Asos, Vibe Coding, Ai

AI STYLIST

Asos, Vibe Coding, Ai

AI STYLIST

Asos, Vibe Coding, Ai

AI STYLIST

Asos, Vibe Coding, Ai

AI STYLIST

if we made discovery conversational and personal, shoppers would find relevant items faster, feel more confident, and keep returning.

Prototype of AI stylist

Search to Stylist

01 /

After proving how small tests could drive big impact in Search, I joined the AI Stylist team. A project stuck in research cycles.  My goal here was to bring momentum and start shipping again.

Search to Stylist

01 /

After proving how small tests could drive big impact in Search, I joined the AI Stylist team. A project stuck in research cycles.  My goal here was to bring momentum and start shipping again.

Search to Stylist

01 /

After proving how small tests could drive big impact in Search, I joined the AI Stylist team. A project stuck in research cycles.  My goal here was to bring momentum and start shipping again.

The Stylist sat inside our Loyalty Tier 1 rewards programme, which massively limited our visibility and traffic.

The first priority was simply to get it out to all users, so we could generate enough traffic to start experimenting properly.

The Stylist sat inside our Loyalty Tier 1 rewards programme, which massively limited our visibility and traffic.

The first priority was simply to get it out to all users, so we could generate enough traffic to start experimenting properly.

Reducing Blank Start

02 /

There were two things I focused on straight away. The first was moving the Stylist out of Loyalty. The team now sits under Search, which makes a lot more sense structurally and lets us open the experience up to all users not just those in Tier 1. That move alone gives us the reach we need to properly test and learn. This was done by a lot of back and forth with Head of product, PM's and other teams. Showing them ideas and getting buy in slowly.

Reducing Blank Start

02 /

There were two things I focused on straight away. The first was moving the Stylist out of Loyalty. The team now sits under Search, which makes a lot more sense structurally and lets us open the experience up to all users not just those in Tier 1. That move alone gives us the reach we need to properly test and learn. This was done by a lot of back and forth with Head of product, PM's and other teams. Showing them ideas and getting buy in slowly.

Reducing Blank Start

02 /

Starter chips & prompts, and on the left was the control

My hypothesis is that by giving users clear, contextual prompts — paired with a new greeting that removes the binary yes/no we’ll guide intent and encourage more meaningful engagement.

My hypothesis is that by giving users clear, contextual prompts paired with a new greeting that removes the binary yes/no we’ll guide intent and encourage more meaningful engagement.

Sucess metrics for stylist

Conversation Start Rate (+10–15 pp target)

Average Messages per Session (↑ from 3 → 5)

Pill Engagement Rate (≥ 20 %)

Sucess metrics for stylist

Conversation Start Rate (+10–15 pp target)

Average Messages per Session (↑ from 3 → 5)

Pill Engagement Rate (≥ 20 %)

Sucess metrics for stylist

Conversation Start Rate (+10–15 pp target)

Average Messages per Session (↑ from 3 → 5)

Pill Engagement Rate (≥ 20 %)

Explore, test and repeat

03 /

In parallel to the experiments, I’ve been looking ahead exploring what the next design iteration of the Stylist could be. The focus has been on entry points where and how the Stylist shows up inside the ASOS app.

Explore, test and repeat

03 /

In parallel to the experiments, I’ve been looking ahead exploring what the next design iteration of the Stylist could be. The focus has been on entry points where and how the Stylist shows up inside the ASOS app.

Explore, test and repeat

03 /

By building it, I got to explore all the nuanced chat interactions. It helped me create an experience that feels as close to real as possible.

My hypothesis was that by giving users clear, contextual prompts — paired with a new greeting that removes the binary yes/no we’ll guide intent and encourage more meaningful engagement.

By building it using claude, I got to explore all the nuanced chat interactions. It helped me create an experience that feels as close to real as possible.

One of the earlier prototypes of chat integration

Earlier prototypes of chat integration through search and product listing pages

In parallel, I drove a series of quick-win UI improvements to realign the Stylist experience with the current ASOS brand and design system.

By building it, I got to explore all the nuanced chat interactions. It helped me create an experience that feels as close to real as possible.

In parallel, I drove a series of quick-win UI improvements to realign the Stylist experience with the current ASOS brand and design system.

I will continue to feed learnings from Search into the Stylist applying what we know about discovery, intent, and behaviour to build a more AI-driven and personalised experience.

First iteration in UI changes

North star

07 /

Since the overall UI and UX for the Stylist are still being defined, I’ve also been experimenting with some more alternative directions looking at how the experience could evolve.

North star

07 /

Since the overall UI and UX for the Stylist are still being defined, I’ve also been experimenting with some more alternative directions looking at how the experience could evolve.

North star

07 /

Since the overall UI and UX for the Stylist are still being defined, I’ve also been experimenting with some more alternative directions looking at how the experience could evolve.

This focused on a completely separate entry point

We’ll continue to feed learnings from Search into the Stylist applying what we know about discovery, intent, and behaviour to build a more AI-driven and personalised experience.

In parallel, I drove a series of quick-win UI improvements to realign the Stylist experience with the current ASOS brand and design system.

i will continue to feed learnings from Search into the Stylist applying what we know about discovery, intent, and behaviour to build a more AI-driven and personalised experience.

if we made discovery conversational and personal, shoppers would find relevant items faster, feel more confident, and keep returning.

if we made discovery conversational and personal, shoppers would find relevant items faster, feel more confident, and keep returning.

Search to Stylist

01 /

Search to Stylist

01 /

After proving how small tests could drive big impact in Search, I joined the AI Stylist team. A project stuck in research cycles. My goal here was to bring momentum and start shipping again.

Search to Stylist

01 /

The Stylist sat inside our Loyalty Tier 1 rewards programme, which massively limited our visibility and traffic.

The first priority was simply to get it out to all users, so we could generate enough traffic to start experimenting properly.

The Stylist sat inside our Loyalty Tier 1 rewards programme, which massively limited our visibility and traffic.

The first priority was simply to get it out to all users, so we could generate enough traffic to start experimenting properly.

Reducing Blank Start

02 /

Reducing Blank Start

02 /

Reducing Blank Start

02 /

My hypothesis is that by giving users clear, contextual prompts — paired with a new greeting that removes the binary yes/no we’ll guide intent and encourage more meaningful engagement.

My hypothesis is that by giving users clear, contextual prompts paired with a new greeting that removes the binary yes/no we’ll guide intent and encourage more meaningful engagement.

Sucess metrics for stylist

Sucess metrics for stylist

Sucess metrics for stylist

Explore, test and repeat

03 /

Explore, test and repeat

03 /

In parallel to the experiments, I’ve been looking ahead exploring what the next design iteration of the Stylist could be. The focus has been on entry points where and how the Stylist shows up inside the ASOS app.

Explore, test and repeat

03 /

By building it, I got to explore all the nuanced chat interactions. It helped me create an experience that feels as close to real as possible.

My hypothesis was that by giving users clear, contextual prompts — paired with a new greeting that removes the binary yes/no we’ll guide intent and encourage more meaningful engagement.

By building it using claude, I got to explore all the nuanced chat interactions. It helped me create an experience that feels as close to real as possible.

In parallel, I drove a series of quick-win UI improvements to realign the Stylist experience with the current ASOS brand and design system.

By building it, I got to explore all the nuanced chat interactions. It helped me create an experience that feels as close to real as possible.

In parallel, I drove a series of quick-win UI improvements to realign the Stylist experience with the current ASOS brand and design system.

I will continue to feed learnings from Search into the Stylist applying what we know about discovery, intent, and behaviour to build a more AI-driven and personalised experience.

North star

07 /

North star

07 /

Since the overall UI and UX for the Stylist are still being defined, I’ve also been experimenting with some more alternative directions looking at how the experience could evolve.

North star

07 /

We’ll continue to feed learnings from Search into the Stylist applying what we know about discovery, intent, and behaviour to build a more AI-driven and personalised experience.

In parallel, I drove a series of quick-win UI improvements to realign the Stylist experience with the current ASOS brand and design system.

i will continue to feed learnings from Search into the Stylist applying what we know about discovery, intent, and behaviour to build a more AI-driven and personalised experience.

After proving how small tests could drive big impact in Search, I joined the AI Stylist team. A project stuck in research cycles.  My goal here was to bring momentum and start shipping again.