic and LLM agnostic.
[DEMETRI]: So anything around security, the governance, the orchestration, everything happens after the model.
[DEMETRI]: So it has nothing to do with the model.
[DEMETRI]: And this is what air gap deployment more or less means.
[DEMETRI]: Niraj, the next question came in for you.
[DEMETRI]: And this question came in from a few different people in slightly different ways.
[DEMETRI]: So I'm going to pick one.
[DEMETRI]: I'm going to pick one from Daniel, who is just outside of Chicago, Illinois.
[MICHAEL]: And Daniel's asking, when tech debt is high and budgets are tight, how do you prioritize which parts of the legacy stack must be modernized now to unlock AI value versus what can safely wait?
[MICHAEL]: You know, this actually reminds me of a quote from Edward Deming.
[MICHAEL]: And Deming, who was known for probably the guru in quality, used to say, you don't have to do this.
[MICHAEL]: Survival is not mandatory.
[MICHAEL]: You don't have to do this.
[MICHAEL]: Survival is not mandatory.
[MICHAEL]: If you're living with tech debt, with all the influx of technology and the changes that are happening, it's going to be really, really, really difficult to streamline your business.
[DEMETRI]: So I think the starting point obviously is tech debt.
[DEMETRI]: And I would say even from a tech debt perspective, look at client facing, customer facing systems first.
[DEMETRI]: Those are the ones that are going to give you the maximum return.
[DEMETRI]: Those are the ones that are going to have the maximum impact and then work your way to the back end.
[DEMETRI]: So I would say address tech debt holistically if possible.
[DEMETRI]: If you can't and you've got only 10 percent budget available, which usually is the case, look at making the maximum impact for the customer.
[DEMETRI]: And that's the way to prioritize processes.
[DEMETRI]: Great answer there.
[DEMETRI]: And Max, I'm going to send this last question to you as we're coming to the end of our time.
[MICHAEL]: Sarah, who's in Stockholm, Sweden, asks, when you embed human in the loop review into document workflows, how do you keep it from becoming a permanent bottle deck instead of shrinking safety net over time?
[MICHAEL]: Awesome.
[DEMETRI]: Well, I love having questions from fellow Europeans.
[DEMETRI]: And it's actually a really good one.
[DEMETRI]: So human in the loop is one of the key components that you want to have inside of your document workflow, even inside of your genetic workflow, any workflow whatsoever.
[DEMETRI]: Because I think that human component, depending on what type of organization you are, what type of industry that you're working in, the regulations, compliance, you're going to need it.
[DEMETRI]: Now, whether it becomes a bottleneck or not, that's something else that you can help determine.
[DEMETRI]: But actually that can also be determined by the regulations and compliance that you have to deal with.
[MICHAEL]: Now, specifically to the more question of, OK, how do you avoid that?
[MICHAEL]: It just becomes more and more and creates a choking point for your actual workflow.
[DEMETRI]: I think that's not particularly a problem that we see often happening inside of our customers when they actually deploy our technology for two reasons.
[DEMETRI]: One, the capability to actually take that human input and further train the system on how it should have dealt better with that information that was provided.
[DEMETRI]: Is the thing that truly sets it apart in terms of it being an educator for agents, for the models that are being used.
[DEMETRI]: And the other component is absolutely that the fact that you are able to leverage it now in combination with, for example, like Dimitri was talking about, the agentic capabilities, allow you to actually also accelerate that.
[DEMETRI]: So we're no longer just having human in this loop, but there actually is assistance from agents happening too.
[DEMETRI]: So they're making the humans actually be able to work through it all faster in terms of, you know, not creating that choke point, but actually assisting them, making the work more faster.
[DEMETRI]: And that's what we see all the time, that there's a high request.
[DEMETRI]: So to my earlier point, it does really depend on regulations, industry, the specific use case.
[DEMETRI]: But the main thing is, is that you're always going to want to have it.
[DEMETRI]: And I've never seen it become a true choke point because the systems that you leverage it in should be able to learn from the additional input.
[DEMETRI]: And now there's even more and more techniques to actually further assist the human in the loop to move faster and faster.
[DEMETRI]: I think a lot of great insights there.
[DEMETRI]: I'd like to thank Neeraj, Dimitri and Max for their fantastic presentations earlier in the event and for joining us for these audience questions on this panel.
[DEMETRI]: Coming up in just a moment, I want to keep everybody on target for the next session.
[DEMETRI]: But if anyone in the audience has additional questions or we didn't get to your question, at the top of the site, you'll see that big orange button.
[DEMETRI]: You can submit your, you know, your question there and someone on the team or one of the panelists will respond to you after the live event.
[DEMETRI]: So with that, I will send everybody off to the next session.
[DEMETRI]: Thank you.
[DEMETRI]: Thank you, Michael.
[MICHAEL]: Reflecting on today's episode, it's clear that building the right infrastructure is key to harnessing AI's potential, and that means focusing on sovereignty and accountability. We learned how Superbo AI navigates these challenges, proving that early adoption comes with its own set of hurdles. Demetri, thank you for sharing your journey with us. For those who want to learn more, I encourage you to connect with Demetri on LinkedIn. And if you haven't already, come join us over at softwareoasis.com — that's where all of this really comes together. Hey, if you enjoyed the episode, please consider subscribing or leaving a review. It really helps. Until next time, take care!