A Zurich founder on why reps should rehearse against an AI buyer before they meet a real one, and what that does to the months a company spends waiting to find out.
Fernando Leyra’s father works for a brand that makes its team study every capability of every new product, every three months. “He hates it, and his whole team too,” Fernando says. “They feel it is uninspiring and draining.”
That was the starting point for Kulissa, and the studies Fernando read next told him why his father’s team was right to feel it. After a week, data demonstrates that more than half of what people learned that way is gone. The shape of that curve has been known since the 1880s, and a 2015 replication of the original Ebbinghaus experiment, published in PLOS ONE, found it had held for more than a century: most of what is learned by rote is lost within a day and keeps draining after that. “I decided to build a tool that helps people to acquire new skills in the best known way for us: playing.”
Fernando Leyra is the Co-Founder and CEO of Kulissa, an AI sales simulation system built in Zurich. A company hands Kulissa their playbook and some core information. Kulissa turns that into custom simulations and simulated buyer personas that a rep can roleplay against by voice, then measures how the rep performed. “Imagine your team roleplaying your sales and seeing how they perform around it with auditable metrics before even facing a real customer,” he says. Roleplay itself is not new. What’s new, in his account, is the voice models. “Roleplaying is proven to be the best technology to acquire new knowledge, and now, thanks to the last generation of AI voice models, we can scale it up for big companies that could not afford the scale, continuity and observability of the bottom line in their sales and communications.”
Leader in Spotlight: Read Angela Cutulenco Feature in Executives Diary
Meeting Maya
Kulissa’s first hire on any engagement isn’t a person. “When a company brings us a real sales scenario to simulate, or their documentation, the first thing we do is to meet with our AI Agent called Maya,” Fernando says, and then, in the same breath: “We are human first and AI first at the same time, the symbiosis of this technology applied to enhancing humanity is where the true value exists.”
Maya can build thousands of custom simulations and synthetic personas. Besides that, she exists as a continuous individual that clients can meet via Google Meet or other platforms to receive 1:1 coaching, team auditing and even to hand human resources inquiries to her.
Two people stand behind Kulissa’s core technology. Yanik Almer, Kulissa’s CPO, is the person behind Maya’s personality and the reason the system’s user experience is playful, empathetic and rich. Sergey Novikov, the CTO, is the engineer behind Kulissa’s harness and evals systems, which ensure that the simulations and scoring behave correctly and the agentic pipelines run smoothly. This month Maya was given her own LinkedIn presence, and Fernando’s post on it ran to four words: “WELCOME TO THE TEAM MAYA!”
What Maya measures is the part an enterprise buyer asks about first.
Seven Months Before You Know
Of the many pains Kulissa is built to address, Fernando names two as the most important: the readiness of reps, and the need to digest new information at speed. Maya doesn’t only orchestrate the simulation. She coaches before and after it, and keeps what he calls a metered scoreboard of insights from the team’s progress.
Today, most firms invest large amounts of resources into training people without clear ROI but blind bets. The curve goes up to 7 months before having the information to make a decision.
Fernando Leyra, Co-Founder and CEO, Kulissa
The seven months is not an unusual figure. In a 2025 QuotaPath survey of 114 revenue leaders, six months was the most common ramp period given to a new rep, stretching to 12 to 18 months in companies with long sales cycles. Kulissa’s pitch is to move the reading forward: scale and speed up the training, but also surface the insights earlier so a brand can focus its resources, which Fernando credits to “the highly sophisticated algorithms from Kulissa’s CTO Sergey Novikov.”
He applies the same logic to hiring, which he describes today as a funnel of automated layers that prefilter candidates in “a blind inhuman system.” Scoring CVs by keywords is, in his word, inefficient. “What if we could send 100 candidates to run the same simulation with Maya and extract their soft-skills and readiness before proposing HR the top candidates with scores that actually matter?”
He has run the simulations on himself. “I personally trained a lot with Maya with the ‘sell me this pen’ exercise and now I do it better than the guy from Wolf of Wall Street.”
He came to sales simulation from an unusual direction.
An Actor Who Hacks
Fernando’s profile, a few years ago, read as a creative’s. He studied for a Master’s in Visual Communication at the University of Kassel, spent a year as a trainee concept creative at the JvM Academy in Hamburg, and worked at Jung von Matt LIMMAT in Zurich, where he was a junior art director and where a project for Google took an ADC Switzerland bronze in art direction in 2026. Before that came cinema and the fine arts. He describes himself, in a recent post, as an ex-actor.
“A few years ago if you read mine it was clear that I was creative, with experience both in cinema and the fine arts,” he says of his CV. “But it could not have told you about my experience building hardware from severed pieces, hacking videogames and even soldering components. Things that are inherent to my personality because of my high levels of curiosity.”
This mix of technology, engineering and creativity can’t be taught, as it is not knowledge but a self-discovery trajectory of myself.
Fernando, Co-Founder and CEO, Kulissa
He gets asked what an acting background teaches you about simulating human interactions that a career engineer wouldn’t know, and he says he loves the question. His answer is that the current wave of AI favors people with divergent thinking and multiple backgrounds, because “AI can’t connect dots from unrelated fields like us,” and that this is the creativity that lets him do the two jobs he holds in 2026, which he labels CEO and FDE. The second is a Forward Deployed Engineer role at Family, which he keeps alongside Kulissa. He points to friends on the same path: Manuel Eguren, with a background in cinema and international relations, now an AI Applied Engineer at Celonis, and Matthijs Lothmann, a Senior Automation Specialist at Tesla who also came from film. What they share, he says, is an obsession with automating processes and understanding them from the inside out, and a sense that the old standards of the job market were not built for how they think. “Everything had to fit in a box, to follow an established path and to answer in the same way.” His shorter version: “We, the people, were always more than our CV.”
He first saw what the relationship between a human and an agent could look like on his own laptop.
Clark
Outside Kulissa, Fernando built a personal assistant he calls Clark. “That proved me how the relationship between humans and agents can look like,” he says. “It is fascinating.” In a post this year he described what Clark does: it phones people in Fernando’s cloned voice, holds a real conversation, and turns what was said into action points in Notion or entries in his calendar. Everything runs on one Mac. The only thing that leaves the machine is the transcript.
Then Clark made a call it was not supposed to make in the way it made it. It was built to open by saying it is an AI, and it did say it. Nobody heard it, because of a one-second bug in Fernando’s own code: it started speaking before the audio channel was live. The colleague it rang heard Fernando’s voice asking about a pitch and replied thinking it was him.
That is the risk of this technology in a single sentence. The safeguard is one line of audio, and I broke it by accident. Anyone doing this deliberately just never adds the line.
Fernando, Co-Founder and CEO, Kulissa
Fifteen seconds of someone’s voice, he wrote, is enough to clone it, and he ended the post with a warning to anyone who gets an urgent call about money from a voice they know: “A familiar voice is no longer evidence of anything.” It is a strange thing for a man selling simulated humans to say in public, and it is the clearest statement of what he thinks the technology is: not only, as he puts it, “a game, a tool and the best thing humanity built, but the bridge for humans to do and build a better and more human world.”
Kulissa itself has a start date and a city.
Paris, Then a Skyscraper in Zurich
In July 2026 Fernando flew to Paris for the RAISE Summit Hackathon with a team called TN1, which included Sergey and Yanik. The project reimagined enterprise learning through AI voice and video avatars, roleplaying and gamified learning experiences, built to increase the speed and penetration of company knowledge across a business. It won the Cursor track and $20,000. “Still processing that I flew to Paris to participate and ended up winning,” he wrote afterward, from the Moulin Rouge, and added that the story “didn’t finish with the $20K (just started.)” Press in Spain, where he is from, picked it up, El Español and La Voz de Galicia among them.
Kulissa, the company, followed the same month, with Fernando as CEO. It has had support since from the ETH Entrepreneur Club, where he has been a member since early 2026, and from AltaLab, BCG and Google. Asked for a hard moment in the company’s life so far, he can’t produce one. “I can’t recall one, I feel God is on our side because Kulissa had run smoothly from day zero, we even got a sponsored office at a skyscraper in Zurich for free thanks to people who believed in our mission and support entrepreneurs like us.”
Where Fernando Is Now
This month Kulissa gave Maya a desk on LinkedIn and, on the company page, posted a job opening in binary, for agents only. Fernando’s own post above it read: “We are hiring and nobody told me… I’m thinking that Maya went rogue.” The product can be tried free on kulissa.com, where visitors are invited to sell Maya a pen. The people he most wants reading this are enterprise L&D and sales leaders in telecommunications and insurance, and the ask is direct: more enterprise clients for Kulissa, and recognition of the crucial return on investment that their technology represents.
He keeps one habit that has nothing to do with AI: an alarm five minutes before every appointment in his calendar. Asked what he does outside work, he answers, “I work. That is why I have a business and a work at the same time. Being productive is my passion, and when I need to Netflix and chill, I just do it.” The people who shaped his values, in the order he gives them, are Amancio Ortega, Adrian Saenz, Alejandro Estebaranz, Queen Letizia of Spain and Andy Warhol.
What’s next, in his own sentence: “Fernando Leyra, from 0 to 100M ARR selling AI for humans who also like selling.” The line under all of it is older than Kulissa. “Understanding how things work from the inside out has always been my north star.”
Five Lines from Fernando’s Diary
- “Understanding how things work from the inside out has always been my north star.”
- “Roleplaying is proven to be the best technology to acquire new knowledge.”
- “We can use AI to be more human than ever… or the opposite.”
- “We, the people, were always more than our CV.”
- “Developing at the core of humanity to elevate it. Communication is what makes us human, from cave paintings and the first texts to simulated human interactions with synthetic personas.”
From the Diary of Fernando Leyra
Fernando Leyra is a Zurich-based founder and the Co-Founder and CEO of Kulissa, an AI sales simulation system.
| Role | Co-Founder and CEO, Kulissa |
| Based in | Zurich, Switzerland |
| Previously with | Jung von Matt, 2024 to 2025 |
| Also with | ETH Entrepreneur Club, member since 2026 |
“Understanding how things work from the inside out has always been my north star.”
Editor’s Note
Executives Diary features Fernando Leyra because he answers the question enterprise training rarely does, which is what a person retains a week later, with a product a stranger can picture: a rep, a voice, an AI buyer built from the company’s own playbook, and a score at the end. The route into it, from film school and an art director’s desk to a hackathon win in Paris and a company in Zurich, is told in his order and his words. What convinced the editors is the Clark post, where he published his own mistake and the risk it exposed.
Executives Diary, Editorial
