Sila Digital Wardrobe
Sila is an AI-powered wardrobe app designed to help users effortlessly style outfits using the clothes they already own.
I designed Sila as a product concept, exploring how AI and UX can bridge the gap between fashion inspiration and real wardrobes.
The Wardrobe Paradox
Many women experience uncertainty and frustration when deciding what to wear. Lack of visibility, difficulty coordinating and overwhelming fashion options lead to impulse purchases or rewearing the same, few items. Despite a growing popularity of digital closet apps, the current options fail to create an easy-to-use, time-saving experience that tech-savvy women need.
For users to find value from Sila, they need to upload items from their closets first. Competitive analysis revealed that users spend an estimated 10 hours setting up digital closets. Sila’s Gmail Sync feature eliminates this pain point by automatically importing past and future purchases in just minutes and provides real-time visibility—something no other digital wardrobe app offers.
Through research, I recognized a need to address users’ struggles with outfit coordination. Style Match is an AI styling feature that helps users easily put together outfits. Users can upload or choose fashion inspiration, and Sila recreates the look using their own clothes.
Research Insights
Analysis from survey data, in-depth interviews, and iOS app reviews of competitors:
Seeing is styling: Only 18% of the average wardrobe is worn regularly, with 77% admitting they forget about clothes they own.
Inspiration drives outfit coordination: 92% decide what to wear by referencing external sources but 86% cited lack of outfit inspiration as their biggest wardrobe challenge.
Exploration
Challenge 1: Setting Up a Digital Closet
To solve for the slow, manual process of uploading closet items, onboarding needed to be fast and effortless. I prioritized automation and explored interactive alternatives beyond static image uploads, including AR filming for bulk detection, a chatbot assistant, and a style quiz that generated a wardrobe. But when I mapped these ideas on an impact/effort matrix, I realized that prioritizing innovation was adding complexity rather than reducing friction. I went back to basics, focusing on user behavior and what aligned with their existing habits.
Challenge 2: Styling Outfits
My early ideas for outfit coordination felt exciting and included ideas such as a "roulette" that created outfits based on location, weather, and mood. However, I realized this approach repeated the same pitfalls as other wardrobe apps. When AI models rely on rigid, rule-based parameters, they fail to capture the nuance of personal style and what users envision a great outfit to be. The solution needed a different approach—one where AI learns from users rather than dictates choices.
Designing AI for Style Aspirations
Most AI fashion apps fail because they rely on rigid tags instead of learning from a user’s actual style preferences. Through hands-on testing, I found that ACloset relied heavily on predefined filters (i.e. “Casual,” “Black Top”), which restricted creativity and led to outfits that felt disconnected from real-life style choices. The issue wasn’t just UI—it was how the AI was trained. Sila actively refines its suggestions over time through structured feedback loops.
Introducing Sila: Wear What You Own
Solution 1: Gmail Sync
Since most users primarily shop online, I designed Gmail Sync to automatically import purchases into their digital closet. By integrating with Gmail, this feature leverages retailer order receipts already in users’ inboxes, reducing onboarding from hours of manual entry to just minutes. Unlike any other closet app, it also keeps wardrobes up to date by automatically adding new purchases, ensuring users always have a real-time view of their closet.
During the first round of user testing, I identified the need to reassure users hesitant about the Gmail feature. Adding a tooltip with an explanatory modal addressed privacy concerns and increased Gmail adoption in first-time onboarding by 30%.
Testers who skipped Gmail onboarding were disappointed by an empty closet. The “Add your first items” path eased this frustration, earning positive feedback. Card sorting led to an average of 18 items added with pre-chosen basics.
Solution 2: Style Match
Because traditional AI models fail to capture how people truly want to dress, I designed Style Match. Users upload fashion inspiration or choose from curated looks, and Sila finds the closest matches in their wardrobe to create cohesive outfits. This makes styling effortless, helping users achieve their aspirational looks with clothes they already own.
While engaging with testers, there were questions on why Sila chose certain items but appreciated that they were in control to replace AI-chosen options. To provide more transparency and build trust in AI recommendations, Sila provides a match score for each item selected.
Gmail Sync and Style Match are positioned as specialized features rather than included on the main navigation. To support users who prefer a more hands-on approach to assembling looks, I designed the Canvas feature, an intuitive drag-and-drop styling interface that allows users to manually create outfits.
Silhouette, Sila’s Design System
Silhouette follows three core principles. AI acts as a co-stylist, learning from user preferences rather than dictating outfits. Clarity in decision-making ensures users see only relevant, wearable choices. The system adapts to context, considering factors like weather, occasion, and personal style evolution.
Learnings
Building Sila showed me that AI in UX should enhance, not replace, human intuition. Users want guidance, not control, making trust and transparency essential. Iteration was key—each test refined the experience. Now, as I explore developing Sila in Cursor, I’m excited to bring these ideas to life.