Project Name
Project Overview
Deliverables
Duration & Team
Tools
Gear Advisor - Redesigning Trust Logic in AI-Powered Decisions
Gear Advisor is an AI decision agent that acts as a Neutral Broker for users buying high-value ski equipment solving trust deficits and cognitive overload at the highest-stakes moment: the product page itself.
I designed the browser plugin that lives inside e-commerce pages and gives users honest, sourced, skill-matched advice in real time without redirecting them away from where the decision is happening.
-
Plugin UX design & prototyping
-
E-commerce integration research
-
User interviews (co-research)
-
Persona development (3 personas)
-
Usability testing
Winter 2026 · 10 weeks · 2 designers (Brinda & Alina) + 1 mentor (Jason Levine)
Figma · Claude · Vite.js · TypeScript · Kiro
What we Built
Before the process here's the finished product. An AI browser plugin that evaluates any snowboard on screen against your skill profile, flags mismatches honestly, and suggests better-matched alternatives with traceable sources.
The Challenge
Problem Statement
Severe cognitive overload and a trust deficit in ski equipment purchasing
When purchasing high-value, technically complex equipment like snowboards, users face three core challenges:
Information Overload: Technical specs like Sidecut, Flex, and Camber are impossible for beginners to translate into how a board actually feels.
Trust Gap: Retailers are driven by inventory pressure. Users can't trust advice from anyone with a financial stake in the sale.
Decision Fatigue: At $600+ with a 7-year ownership cycle, making the wrong call means years of regret not just a bad purchase.
Users aren't afraid of high prices - they're afraid of buying the wrong thing.
- Core insight from user interviews
6
In-depth user interviews
7yr
Average purchase cycle
3
Core user personas
USER RESEARCH
Research
From emotion to insight: behavioral psychology revealed through interviews
Through in-depth interviews with 6 skiers of varying skill levels (using Affinity Diagramming), we uncovered three core psychological patterns behind purchasing behavior:
-
Expert Ego vs. Safety Fear
Many users choose equipment beyond their ability due to aesthetics or brand preference, creating safety hazards. Behind this is a desire to "look expert" coupled with avoiding confronting their true skill level. For the plugin, this meant skill-fit warnings had to be clear and honest without feeling like a lecture.
-
Mathematical Permission
For beginners transitioning from rental to purchase, the biggest barrier isn't the price — it's the lack of a clear break-even point. Users need concrete numbers to grant themselves psychological permission to buy. This directly shaped the Value Analysis card in the plugin panel.
-
The 7-Year Communication Gap
Snowboard purchase cycles span 7 years, and traditional e-commerce completely loses contact with users after the sale. This framed the plugin's longer-term role — not just a purchase tool, but a persistent equipment companion. Being useful in the off-years builds the trust needed for the next $600 decision.
Personas
Three core user personas
Persona 01
Hesitant Outsider
Rides <5 days/year. Primarily rents. Hesitant about committing to a purchase.
-
Needs "mathematical permission" to justify buying
-
Worried about overpaying or wrong fit
-
Lacks knowledge, relies on others' advice
Persona 02
Informed Progressor
Owns basic gear. Actively seeking an upgrade that matches growing skill.
-
Values performance and technical fit
-
Willing to invest - needs credible guidance
-
Wants specs translated into felt experience
Persona 03
Aesthetic Motivator
Brand and appearance drive decisions. Highest risk of skill mismatch.
-
Brand image is the primary purchase driver
-
May choose gear beyond their ability
-
Needs honest nudge without feeling judged
Key Insight - Shared pain points across all personas
Regardless of skill level, all users cannot translate technical specs into felt experience, distrust retailers with inventory pressure, and lack objective neutral decision support.
DESIGN SOLUTION
Solution 01
Designing AI behavior: from sales assistant to neutral broker
Our core strategy is transforming AI from a "sales assistant" into a Neutral Broker. This isn't just about changing the UI; it's about fundamentally redefining AI's behavioral principles and reasoning logic.
Core Design Principles
Principle 01
Anti-Sycophancy
AI must be willing to say "no" to users. When the system identifies a skill mismatch or safety risk, it must prioritize honest feedback over user compliance.
-
Prioritize survey results over stated user preferences
-
Clearly flag risks and skill mismatches
-
Offer better-matched alternatives instead of simple rejection
Principle 02
Transparency
Every AI recommendation cites its sources, PSIA-AASI standards, OutdoorGearLab reviews, and Reddit community data, allowing users to trace the reasoning process.
-
Cite objective reviews (OutdoorGearLab)
-
Cite community feedback (Reddit)
-
Cite industry standards (PSIA-AASI)
Solution 02
AI Shopping Assistant Plugin
To validate the Neutral Broker concept in real shopping scenarios, I designed a browser plugin that integrates directly into e-commerce product pages. The plugin acts as a persistent decision companion offering contextual guidance without disrupting the browsing experience.
The critical behavior: it reads the product currently on screen and evaluates it against the user's stored skill profile the moment it opens. Assessment first. Alternatives second. Always.
Core Feature Logic
Feature 01
Assessment vs. Preference Separation
Logically, enforce skill assessment based on industry standards (PSIA-AASI) first, then layer on user aesthetic preferences.
-
Assessment (hard constraint) - skill level evaluation
-
Preference (soft constraint) - brand and aesthetics
-
When conflicts arise, safety takes priority
Feature 03
Value Calculator
Help "Hesitant Outsiders" make purchase decisions through 3-year Total Cost of Ownership (TCO).
-
Formula: Purchase price + maintenance - resale value
-
Visualize break-even point
-
Provide "mathematical permission"
Feature 02
Traceable Truth Sources
Establish a multi-dimensional "truth library" to ensure recommendation neutrality and objectivity.
-
Professional standards - PSIA-AASI manual
-
Objective reviews - OutdoorGearLab
-
Community voice - Reddit long-term feedback
-
Financial data - rental prices and resale value
Feature 04
Product Breadth Strategy
Expand product range from boards to maintenance consumables, solving the 7-year cycle problem.
-
Tier 1- Boards, boots (core decisions)
-
Tier 2 - Helmets, outerwear (essentials)
-
Tier 3 - Wax, tuning tools (retention)
-
Maintain contact during off-season with maintenance reminders
Impact & Learnings
Outcomes
Key achievements & future directions
Core Achievements
-
Plugin UX pattern for AI trust. The context-first evaluation framework - assess current product before alternatives is transferable to any high-value purchase category: cameras, bikes, audio equipment, golf clubs.
-
Persona-led information architecture. 3 personas each needed a different primary message and tone in the plugin panel. Persona work that directly shaped component-level decisions.
-
Real API integration validated. Anti-sycophancy and transparency principles held up in live Claude responses not just scripted demos.
-
Neutral Broker business model. Proposed commercialization via Tier 3 consumables (wax, tuning) solving the 7-year retention problem and reframing the AI as a long-term relationship tool.
Key Learnings
Learning 01
Design AI behavior, not just UI
True AI product design means defining what the AI refuses to do, not just how it looks. UX designers in AI need systems thinking
Learning 02
Context is the product
The biggest lever wasn't the visual design or the model, it was context. Knowing which product a user was looking at changed everything.
Learning 03
Trust = willingness to disappoint
Users only trust advisors willing to say no. Every design decision: does this build trust, or optimize for conversion at trust's expense?
Let's Connect
Interested in discussing UX design, AI experiences, or collaboration opportunities?