Autorenprofil
When Tony started Zephior, among all the mistakes he made, there was one that had the most impact on the company’s revenue. Over the last two years of Zephior, he made a lot of classic first-time founder mistakes:
But the most critical of them all… the one mistake that still impacts Zephior’s revenue today, is:
Giving away stuff for free.
Tony cold-approached 50 different companies on LinkedIn, and offered free pilots to anyone who would be considered a „potential“ customer that was never going to pay. It took him six months to realize that free attracts the wrong customers, and that the best deals come from people willing to pay – as this willingness in turn is a validation that what you’re building actually creates value to those people.
Every year, companies spend countless hours responding to RFPs, questionnaires, and tenders. A single response can take 25-30 hours of manual work: copying answers from previous submissions, hunting for compliance documents, and reformatting the same information over and over. For consulting firms, insurance companies, and government contractors, RFPs can account for nearly half of their revenue – yet the process remains painfully manual.
Zephior automates this. The company uses AI to build a knowledge library from previous responses, then generates draft answers for new RFPs in minutes instead of days. The platform learns from every completed project, so it gets smarter over time.
But getting there required learning some hard lessons about what early-stage B2B actually looks like.
The conventional wisdom says companies should offer free trials to reduce friction. In enterprise software, Tony has found the opposite to be true. Free trials attract people who want to „explore“ but have no budget, no urgency, and no authority to buy.
Zephior now charges a minimum of €299 for the trial period. It’s not about the money. It’s about filtering for customers who have a real problem and real intent to solve it. When someone pays for a trial, they show up to meetings prepared, provide actual documents to test with, and give honest feedback because they have skin in the game.
The result? Zephior’s trial-to-paid conversion rate jumped significantly, and the team stopped wasting weeks on prospects who were never serious.
Instead of unlimited usage, Zephior allocates a fixed number of credits per month. This creates natural checkpoints where customers evaluate whether they’re getting value. If they consistently run out, it’s a signal they need to upgrade. If they never use them, the team reaches out to understand why.
Constraints aren’t limitations. They’re decision-forcing mechanisms that help both the company and its customers make better choices. The credit model also ensures controlled expense management for Zephior – as the company mainly pays for token consumption—without surprise costs.
The traditional SaaS model assumes customers can self-serve. In Zephior’s market, that’s rarely true. Setting up a knowledge library, onboarding the first users, integrating with existing workflows: these require hands-on „Forward Deployed“ work.
Rather than fighting this, Tony embraced it. He calls the model „Service as Software“: high-touch, paid onboarding that transitions into automated, scalable product usage. The team sets up the knowledge library for every customer during their trial. It takes a few hours but saves customers weeks of frustration.
Zephior is bootstrapped with CHF 40K in ARR, seven clients, and a clear path to CHF 500K by end of 2026. The company has won grants, secured GPU compute worth over €2.5M, and built a product that genuinely saves customers time, at a price point that makes sense.
For anyone building a B2B startup, Tony’s advice is simple: charge money, embrace constraints, invest in trust, and don’t be afraid to do things that don’t scale. The insights from those early, hands-on customers will shape everything that comes next.
Frankfurt School gGmbH
Adickesallee 32-34
60322 Frankfurt am Main
