Porcelain Raven

Porcelain Raven is an original product development effort — not client work. It is a consumer-facing AI stylist concept built to test a specific question: can a conversational assistant be genuinely useful at the moment a person is deciding what to wear, rather than just recommending more things to buy?

Most styling and fashion tools optimise for discovery and purchase. They ignore the wardrobe someone already owns, the fit issues they actually have, and the occasion in front of them. The result is recommendations that feel generic and advice that never survives contact with a real closet.

Prototype-first. Rather than specifying a full product up front, the work started with the narrowest useful loop — describe an occasion, get a wearable answer grounded in items the user actually has — and expanded outward only where the loop broke down. Taste, fit, occasion, and inventory were modelled as distinct inputs so each could be tested independently.

A working prototype with a conversational styling interface, a structured wardrobe model, and an editorial content layer for longer-form styling guidance. Content editing, image handling, and publishing flows were built so the product could carry real editorial voice rather than only machine output.

AI conversational interfaces, structured product/content modelling, rich text and media editing, prototype-driven iteration, consumer UX design.

Active prototype. The core styling loop and editorial layer are working; the concept continues to be refined against real-use scenarios rather than pushed toward premature launch.

How to take an ambiguous consumer AI idea and turn it into something testable — narrowing to the smallest honest loop, deciding what the model actually needs to know, and building enough product around it that the idea can be judged fairly.