A concept project: an AI-first scheduling and management tool for a rural yoga shala. The interface is a conversation. No setup screens, no configuration panels, no dropdowns.
Sua Casa Escolástica is a small shala in Castrillo de Bezana, a village in northern Spain. Amaya runs it alone and offers three services: regular yoga classes, themed workshops, and individual treatments.
She manages everything through WhatsApp, a paper notebook and a spreadsheet that has grown beyond what it was made for.
This is a concept project. I wanted to explore one specific design idea: what happens when AI conversation replaces every form in a product, including setup.
I wanted to design what the right tool would actually look like for someone like Amaya, and explore one specific design idea: what happens when AI conversation replaces every form in a product, including setup.
Booking apps exist. Dozens of them. The problem is that they’re all designed for people who treat their business as a system. Amaya doesn’t. She treats it as a practice.
She doesn’t want to configure pricing tiers and automation triggers. She wants to say “I give hatha on Tuesdays at 7pm” and have that work. She wants to cancel a class at 8am because her son got sick and notify six students in thirty seconds, without anyone losing their monthly credit.
The product is not a booking app. It’s an assistant.
The rule was simple: the AI replaces forms, not people.
No setup, only conversation. The first time Amaya opens the app, she doesn’t see a settings panel. She has a six-message conversation. What would have been four screens of forms takes ninety seconds of talking.
The AI proposes, Amaya confirms. Nothing irreversible happens without preview. When she asks to cancel a class, the AI shows the exact message it will send to each student before sending it.
The AI handles consequences, not just actions. When a class gets cancelled, the AI also tracks who had paid for a monthly bundle and adjusts the credit. This is where it earns its place.
This was the hardest part of the project. The final list of what the AI doesn’t do is as long as what it does.
It does not suggest optimal schedules. Amaya knows hers better than any model.
It does not generate analytics. With fifty students, there is nothing to analyse that she doesn’t already know.
It does not replace WhatsApp for students. They keep using WhatsApp. The AI sits behind it, not in front of it.
It does not learn her voice to write in her name without preview. Everything it writes, she sees first.
These restraints are not limitations. They are the product.