Michael Economou on building Exyde around AI teams

We started Exyde as a marketplace for activities and experiences. It taught us how operators work, but it was a crowded market where building a real edge would be hard. We moved into building AI tools for those operators. Our first product, Lucy, was an assistant for their daily work, but still far from an AI employee. The turning point came from inside our company. We were bootstrapped and close to hiring an operations specialist when I realised the technology was mature enough to build that role with AI instead. We built Kira, always on and proactive, and I found myself chatting with her from my phone as I would with any teammate. It worked, so we built Alex for our catalogue, Aria for finance and Max for business development. When we told friends and partners how we run the company, they asked us to build the same for them. They wanted the same method.

That is when I realised the product was the way we work. Roles and human oversight ChatGPT and Claude are fantastic tools, but they depend on the person using them. In a team, everyone uses them in their own way, and quickly nobody is managing the result. AI employees work the other way around. Each has a defined role, a place in the team and someone managing it, whether a human or an AI chief of staff. Our core philosophy is to treat AI as we would people: we give them names, onboard and train them, provide feedback and correct mistakes. Human supervision has to stay wherever someone is accountable for the outcome. How much depends on the risk and what the technology can do today. But there is one difference I keep coming back to. A person avoids mistakes because a mistake can cost them their job, which is a kind of survival instinct. AI doesn’t have that. Building differently In the past, every new service and…