Mindvalley Stories
Case Study · AI Mastery

A real-estate CEO with zero developers shipped a live construction app in 5

A non-technical CEO built a live app where site teams tap a floor plan to log incidents, with a full approval workflow, in five weeks.

The CEO of a Spanish real estate developer had a 40-home site, scattered incident reports and no developers. In five weeks they shipped a production app their builder, site director and architect now use. It shows that what you need is method and discipline, not a technical background.

Iratxe Rodriguez Garcia
CEO of Investleading, a Spanish real estate developer co-investing with family offices.

Maia Obra: construction incident app for Vive Ginkgo

Before

On the Vive Ginkgo site in Valencia, 40 homes were going up, and incidents arrived by WhatsApp, email or word of mouth. Nothing could be traced or pinned to a floor plan. When defect disputes came after a sale, that scatter became a real liability risk. They had no developers to build a fix.

“On our 40-home construction site (Vive Ginkgo, Valencia (Spain)), incidents were reported via WhatsApp, email or verbally — scattered, untraceable, impossible to reference on floor plans, and a liability risk when post-sale defect disputes arise.”

In their own words

The turning point

They took a spec-first path: Dream, Design, Develop. Before building anything big, they tested the riskiest idea with a one-screen prototype where you pin a spot on a plan. Then they built in buckets that only closed once real users had tested them, with Claude Code writing all of the code.

“Golden rule: everything that varies is CONFIGURATION in Airtable, never code — trade catalogs, approval chains, teams.”

In their own words

After

Now they run a live web app where anyone taps the exact spot on one of 46 real floor plans to log an incident, numbered, photographed, assigned and dated. An approval chain runs from builder to site director to architect, with a signed audit trail, designed with the real team. They have named it Maia Obra and see it as a future platform for other developers.

What it lets people do

Site teams log every incident in one place, tied to the exact spot on a plan, with photos and documents filed automatically. Each person sees what is waiting for them, and the company keeps a clear record when defect questions come up later.

What they learned

They learned that AI proposes and humans decide, and that studying real documents first shapes a better design: 35 real meeting minutes reshaped their data model. Most of all, they found a non-technical leader can ship production software with method, real-user testing and good version control habits.

“AI proposes, humans decide (validations, liability attribution).”

In their own words

Maia Obra: construction incident app for Vive Ginkgo was built during AI Mastery.

Read the full submission
THE PROBLEM: I'm the CEO of Investleading, a Spanish real estate developer co-investing with family offices. On our 40-home construction site (Vive Ginkgo, Valencia (Spain)), incidents were reported via WhatsApp, email or verbally — scattered, untraceable, impossible to reference on floor plans, and a liability risk when post-sale defect disputes arise. ACHIEVEMENTS (5 weeks, zero developers): - A live production web app where anyone taps the exact spot on a floor plan to log an incident — auto-numbered, photographed, assigned, with deadlines - All 46 real floor plans served live from Google Drive; photos & documents auto-filed into one Drive folder per incident - A fully configurable approval workflow (builder resolves with photo evidence → site director validates → architect validates → auto-close), including per-incident circuits and a signed audit trail — designed with our real site director and architect - Role-based action bars ("4 incidents awaiting YOUR validation — oldest: 6 days"), analytics tab with drill-down pivot, generic incidents spanning multiple homes, multi-trade incidents - 32 historical site meeting minutes migrated; next step (in build): weekly meeting agendas and minutes that generate themselves from the data - Named and branded as a future product for other developers: Maia Obra (maiaobra.com secured) TOOLS: Claude Cowork (spec, design system, Airtable/Drive infrastructure, knowledge capture) · Claude Code (100% of the code) · Airtable (database + all configuration) · Next.js on Vercel (auto-deploy from GitHub) · Google Drive with a service account · GitHub (private repo). METHODOLOGY: Spec-first (Dream→Design→Develop) → validate the riskiest piece with a one-screen prototype (pin-on-plan) before building → build in "buckets" that only close when real users have tested them → collect feedback in a numbered improvement log that becomes the next consolidated build order → release checklist before every deploy. Golden rule: everything that varies is CONFIGURATION in Airtable, never code — trade catalogs, approval chains, teams. That's what turns one site's tool into a multi-project platform. LEARNINGS: AI proposes, humans decide (validations, liability attribution). Analyze the real documents before designing (35 real meeting minutes reshaped our data model). Measure before optimizing. Service accounts need Shared Drives. Secrets never touch GitHub. And the big one: a non-technical CEO can ship production software — what's actually required is method, real-user testing discipline, and version control hygiene. Everything we learned is codified in our company knowledge base (ADN doc META_02: "How we build internal apps with AI") so the next app takes days, not weeks. URLs: Live app (production pilot): https://app-incidencias-obra-omega.vercel.app

Written 11 August 2026 Submitted by the storyteller

Every story is written by a Mindvalley student or by someone our teachers have worked with. How we collect and check stories

More from AI Mastery

Maia Obra: construction incident app for Vive Ginkgo: a AI Mastery case study | Mindvalley Stories