gen-ai
search

Designed for Decathlon GenAI Search, experimented user flows, defined core interaction and visual design patterns.

E-commerce search has a translation problem: users think in contexts and needs, while search engines think in keywords and categories. Large language models can bridge this gap by understanding semantic intent behind queries.

The Challenge

I was part of AI Innovation and Trust team, tasked with imaging an GenAI augumented search experience. As the UX designer, I led the product design process for adding a smart cluster feature to the search results page. Our challenge was to create a user-friendly interface that would help users narrow down to relevant products quickly and easily.

Challenge visualFinal results

Defining the Problem

I conducted a competitive analysis to identify commonalities among competitors and gather valuable insghts for building our experience. This research allowed me to identify design patterns like onboarding, loading experience, explanability, etc.

Research visual
Competitive analysis

To gather insights from stakeholders on what we actually want to build, I initiated a workshop with the product team. We discussed the potential of GenAI in search through Figma prototypes, which helped us visualize the user journey and identify key interaction points. This collaborative approach ensured that we were aligned on the vision and goals for the project.

Animated search demo
Stakeholder workshop guiding slides

As results, we defined the prioritised core feature and user journey, with the focus on creating an intuitive and efficient search experience that would help users find relevant products quickly through a smart cluster feature.

Animated search demo 2
Decision matrix and core feature low-fi demo

Focus Area

Then following the design system, I created a new component for the smart cluster feature.

Smart cluster component design
Anatomy of the smart cluster component

The design goal was to create a familiar pattern that users recognize as groups of recommendations with similar product traits, without distracting users with any elements of AI.

Smart cluster component anatomy details
Part of the documentations

To help the devs create the demo better, I also specified the design of other components and the anatomy of the search results page, which has made the implementation process smoother.

Key Takeaways

  • Designing with GenAI requires balancing innovation with user familiarity. Users may not care if a feature is powered by AI, but they do care about how it helps them achieve their goals.
  • Early-on demos are extremely helpful for aligning on the vision and goals of the project with stakeholders. They provide a tangible reference point for discussions and decision-making.