Between code and consciousness: How Marco Moosbrugger makes AI human

Sometimes the best ideas come from where you least look for them, at a poker table, for example. Between chips, laughter and three hours of conversations about life. Neither about work nor about money.

At the poker table among friends, Marco Moosbrugger met the man who would later become his business partner. Today, together with Andreas Hohenstatt and a second colleague from the poker circle, Armin Gaiser, he is building one of the most exciting AI companies in Vorarlberg. A company that actively applies artificial intelligence without having to talk about it too much. Pragmatic and down to earth, above all by people for people.

From a social media project to an AI vision

The story begins in a world that many people know: social media. I know Moosbrugger from his time as a co-founder of Castl, an agency that built a social media platform, managed content, promised reach. But the algorithm did not mean well. “We simply could not get a foot in the door,” he says. So he pulled the ripcord and started over.

Together with a small team from Ukraine, they developed their own language model in 2022, long before AI exploded in the feeds. He pitched to large companies, learned, failed, kept researching. And because the acquisition of new knowledge also made his talent for passing on knowledge known, he founded his own company in the field of AI training for SMEs.

In parallel, Byti emerged, an AI agent system that lets digital tools talk to one another instead of making people jump back and forth between programs.

“I never wanted to sell AI, I wanted to explain it.”

Perhaps that is the difference between a trainer and a technician.

Byti: The quiet revolution in the back office

What Byti does sounds simple: it connects a company’s existing software solutions, accounting, CRM, calendar, email, in a single interface. One voice command is enough, the rest runs in the background. The data stays on servers in Feldkirch, that is, in Austria.

Moosbrugger calls this “agent intelligence,” an architecture based on the idea that AI does not appear as one large system, but as a network of many small, specialized agents. Each person can create their own, each can shape their digital team themselves.

“We wanted to build a system that does not create dependency.”

The data stays in Austria. Control stays with the user. It is an approach that says less about the individual product and more about the question of what European software architecture can even look like.

Teaching instead of selling

Moosbrugger is reluctant to call himself an expert. “I am a trainer, not a prophet,” he says, and laughs. Perhaps that is exactly why people listen to him. He does not sell fear of the future, but competence. With the domain KI-Führerschein he has secured a brand that fills his workshops and wakes people up from their digital sleep.

With the help of his colleague Marco Esposito, he creates places where people grasp for the first time what artificial intelligence really means: neither magic nor science fiction, but logic and craft.

Over 70 percent of participants, he says, have hardly any prior knowledge. And of those, only 30 percent really wanted to go deeper. “That shows where we stand. The train is in the station. You really should just get on.”

From being overwhelmed to finding direction

Moosbrugger knows the skepticism. “Many say: AI does nothing for me. I do not need it.” But as soon as they hear that they will no longer have to write quotes themselves, no longer have to type up accounting, “they prick up their ears.” By now he works with hotels, tax advisors, cities, manufacturing companies and insurers, everywhere processes repeat themselves.

“Back office, support, marketing, that is where the greatest leverage lies.”

But he goes further: “I do not want to sell everyone something. I want everyone to understand what they need AI for.” That has become rare, an entrepreneur who does not put his product at the center, but the person in front of it.

Data sovereignty as a stance

In a world where data is the new currency, Moosbrugger’s vision is almost romantic: language models, trained in Feldkirch. No server farms in California. No passing on to third parties. A closed, secure data cycle, as a symbol of European independence.

He wants small, high-performance models, specialized rather than gigantic. He believes in the principle: local instead of global, human instead of anonymous.

“I think everything is good that moves away from centralization, that does not send everything to America.”

Between pragmatism and philosophy

What impressed me about AI from the start: you can do everything you can imagine. And if it does not work today, then try again in three months.

Moosbrugger acts hands-on. He does not talk about singularities, but about Excel spreadsheets. Not about consciousness, but about calendar integration. But that is precisely where his strength lies: he translates the abstract into something usable in everyday life.

The lesson from the one who teaches

In the conversation with ahead, one sentence stuck especially:

Many people ask me how they can tell what is real. I tell them: believe nothing at all, but learn to understand.

In a world that is digitizing ever faster, that is perhaps the most honest answer one can give.

So that we finally take not only the future, but also the present, more seriously, in a time in which speed defines our existence and exerts pressure where it is hardest to breathe.