Almost ten years, multiple products, one then two development teams, and now a brand-new AI team. Markus Happe has seen Exeon from nearly the very beginning. We sat down with our CTO to talk about what it means to build cybersecurity products in the middle of an AI race, how you grow a team when the tools change every few months, and why being careful and being innovative don’t have to be opposites.
You’ve been building security products at Exeon since long before ‘AI-powered’ appeared on every pitch deck. If you had to explain to a skeptical CISO what’s real and what’s marketing, where would you draw the line?
I’d tell a skeptical CISO that AI can deliver real productivity gains, but the value depends on the use case and how the system is governed. AI isn’t inherently unsafe. The risk comes when an agent can access sensitive data, encounter untrusted content, and send information or take actions through tools such as browsers, email, APIs, or MCP servers. That combination is often called the “lethal trifecta.” It’s a useful way to assess exposure, not a reason to avoid AI altogether.
Marketing will tell you to plug in any AI tool, and everything magically gets better, risk-free. It doesn’t work that way. CISOs still need to evaluate the technology, set appropriate permissions and safeguards, and ensure activity can be monitored and incidents addressed. Those responsibilities can’t be offloaded to an agent or a product. However, we want to make the task as easy as possible. At Exeon, we’ve spent years monitoring what happens on networks, and we’ve now built Exeon.AgentGuard to monitor what AI agents do.
Swiss companies have a reputation for being careful and precise. Is that an advantage or a handicap in the AI race?
In cyber security, being careful and precise is a strength, with or without AI. But right now, AI is moving so fast that most companies feel enormous pressure not to miss out. The common belief is: if you don’t switch to AI now, you will be irrelevant tomorrow.
The problem is that rushing change comes with risks. AI is a very powerful tool, but it needs strong guardrails, oversight and human experts who control what agents can do. Without that, companies open themselves up to entirely new kinds of cyberattacks.
So yes, it may sound counterintuitive, but I believe that being careful and precise with AI is a clear advantage. And I am curious to see how the AI race plays out to the big tech players and whether their investments will pay off as much as they hope.
What’s one thing that you’ve deliberately decided not to let AI do, and why?
We don’t let AI approve code reviews or merge code into production. AI supports our work, but engineers and I remain accountable for every change.
Same goes for testing. If AI writes tests without supervision, it often produces tests that look fine but don’t actually check whether the feature works. So we stay in charge. AI is our assistant, not the other way around.
How has AI changed the way your engineers work day to day, and has anything gotten harder because of it?
AI has certainly improved our productivity. We can build more without growing the team at the same pace. But AI models need to follow your coding and architectural guidelines, so we used that as an opportunity to sharpen ours. That’s been a real benefit itself.
What’s gotten harder? Today, we review more code than ever before. As development speeds up, it becomes more challenging to stay in control and keep questioning what AI generates. The way we build software is drastically changing and keeping on top of that takes real effort.
You’re building a new AI team right now. If you were hiring an engineer today, would you look for different things than five years ago?
In some ways, yes. Five years ago, AI wasn’t even part of our job descriptions. Today, experience with AI-assisted development is something we actively look for, because it’s become part of how every engineer works.
That’s especially true now that we’re building a dedicated AI team. They’ll be building AI capabilities directly into our products, starting with tools like the ExeonAssistant. For security vendors like Exeon, the bar for accuracy and trust is much higher than in most industries, so beyond the technical skills, we’re looking for people who are curious about AI but also critical of it.
Bonus questions
What would your team say is your most recognizable quirk as a leader?
– Probably that I always say “we”. But I really believe that we win together, we fail together and we grow together.
If Exeon were a Swiss dish, what would it be and why?
– Raclette. Originally Swiss, but enjoyed across Europe and far beyond. Everyone builds their own plate the way they like it, and it’s best enjoyed together, just like working with Exeon and our partners.
Finish the sentence: “AI will never replace…”
– …the people who inspire us. The ones who find words for whatever we cannot express and who shape our culture. AI will never replace responsibility for what we do, our values, or the great feeling of achieving something together with other humans. And it will never replace our humor and love.
