PropTechMarketplaceAI-supervised

PropA2Z

A property marketplace where I gave AI the flows and the IA, kept the visual layer for myself, and hand-designed the hero.

Role: Design supervisor & product validator
Timeline: 2025 · in development
Platform: Web
Status: In development.

// the experiment

Give AI the designer’s seat

PropA2Z is a property marketplace for buyers, renters, and sellers — search, map-based discovery, listing management. But the case study isn’t really about the product. It’s about the process: what happens when you give AI the designer’s seat and take the director’s chair instead.

// the split

What AI did well, and where I came in

The honest division of labour is the whole point, so here it is plainly:

Minimal intervention — user flows and information architecture. The AI researched competitors and optimised from there, and was genuinely good at it.
Where I came in — the visual layer. Several iterations, feeding clean visual references in repeatedly until it was shippable.
Hand-designed by me — the homepage and hero, driven by client feedback and one specific client request.

// what the human was for

Director, not prompter

The quality of AI output is proportional to the quality of the input, so my work moved upstream — constructing the brief, feeding references, and judging every pass against a real user goal. AI cleared the flows and the structure. Taste, and the calls it couldn’t make, stayed mine.

// the work

Search, discovery, and a hero I drew by hand

Core surfaces cover property search, map-based browsing, and listing management for two opposite users — people looking, and people listing.

PropA2Z home page with search entry across 200+ cities — The one surface I designed by hand — the opening promise: owners directly, zero brokerage on rentals.

How PropA2Z works: search and shortlist, talk to owners directly, close the deal — The three-step promise as a flow, not just a headline: search, talk to the owner directly, close.

PropA2Z search results with filters, applied chips, and a recently-listed rail — AI drafted the flow and structure from the brief; I directed the visual layer to shippable.

PropA2Z property detail with specs grid, contact owner, EMI calculator, and scarcity nudge — The detail page carries the decision-making tools — specs, an EMI calculator, a direct line to the owner — not just photos.

PropA2Z property gallery with hero image, thumbnail grid, and more-photos tile — A gallery pattern AI drafted from the brief and I directed to shippable, same as the search results.

// outcome

Designs approved, build in progress

The designs are approved and the build is under way; the public site isn’t live yet. What is durable here is the method — a clear, repeatable division of labour between AI and a human director, and a sharper read on where AI creates real leverage and where it hits a ceiling.