Agis Stathakis is Co-Founder and CFO of Superbo, an enterprise AI company that helps organizations deploy secure, governed AI agents across their existing systems. He brings more than 20 years of experience in corporate finance, investment banking, restructuring and asset management, including managing a large EMEA portfolio at Piraeus Bank Group and serving as CFO of a multi-industry group. We spoke with him about his journey from finance to AI entrepreneurship and the future of trusted enterprise AI. Here’s what he told us.

Tell us a bit about Superbo — how did the idea come about, and what made you realize there was a market need for your solutions?

Superbo started in 2020 from a very simple idea: companies were investing heavily in digital transformation, but the customer experience was still often slow, fragmented and frustrating. We felt there had to be a better way for people to interact with large organizations, especially in sectors like telecom and banking. We started with conversational AI and natural language understanding, and as the technology evolved, so did we. We moved into generative AI and today into agentic AI.

The real proof for us came when we started seeing the technology work at scale in large, demanding organizations. That is when you realize there is a big difference between building an impressive demo and building technology that has to work every day, connect with real systems and actually solve a business problem. Today, Superbo is really a lot more than just a conversational AI firm, it is more about helping enterprises use AI in a practical way, making interactions, decisions and processes simpler, faster and more intelligent.

What sets Superbo apart from similar solutions in the market? What specific value do you deliver to your customers and partners?

We solve the trust problem that stops enterprises from moving AI into production. Most AI failures aren't about model capability, they're about confidence. Organizations can't verify what data informed a decision, trace how an agent reached its conclusion, or prove compliance. We close that gap.

Our stack is built trust-first: governance, traceability, and audit-ready evidence are enforced at runtime, not bolted on afterward. Unlike agent frameworks that leave you to solve data semantics, compliance, and audit evidence yourself, we deliver the full stack, from data semantics to orchestration to domain agents. And unlike pure governance tools that observe systems after the fact, we govern during execution and produce Evidence Packs as a native output of every action.

The result: you move from AI experimentation to real production with confidence, not just a toolkit.

What has been your biggest challenge so far — and how did you overcome it (or how are you addressing it)?

One of the biggest challenges has been the speed at which everything changes. In AI, something can feel groundbreaking today and almost standard a year later. At the same time, the large companies we work with need stability, reliability and trust. So for us, the challenge has always been to move fast without chasing every new trend. Over time, we learned to focus less on what is fashionable and more on what actually creates value for the customer.

Superbo has changed a lot since 2020, but the core problem we are trying to solve has stayed surprisingly consistent: how can we use technology to make complex organizations easier to deal with? I think one of the most important things we have learned is how to evolve constantly without losing sight of why we started in the first place.

What are the most important lessons you’ve learned along the way?

The biggest lesson for me is that entrepreneurship is rarely about having one great idea. It is much more about adapting that idea again and again as reality changes. Markets change. Technology changes. Customers change. And sometimes the things you were absolutely convinced about at the beginning turn out to be wrong. 

I have also learned that technology by itself is rarely enough. Especially today, when powerful AI tools are becoming available to almost everyone. The real advantage increasingly comes down to judgment: knowing what to build, what not to build, which problems are worth solving and what people will actually value.

And one more lesson that probably comes later in the journey: as a founder, your goal should not be to make the company depend on you forever. In the beginning, you are involved in everything. But eventually, you have to build the right people, culture and structure so the company can grow beyond you. For me, that shift from building a product to building an organization has been one of the biggest parts of the journey.

What are your impressions so far on the EU AI Act. What are the fundamental actions that enterprises should consider to be compliant with the law and safe with the use of AI?

The EU AI Act is broadly positive. The risk-based approach makes sense, and in many ways the regulation is formalizing what responsible enterprise AI should already look like: accountability, traceability, human oversight, data governance and continuous monitoring. For enterprises, I would focus on a few fundamentals.

First, know where AI is being used and classify the risk of each use case. Second, establish clear ownership and governance: who approves the system, who is accountable for it and what it is allowed to do. Third, build for traceability, so you can understand what data an AI system used, why it produced a result and what actions it took. Human oversight is equally important. “Human in the loop” should not just be a phrase; organizations need clear rules about when AI can act autonomously, when approval is required and when a case must be escalated. Finally, enterprises need to continuously monitor AI systems and invest in AI literacy, so employees understand both the capabilities and the limitations of the technology.

My advice would be not to treat the AI Act simply as a compliance checklist. The better question is: how do we make AI accountable inside the organization? If you know what your AI can access, why it reached a conclusion, what it is allowed to do and who remains responsible, you are addressing both the regulatory challenge and the much bigger issue: trust.

Governance is not what slows enterprise AI down. Done properly, it is what allows it to scale.