Carousell

Shop with AI

A report of my WIP work with a mix of released and ongoing efforts.

The potential to use AI to improve buyer experience at Carousell was broad. Multiple teams have plans to implement AI enhancement for their buyer problems.

As the lead designer, I consolidate buyer problems across different categories and phases in the user journey to:

  • Design a coherent AI-enabled shopping experience

  • Use AI-enabled design workflow to improve productivity and expand exploration capacity.

Role

Lead designer

Responsibility

IA, Interaction design, Visual design

Timeline

Jan 2025 to present

Context & problems

Search experience on Carousell, as a secondhand marketplace, was time-consuming in two main ways:

Good deals are hard to come by. Listings aren’t described in a standard way. Buyers have to learn the right keywords and switch between different search results to monitor for good deals.

Effort of treasure hunting

The marketplace has rare items and non-standard custom services. Buyers manually chat with each seller to negotiate and get a quote. 

Manual negotiation

Solution overview

I designed:

  1. A common way for buyers to find AI-enabled shopping flow across shared pages

  2. Two different AI-shopping modes: treasure-hunting mode and utility-focused mode

Natural gateways to AI shopping experience

Introduced natural gateways to the AI shopping mode along the user’s journey.

Tresure hunting mode

Buyers describe their intent in words or images to get a collection not bounded by hard filter attributes.

Utility-focused mode

Home agent helps buyers filter options quickly, collect quotes and initiate booking flow

As part of this project, I also introduced improvements to our design process and design guidelines:

  • Ideation with AI agents

  • Establishing AI design patterns for our design system

Process

Collecting user needs & problems across different categories. 

User needs <> category
Map categories into behaviour dimensions

Creating guiding principles

01 Integrate into buyer's natural behaviour

AI shopping mode should not requires an isolated and bifurcated user flow.

02 Global consistency, local optimization

Shared surfaces should be consistent. In-dept category experience should meet local needs.

Designed shared surfaces to handle multi-category behaviour

Same search suggestion page behaves in expected ways, while showing relevant AI suggestion paths for different category keywords

Treasure hunting UX: Ideating interaction, micro-interaction and motion with AI agent

Sandbox Interaction ideation with design agents

Sandbox motion & micro-interaction ideation with design agents

Created event-tracked prototype for unmoderated interaction testing using Figma Make + Supabase

Transactional flow & conversational flow mapping with AI agent support.

Transactional flow generated with Claude and Figma MCP support. Reviewed and edited by human on Figma :)

Conversational flow

From the booking mgmt flow, ideate inline chat action components with AI

From the booking management flow, generate the comprehensive list of in-line chat action components.

Rapid ideation of UI treatment with sandbox preview.

WIP: AI design library for Carousell

I initiated creating the guidelines and worked with a senior designer to define emerging patterns on our app. We aim to cover: product principles, tone of voice, interaction patterns and design system elements.

In the oven ♨️

This work is still WIP. Reach out to me if you are interested to hear about what’s cooking :)

Go to next project

→ Re-architect a shared ‘Me’ page