CASE STUDY
Ithaca : AI Agents
Ithaca wanted to make crypto options accessible through AI agents that trade for users. The idea was strong, but the product was just a demo: jargon-heavy agent cards, no deposit flow, and no way to track performance.
I joined as the product designer to turn the vision into a functional, confidence-building experience.
My Contribution
Research
UX Design
Design System
User Testing
The Team
The Problem
Users loved the promise of AI-powered trading but hit friction immediately:
Confusing agent cards, no clear explanation of how they worked
No confidence in selecting agents
No visibility after depositing - trust broke down
This gap between vision (accessible AI trading) and reality (opaque, confusing prototype) led to drop-offs and low adoption.
“If I don’t understand the agent, I can’t trust it with my money.”
The User
I centered the redesign around Jason, a persona built from interviews and user testing:
Jason
Comfortable with staking and swapping, but new to options.
Wants to grow his money without diving into complex strategies.
🧠 Mental Model
“I don’t need to know how trades happen, I just want a simple, safe way to earn.”
😣 Pain Points
Overwhelmed by jargon-filled agent cards
Confused about what happens after depositing
Sees raw numbers but lacks clear context
🎯 Design Implication
The app should feel transparent, and approachable, more like a trusted companion than a trading terminal.
My Role
I was responsible for creating Ithaca’s end-to-end product AI agent Experience:
Restructured information architecture
Designed the first deposit flow
Redesigned agent cards to build trust
Created tracking dashboards with clear, contextual insights
Prototyped and validated with internal teams and external crypto users
I worked closely with the CEO, CTO, and engineers to turn abstract specs into user-facing design.
Research
Since I was new to options too, my first challenge was to demystify the product. Learning it myself gave me empathy for users who were also encountering these concepts for the first time.
I evaluated the app using three methods:
Heuristic evaluation
I mapped where the Information Architecture and interactions broke down. The flow Pick agent → Deposit → Track was technically sound, but execution was weighed down by jargon-heavy cards and data that didn’t explain itself.
Usability walkthroughs
I tested with crypto users who were comfortable with DeFi basics (staking, swapping) but had never traded options. Asking them to pick an agent, deposit, and check results revealed the same friction: they could complete the steps, but they didn’t understand what they were doing or why.
Competitor Analysis
There were no direct competitors, so I looked at adjacent platforms:
Hyperliquid Vaults → for simplifying complex trade data into digestible metrics.
Virtuals → for making AI agents feel human and trustworthy.
Result
The research confirmed that Ithaca’s strongest idea, agents as personas was buried under friction. Users needed three things above all: a clear journey, approachable agent profiles, and transparency that explained actions without overwhelming them.
User Flow
I reframed the IA around the natural user journey:
Pick agent → Deposit → Track performance
This stripped the product back to what mattered most and gave users a flow that felt both intuitive and confidence-building.
New Design
The redesign built on what worked in the old design while addressing its flaws.
Agent selection became approachable, with scannable profiles, visuals, and clear performance history. The persona model stayed but was refined, making agents feel less abstract and more relatable.
On the details page, performance was reorganized into simple charts, while a “Trader’s Journal” gave short, plain-language updates on what the AI was doing. Dashboards shifted to highlight outcomes — what users had earned, and how their agent was performing — instead of overwhelming them with raw data.
Design System
Alongside the redesign, I created a scalable design system that balanced speed and consistency. It wasn’t a reinvention of Ithaca’s brand, but an evolution — building reusable patterns that could flex as new agents and features were added.
The system was integrated directly into the team’s workflow. Using Cursor, developers could quickly generate new components based on the tokens and guidelines I defined. This made it faster to experiment, reduced inconsistencies, and ensured the app could scale without slowing down design or engineering.
Outcome
The new version delivered immediate improvements:
Usability Testing: 70% of users could complete deposits without help
Agent Understanding: Comprehension rose from 20% to 70% after redesign
Confidence: Average confidence score improved from 3.1/5 → 4.2/5
Adoption: In the first month, ~1,200 deposits were made by ~800 users
Most importantly, the product finally aligned with its promise: approachable for beginners, transparent for advanced users.
Reflection
Ithaca reinforced a lesson I carry into every project: simplifying isn’t about hiding complexity, it’s about reframing it so people can trust and understand it.
By preserving the persona model and redesigning the flows around clarity and trust, I helped turn an intimidating MVP into a guided, confidence-building experience.