AI Relationship
Advisor
I redesigned and helped launch ChatEZ, an AI dating assistant that turns chat conversations into clear insights and editable replies.
Role
Working with Lin over four months, I led the UX/UI redesign and visual identity for ChatEZ. I evaluated the earlier version, conducted competitive research, and redesigned the core experience around conversation analysis, relationship guidance, and editable replies. I also built the interface design system and created the logo, mascot, posters, and launch assets.
Working with Lin over four months, I led the UX/UI redesign and visual identity for ChatEZ. I evaluated the earlier version, conducted competitive research, and redesigned the core experience around conversation analysis, relationship guidance, and editable replies. I also built the interface design system and created the logo, mascot, posters, and launch assets.
TEAMAgnes Liu, Lin Yang
Product and Engineering
Product and Engineering
Timeline4 Months
Launch and released
Launch and released
Scope User Interface (iOS / Mobile + Web Admin), Product audit, UX strategy, UI design, design system, handoff and QA
ToolsUser interviews, Figma (design and prototyping)
The features were there. The experience made getting started difficult.
ChatEZ is an AI dating assistant that helps users interpret conversations and decide how to respond. Designed for college students and working adults navigating dating conversations, it offers conversation analysis, relationship insights, and suggested replies.
The product already had a working first version. As its features expanded, I was brought in to rethink the core experience and establish a consistent visual identity for its next release.
ChatEZ is an AI dating assistant that helps users interpret conversations and decide how to respond. Designed for college students and working adults navigating dating conversations, it offers conversation analysis, relationship insights, and suggested replies.
The product already had a working first version. As its features expanded, I was brought in to rethink the core experience and establish a consistent visual identity for its next release.
How might we help users move from an emotionally charged conversation to useful, trustworthy guidance and an editable response—without overwhelming them or taking away their agency?
ChatEZ already had a working first version, but its growing feature set made the experience very difficult to understand and maintain. I was asked to rethink the product rather than simply reskin it. I audited the existing experience, reorganized the core journey, and rebuilt the interface around the moments users valued most: understanding a conversation, interpreting relationship signals, and deciding how to respond.
ChatEZ already had a working first version, but its growing feature set made the experience very difficult to understand and maintain. I was asked to rethink the product rather than simply reskin it. I audited the existing experience, reorganized the core journey, and rebuilt the interface around the moments users valued most: understanding a conversation, interpreting relationship signals, and deciding how to respond.
Rather than treating each AI feature as an independent destination,
we reframed the product as one continuous decision-making journey.
we reframed the product as one continuous decision-making journey.
I used low-fidelity wireframes to explore navigation, chat history, and Chat Profile creation before refining the visual design
Product Experience
Feature 1
Reply & Conversation Analysis
Upload chat screenshots to get conversation insights, suggested replies, and a preview of how your response could fit into the conversation
Upload chat screenshots to get conversation insights, suggested replies, and a preview of how your response could fit into the conversation
Feature 2
Refine your own reply
Polish a draft reply to better express what you want to say.
Polish a draft reply to better express what you want to say.
Feature 3
Replies from longer
conversations
Upload chat text to generate replies informed by a longer conversation.
Upload chat text to generate replies informed by a longer conversation.
Feature 4
Profiles & History
Saving conversations and returning to earlier analyses without restarting the process.
Saving conversations and returning to earlier analyses without restarting the process.
Turning the design audit into clear priorities
I mapped the existing journey and reviewed key screens to identify unclear entry points, disconnected interactions, and gaps in conversation context.
I mapped the existing journey and reviewed key screens to identify unclear entry points, disconnected interactions, and gaps in conversation context.
Original conversation-to-reply flow
Unclear starting point
Weak visual hierarchy, limited guidance, and inconsistent visual language made it unclear how to get started.
Limited support for different needs
Users needed support beyond screenshot-based replies. We needed to evaluate which additional tasks to prioritize.
An incomplete supporting experience
Supporting pages and navigation were underdeveloped, leaving gaps beyond the main reply flow.
Clarify Sign-In and Entry Points
The original sign-in screen used low-contrast supporting text and hid alternative login options, while the wider interface lacked consistent colors, typography, and component styles.
I made sign-in choices visible, clarified the visual hierarchy, and established a cohesive visual identity and reusable UI system across the app.
The original sign-in screen used low-contrast supporting text and hid alternative login options, while the wider interface lacked consistent colors, typography, and component styles.
I made sign-in choices visible, clarified the visual hierarchy, and established a cohesive visual identity and reusable UI system across the app.
Clearer sign-in choices
Made WeChat, phone, and Apple sign-in visible on one screen, with guest access grouped below.
More consistent visual hierarchy
Used clearer text contrast and consistent button styles to distinguish actions from supporting content.
Design around three distinct needs
I interviewed college students and working adults to understand where they needed support in dating conversations. Three needs emerged: finding a reply when they felt stuck, refining a draft they had already written, and making sense of a longer exchange.
I translated these needs into three dedicated modes: Generate a Reply, Refine My Reply, and Analyze a Conversation. Each mode provides a clear starting point based on what the user wants to do.
I interviewed college students and working adults to understand where they needed support in dating conversations. Three needs emerged: finding a reply when they felt stuck, refining a draft they had already written, and making sense of a longer exchange.
I translated these needs into three dedicated modes: Generate a Reply, Refine My Reply, and Analyze a Conversation. Each mode provides a clear starting point based on what the user wants to do.
BEFORE — One screenshot-based path
AFTER — Three task-specific paths
BEFORE — Original generate reply workflow
I kept conversation analysis and reply generation in one workflow, since both support the same goal: helping users decide what to say next.
Rather than repeating the uploaded screenshot across screens, I shifted the focus to AI insights and actionable replies, with clearer transitions between reviewing, generating, and editing.
Rather than repeating the uploaded screenshot across screens, I shifted the focus to AI insights and actionable replies, with clearer transitions between reviewing, generating, and editing.
AFTER — Redesigned generate reply workflow
01 — Clearer starting points
Introduced three task-specific modes and clearer upload guidance, so users know where to begin and what to expect.
02 — More space for useful insightsReduced the prominence of uploaded screenshots and prioritized AI insights and reply suggestions.
03 — Easier-to-scan analysis
Organized longer analyses into clear sections with stronger typography and visual hierarchy, making key takeaways easier to find.
04 — More control over repliesAdded reply editing, regeneration, and simulated previews, so users can adjust a suggestion before copying it.
Keep conversation context connected
The original version supported folders and saved chats, but the connection between saving a conversation, choosing a destination, and revisiting its analysis was unclear.
I reorganized this experience around Chat Profiles, clarified profile creation, and brought past analyses and the option to add more chats into the profile detail page.
The original version supported folders and saved chats, but the connection between saving a conversation, choosing a destination, and revisiting its analysis was unclear.
I reorganized this experience around Chat Profiles, clarified profile creation, and brought past analyses and the option to add more chats into the profile detail page.
BEFORE — Original history workflow
01 — Unclear folder organization
The layout and “Add” label left users unsure which chats belonged to folders, what would be saved, and how folders supported ongoing conversations.
02 — Hard-to-find unfiled chatsAbrupt scrolling transitions made chats outside folders difficult to locate. Visible delete icons competed with browsing and increased the risk of accidental deletion.
03 — Limited conversation previews
Image-only thumbnails made saved chats look like screenshots and offered little context about the conversation or its analysis.
04 — Disconnected access to historyUsers had to exit the current screen to access History, interrupting the flow of reviewing saved conversations.
AFTER — Redesigned history workflow
01 — Direct access to history
The side menu brings History and chat profiles together, making saved conversations easier to find and revisit.
02 — Clear profile creationUsers can create a profile directly from a conversation, with a clear explanation of how it supports personalized replies.
03 — More informative records
Message previews, timestamps, and personality tags provide context for reviewing past conversations and understanding each person.
04 — Guided empty statesNew profiles explain their purpose and offer a clear “Upload Chats” action to help users get started.
Brand Refresh
I refreshed ChatEZ’s visual identity to make the AI chat assistant feel more approachable, expressive, and cohesive across the app and its communications.
I designed the logo, mascot, and a range of avatars, character expressions, and chat stickers. I refined the typography, spacing, and icon system to create a consistent visual language throughout the app, then extended the identity into social media campaigns and large-format posters.
I refreshed ChatEZ’s visual identity to make the AI chat assistant feel more approachable, expressive, and cohesive across the app and its communications.
I designed the logo, mascot, and a range of avatars, character expressions, and chat stickers. I refined the typography, spacing, and icon system to create a consistent visual language throughout the app, then extended the identity into social media campaigns and large-format posters.
Internal workspace for managing chat samples, testing prompts, and reviewing agent runs.
The interface brings an asset library, a visual workflow, and run history into one workspace. Teams can inspect each node’s inputs, prompts, and outputs, adjust expert and evaluator prompts, and revisit previous runs to continue testing.
The interface brings an asset library, a visual workflow, and run history into one workspace. Teams can inspect each node’s inputs, prompts, and outputs, adjust expert and evaluator prompts, and revisit previous runs to continue testing.
Visual Agent Workflow
A connected canvas shows how chat inputs move through expert and evaluator prompts to produce a reply
A connected canvas shows how chat inputs move through expert and evaluator prompts to produce a reply
Prompt Inspection & Testing
Node details expose inputs, prompts, outputs, and metadata, helping the team inspect context and test prompt changes.
Node details expose inputs, prompts, outputs, and metadata, helping the team inspect context and test prompt changes.
Samples & Run History
A shared sample library and run history let the team reuse conversations and reopen previous runs for further testing.
A shared sample library and run history let the team reuse conversations and reopen previous runs for further testing.
Bringing the design into implementation through clear specifications and visual review.
I prepared the final screens, reusable components, and interaction notes for development. During implementation, I reviewed the UI against the designs and documented adjustments to spacing, typography, icons, and component states.
I prepared the final screens, reusable components, and interaction notes for development. During implementation, I reviewed the UI against the designs and documented adjustments to spacing, typography, icons, and component states.
Style Guide
Visual QA
That isn’t all
Over four months, I led ChatEZ’s UX/UI redesign and visual identity and supported the release through developer handoff and visual QA. The work extended across the mobile experience, a reusable UI system, brand assets, and an internal agent operations tool.
The redesigned experience brings conversation analysis, reply generation, and editing into a connected flow. Three task-specific modes give users clearer starting points, while Chat Profiles let them revisit earlier analyses and add more conversation context.
Over four months, I led ChatEZ’s UX/UI redesign and visual identity and supported the release through developer handoff and visual QA. The work extended across the mobile experience, a reusable UI system, brand assets, and an internal agent operations tool.
The redesigned experience brings conversation analysis, reply generation, and editing into a connected flow. Three task-specific modes give users clearer starting points, while Chat Profiles let them revisit earlier analyses and add more conversation context.
Reflection
This project taught me to design around the decisions users need to make when working with AI. I focused on helping users understand the analysis, adjust suggested replies, and choose what to send. Working through implementation also showed me how much the final experience depends on clear specifications and consistent visual review.
This project taught me to design around the decisions users need to make when working with AI. I focused on helping users understand the analysis, adjust suggested replies, and choose what to send. Working through implementation also showed me how much the final experience depends on clear specifications and consistent visual review.
Next Step
I would validate the redesigned flows with users, focusing on whether they can choose the right mode, understand the AI’s guidance, and save or revisit a Chat Profile without assistance. I would also examine how users edit suggested replies to identify where the output needs more flexibility.
I would validate the redesigned flows with users, focusing on whether they can choose the right mode, understand the AI’s guidance, and save or revisit a Chat Profile without assistance. I would also examine how users edit suggested replies to identify where the output needs more flexibility.