Leveraging AI to Accelerate a Product Tile Redesign

This project focused on redesigning product tiles on the product listing page to improve clarity, hierarchy, and decision-making. But beyond the interface itself, the larger challenge was execution.

• A compressed two-month timeline
• A downsized team (solo designer)
• A need for high-quality exploration and validation

A traditional UX process wouldn’t allow for enough iteration or depth within these constraints. So I restructured my workflow using AI to compress timelines, expand exploration, and reduce manual overhead across the entire design cycle.

Instead of following a traditional linear process, I used AI to move faster. Using Claude throughout, I cut down on manual work and focused more on decision-making and iteration.

I ran a design audit using Claude’s design critique skill to quickly identify issues with the existing tiles.

I ran an unmoderated usability test on the existing product tiles to establish a baseline and identify issues with clarity and hierarchy. To move quickly and reduce bias, I used Claude to help generate the test structure and questions. Rather than asking it to “write a test,” I prompted it with a clear framework:

What shoppers noticed, trusted, and misread within the product tiles:

To validate findings beyond unmoderated testing, I conducted a moderated testing session in-store with real customers.

I printed individual product tile elements, cut them into cards, and asked users to rank them from most to least important when deciding what to purchase.

In-store testing grounded the redesign in real-world shopping behavior and reinforced key behavioral patterns:

• Price consistently ranked as most important
• User and expert reviews were the strongest trust signal
• Merch badges were secondary and often unclear

Using Claude Cowork with agent-based browsing, I analyzed competitor experiences to identify patterns in:

• Product information hierarchy
• Badge usage and placement
• Trust signal presentation

Insights and opportunities were automatically structured and documented directly into Confluence, reducing the time I would have spent manually synthesizing.

Instead of starting with wireframes, I used Claude to generate 10 high-fidelity design directions based on research findings, identified usability issues, and competitive patterns.

Because we already had an established design system, I could skip low-fidelity exploration and focus directly on layout and hierarchy in realistic UI.

This approach allowed me to:

• Explore more directions than a traditional process
• Quickly visualize different hierarchy models
• Facilitate more concrete stakeholder discussions

From the generated concepts, I used AI to help identify which directions were most valuable to test based on usability risk and potential impact.

AI then assisted in writing unbiased test scripts, structuring realistic user scenarios, and deploying tests directly into UserTesting.

After testing the design variants, I exported the results into structured Excel files and uploaded them to Claude for rapid synthesis.

I prompted it to identify patterns, recurring behaviors, and areas of confusion across participants. This helped quickly surface themes in the raw data, which I then reviewed and refined to ensure accuracy before translating them into final insights.

I used Clause to quickly create readouts for presenting to stakeholders.