AI-driven product building · Marketplace · UX design
I built Čistý kout on my own as an AI-first experiment, turning a personal need for cleaning help into a live marketplace.
Visit cistykout.cz
Čistý kout began with a personal need: I was looking for someone to help with cleaning. I also founded a Facebook group around cleaning services, which grew to 70,000 members.
From that starting point, I built a marketplace connecting households with cleaning professionals and small firms. The product brings together finding a provider, describing a job, making a booking and managing the work afterwards.
I was working on a need I had experienced myself, alongside a community I had built around the same topic. The product needed to connect the initial search for help with the practical details of arranging a service.
I also wanted to test an AI-first approach to building a complete product. Čistý kout became the first application of my own agent system built on OpenClaw.
I created the entire project on my own, taking responsibility for product direction, UX design and implementation. AI was part of the development approach from the beginning, using my OpenClaw agent system to support the work.
The product work connects discovery, provider profiles, AI-assisted requests, conversational onboarding and booking workflows. I developed the flows and refined the visual language across the public site and application. Recorded walkthroughs in the portfolio show the real application with demo data.
I chose Čistý kout as the first application of my OpenClaw agent system and used an AI-first workflow across the project, while retaining responsibility for the product, design and implementation.
The experiment resulted in a working product that I built solo and brought into production with real users.
I initially tried to use AI throughout the experience, including writing cleaning listings for users. That approach did not suit them: they preferred to enter the details themselves in a simple form.
I changed the flow so AI generates a draft from their answers. Users can then edit the text before publishing. This keeps the familiar form as the starting point and uses AI to help with the wording.
Providers were reluctant to fill in their profiles. I introduced an AI onboarding chat to guide them through creating a profile in conversation.
The conversational approach substantially increased the profile completion rate. While users preferred a simple form for creating listings, AI-guided conversation worked well for profile onboarding.
In production · Real users
Čistý kout is in production and has real users. What began as a personal search for cleaning help became a complete product built solo through an AI-first workflow.
AI onboarding substantially increased the profile completion rate, addressing providers’ reluctance to fill in their profiles.
The screenshots document redesign work captured from the local application; they do not establish that every screen shown has shipped.
Captures from the working local application, showing discovery, the provider experience and AI-assisted requests.
