Designing an AI-powered
personal stylist system
Cloh is an AI-powered personal stylist designed to help women build wardrobe confidence through guided style discovery.
I designed the full conversational system inside WhatsApp, enabling users to express frustrations, preferences and personal style intentions, while progressively building structured style profiles that power future personalised recommendations.
My focus: empowering users to define their personal style through guided AI conversations, while capturing structured data to support future personalised recommendations and ecommerce scaling.
Timeline
Jun 2024 - Present
Role
UX & Product Designer
Team
1x Product Designer
1x Product Manager
1x Back End Developer
1x Fashion Stylist
Impact
- 67% quiz completion on first prototype, validating engagement for style segmentation
- Structured conversational data feeding personal style profiles, supporting future personalised recommendations.
- Built data foundation to support future ecommerce filtering based on personal style tags, colour palettes, and fit preferences.
One-minute project overview
Problem
Women often feel overwhelmed when making fashion choices. Most styling tools either feel too generic, too rigid, or too expensive. Many struggle to define their personal style, make confident purchasing decisions, and avoid impulsive buys.
Problem Space Challenges & Realisations
Building trust in an AI-powered stylist requires more than offering quick recommendations. We needed to design a conversational experience that feels like a supportive friend, while still capturing structured data that could support both the user's personal style discovery and future product recommendations.
Solutions
- Designed Cloh's full conversational system inside WhatsApp, breaking the style analysis into lightweight, engaging micro-interactions:
- Context questions around routine and frustrations
- Outfit scenario tests for real-life situations
- Visual mood-board selections
- Style identity mapping ("style words")
- Structured conversational data mapped into personal style profiles and future ecommerce filtering
Results
- Fully functional MVP system ready for closed beta
- ~67% quiz completion rate on early prototypes
- Personalised style profiles generated via structured conversational data
- Technical foundation built to support future ecommerce scaling and affiliated-based filtering
if you want to see more, keep scrolling :)
Turning style discovery into structured market data
In Brazil, over 480,000 shoppers browse fashion digitally and 63% discover trends via social media. Yet brands miss the chance to capture real-time style intent, relying instead on demographics and historic purchases.
We saw the opportunity to structure discovery at the moment it happens, before taste signals are lost.
The personalisation gap in fashion retail
Without real-time style profiles, brands launch campaigns on guesswork:
- Discovery happens socially, but sales campaigns are generic
- Retargeting lacks aesthetic segmentation
- Loyalty is low, and return rates are high
Brands need better ways to listen, organise, and respond to style intent, not just click behaviour.
A system that personalises style and feeds brand insights
Cloh is intentionally lean — designed to deliver value quickly, with minimal effort. From day one, I focused on structuring user input into taggable, brand-usable data. That meant designing for both personalisation and segmentation at once.
Behind the screens, the system works quietly:
- Inputs like moodboard taps, colour picks and product clicks are invisibly tagged
- Style clusters (e.g. Minimal, Romantic, Street) form automatically based on behaviour
- Product suggestions build over time — informed by previous purchases.
This allows Cloh to act as a "living" style engine: building trust with users, while helping brands target with precision.
01.Instant entry from Instagram or web CTA
Users arrived from different entry points, either Instagram followers and/or website visitors, but all had the same need: start styling immediately.
The flow starts when a user taps either the link on our profile or the CTA button in our landing page. A quick-reply button instantly launches the style quiz — no form fields, no loading screen, just instant interaction.
Strategic impact:
- Captures interest before drop-off
- Adapts natively to Instagram and web behaviour
- Establishes a seamless user expectation from first tap
WhatsApp flow triggered from Instagram or Landing page CTA
02.Adaptive quiz builds style profile invisibly
The chat flow combines multiple choice replies, open-text inputs and voice messages, alongside with carousels (e.g. style scenarios, moodboards, fabric preferences). Every interaction was mapped to a tag, progressively building a structured profile behind the scenes, without overwhelming users.
Design rationale
- Flexible enough to adapt to different behaviours and input preferences
- No signup or onboarding friction
- Tied engagement directly to valuable style data
Carousel interaction prototype
03.Personalised style results before product push
Before suggesting any items, users received a shareable style guide based on their cluster. This built trust, giving them a sense of clarity before seeing product links.
Strategic impact:
- Delivered immediate value (not just a shopping link)
- Strengthened user confidence
- Built emotional buy-in through validation
Style profile result screen (highlight key style cluster traits)
04.Smart shopping suggestions via affiliate links
Product suggestions only appear after the user receives their style profile. Each recommendation is matched to their cluster and powered by affiliate links from major brands.
As users click, save or buy, the system learns what works. This closes the feedback loop, letting Cloh improve future suggestions and segment users based on real product behaviour.
Strategic impact:
- Aligns product discovery with personal style
- Increases click and conversion likelihood
- Feeds purchase data back into the style engine
Shopping feed mockup (affiliated products tied to style profile)
Turning early engagement into long-term business value
Our goal wasn't just to engage users, but to structure data that fuels better targeting, smarter product discovery, and future B2B integrations.
By validating a lightweight, scalable quiz experience, we were able to:
- Generate rich personal style profiles
- Establish a strong feedback loop for affiliate-based ecommerce
- Create a technical foundation for future brand partnerships
Designing with intent, scaling with simplicity
This project reinforced the value of early research, simplicity in flow design, and close collaboration with cross-functional partners, building an intuitive user experience while laying the groundwork for long-term business impact.
Balancing both taught me to think beyond screens and focus on long-term impact — how each interaction feeds into smarter data, better targeting and a stronger affiliate loop. It also reinforced the value of early research and working closely with developers to prototype within constraints.
