How to Build a Real Business with AI in 2026: One Live Build, One Founder, 30 Riders on the Road
Watch on YouTubeOn September 16, our founder Kasia Sadowska hosted "How to Build a Real Business with AI" together with Founder Institute Austria. The live online session drew 140+ sign-ups from founders, builders and curious first-timers. It answered one question: if building an app now takes minutes, what does it take to build a business that people pay for?
The session had three parts. Connor Sattely, the Founder Institute's new CEO, explained why validation matters more than ever. Kasia built an app live with Lovable and Claude. Then Kees Wouters, a Founder Institute graduate with no coding background, showed how he built Legendary Rides, a self-guided cycling company with bookings in six destinations. Stephan Kreiger, co-director of Founder Institute Austria, moderated the Q&A.
The challenge: building is no longer the hard part
Connor Sattely opened with a blunt point. The gap between idea and product is now close to zero. Anyone can open Lovable or Replit and have something working before the next meeting. The problem is that most of these products never find out whether anybody wants them. Founders skip the uncomfortable first steps: talking to customers, testing what people will pay for, learning what not to build.
Kees put the same shift in numbers. Before he started, he knew of a company that spent 50,000 to 60,000 euros a month on IT development, and of a hiking trips platform that invested about a million euros and waited a year for its first customer. His own costs run at 50 to 100 US dollars a month. Cheap building is the opportunity. Selling is still the work.
What is the fastest way to turn an idea into an app?
Kasia built Roomly live, a pocket room planner. You enter your room, then drag, rotate and resize furniture on screen before you move anything heavy. The explanation, the prompt and the published app took about 10 minutes in total. Lovable's build itself took 3 minutes 3 seconds and 2.3 credits. The steps are the same for any idea:
Write a plan before you prompt. A product requirements document is a recipe. It says who the customer is, what problem you solve, what the smallest useful version looks like and what success means.
Let Claude interview you. Kasia used the FutureHabits.Tech prompt template in Claude. It asks about the idea, the features and the user flow, then writes the finished prompt for Lovable. Answer in your own words, in a few sentences, and it fills the gaps.
Keep it to one problem, one customer, one solution. This is the Startup Rule of One. Founders who build for everyone and solve three problems at once rarely launch.
Build only the smallest useful version. Lovable and Claude will suggest many features. Each one costs credits and adds no proof that people want it.
Edit cheaply. Change details through the chat with a specific instruction, such as a corner radius of 10. Edit text directly on the page, which uses no credits. Use Plan mode when you are not sure your prompt is right, because it costs less than Build mode.
Check on mobile and desktop, then publish. Publishing gives you a shareable link with a built-in SEO review. A custom domain can come later.
How does a non-technical founder go from MVP to paying customers?
Kees studied finance and spent ten years at an international logistics company, including five years running its office in Qatar. After quitting, he traveled for a year in a camper van and rode a 1,000 km gravel race across Rwanda. On that ride he noticed a gap. If you wanted to cycle through a country like Rwanda, you either organized everything yourself or booked a staged, fully guided tour. Legendary Rides sits in between: self-guided trips with a tested route, arranged accommodation, bikes, luggage and a support network.
He joined the Founder Institute in September 2025 with a co-founder, and graduated in January. The start was rough. Their weekly grades were low, and a few weeks in, his co-founder left, which meant doing everything alone with twice the cost and half the speed. Around then he attended Kasia's Lovable workshop and built his first MVP in a single evening.
That first version taught him two things. Building with many small patches produced a messy codebase. And the site was hard for Google and AI search tools to read, because most of the text was not in the page's HTML. So he rebuilt from scratch with Kasia's help on the technical structure, plus outside help on marketing messages, user experience and SEO. His current stack:
- Lovable for the website
- Claude to write the prompts for Lovable, which he says uses fewer credits than prompting Lovable directly
- Supabase for bookings and pricing data
- Stripe for payments
- Semrush, now built into Lovable, for SEO
- Brevo as the CRM
What were the results?
- The platform launched in February and the first trip ran in March
- Six destinations, 30 riders on the road, all rating the trips five stars
- Almost 1,000 newsletter sign-ups and 2,400 Instagram followers
- About 100 website visitors a day, in two languages
- Running costs of 50 to 100 US dollars a month for Claude and Lovable
The first customers came from the closest circle outward. First his girlfriend and a friend, then former colleagues who saw his LinkedIn posts, then strangers who found him through Google and Instagram. His co-founder sends a personal message to every new Instagram follower, which leads to real conversations about what riders want.
What can you take from this?
Start before it is perfect, and expect to rebuild. If you cannot code, your first version will not be your last. Kees put it simply: "Get going, get learning." The rebuild was faster because he knew exactly what to keep.
Decide the structure first. A friend gave Kees one piece of advice before he built his site: know the page structure in advance. Text and styling are easy to change later. Structure is not.
Validate before you prompt. Research the market and your customer first, then write the prompt. The prompt is only as good as the brief behind it.
Sell before you add more. After six destinations, Kees and his co-founder decided that the only thing that mattered was selling what they already had. New destinations come after that.
Let Claude coordinate your tools. Kees no longer prompts Lovable himself. Claude writes the instructions and keeps the other systems in line.
Use AI to spend time with people. The routes are built by a local partner and tested in person. AI took over the website work, which freed him to talk to riders and partners.
Tell everyone what you are building. Kees kept his Founder Institute plans quiet at first. Once he shared them, people started helping, and he met his current co-founder through a friend of a friend.
Ask people who are one step ahead. His advice: find people who already did the research and copy what works.
Check what AI does not check. Lovable flags security issues, ranks them from minor to serious and often blocks risky updates. It does not check whether a generated logo is too close to an existing one. Look that up in a trademark database for your country.
Use the tool you already pay for. Gemini works for writing the prompt just as well as Claude. Most small teams can run on Lovable and Claude, instead of paying for ten separate subscriptions.
Questions founders asked
Lovable or Claude Code? Lovable is for people who do not want to learn to code. Claude Code and an editor suit people who are comfortable with a terminal.
Can Lovable build native mobile apps? Not directly. You build the web app, then wrap it with another tool to publish in the app stores. This is the topic of the next session on October 14.
Do I need Supabase separately? Not as a beginner. Lovable Cloud includes a database. A separate Supabase account gives more control over queries and tables, and lets you connect other front ends later.
Can I use voice instead of typing? Yes. Kasia dictates her prompts, because speaking gets her into a creative flow. Lovable and Claude both accept voice and have mobile apps.
Can I bring a design I made in Claude into Lovable? Yes. Claude connects to Lovable directly.
What comes next
Founder Institute Austria runs free sessions like this one throughout the year. The next online session is on October 14, where Kasia shows how to take a Lovable web app to the app stores. The free prompt template used in this session is available at futurehabits.tech. If you want this workshop for your accelerator, community or team, get in touch.