VIP.ai: AI Concierge for Luxury Discovery

Designed an AI concierge for luxury shopping on mobile and web. People ask for a piece, save it to a selection, see it on themselves, and find where to buy it, all in one chat.

Context

Luxury shopping is spread across brand sites, boutiques, multi-brand stores, resellers and auction houses. Cars, yachts and art each have their own marketplaces. Shoppers already turn to AI for help, but a general assistant knows nothing about their taste.

Problem

Each tool covers one part of the job. ChatGPT answers questions but doesn't know your taste. Pinterest is for saving, not buying. Stores sell, but only their own brands. Nobody covers the whole path, from what should I get to where do I buy it.

My role

I led product design for VIP.ai on mobile and web. I worked with stakeholders on what to build, ran weekly syncs with the developers, and wrote the documentation the team builds from. I prototyped flows in code to test them on a real phone, and kept our design system in sync with the team, since it's shared with another product.

Key features

Every action tells the concierge more about you. A search, a save or a try-on shapes the next recommendation.

Key decisions

Fixed starters instead of AI suggestions

Why: AI-generated suggestions changed every visit and hid what the product can do. Four fixed starters, like Find a piece & where to buy, show the main features and open a guided flow in one tap.

Trade-off: Less variety on the first screen, but every core feature is one tap away.

Onboarding that tunes the concierge

Why: The concierge is only useful if it knows your taste. People pick categories, styles and brands by photo, and every answer goes into its memory.

Trade-off: More steps up front, so each one can be skipped and finished later.

Memory people can see and edit

Why: Privacy is the top worry luxury shoppers have about AI. Every saved preference shows up in the chat, and all of it lives in one place to edit or delete.

Trade-off: More screens to design and build, but people decide what the concierge keeps.

Outcome

The core experience is designed and running as a working demo on mobile and web, with about 80% of the planned scope built.

Why it mattered

Most of this was designed for the first time. AI shopping patterns are still being invented, and every AI company is racing toward them. VIP.ai let me shape them across two platforms while the technology changed under us.