Li-Yuan Chiou
Interaction Designer
Based in Seattle
Published researcher on Designer-AI Collaboration.
Former UI/UX Intern worked on AI product design at Logitech and ASUS.
Master of Interaction Design at the University of Washington.
2025
Claira
Visualizing Complex Data Relationships for Enterprise Innovation Teams
Context
Problem
USER INTERVIEWS
Workflow Pain Points
Design Principles
KEy functions
Customer Journey Map
Ideation
Solution
Reflection
Role
Product Designer
Developer
Skills
UIUX Design
Prototyping
Coding
Collaborators
Me
Myself
and I
Duration
14 Days
Context
Why do innovation teams struggle to leverage powerful AI tools safely?
Innovation teams need breakthrough insights that come from connecting internal company data with external research, but security policies prevent them from using powerful AI tools with proprietary information. This creates a fundamental tension between staying secure and accessing the most advanced analytical capabilities.
Problem
Research efficiency drops when information becomes fragmented.
Enterprise researchers spend 28% of their time hunting for information instead of actually innovating with it.
Fragmented Information
Critical insights scattered across internal databases, external websites, and research papers with no unified access
Security Barriers
Teams cannot safely use external AI tools with internal company data, limiting analytical capabilities
Manual Research Burden
Hours wasted manually copying, converting, and organizing data before analysis can even begin
Connection Blindness
Difficulty seeing relationships between different pieces of information leads to missed insights and opportunities
USER INTERVIEWS
What we discovered by watching innovation teams work
In the flow of chatting with AI, sharing sensitive details feels natural and harmless.
Observations
Spends 2+ hours on data gathering
Juggles 20+ browser tabs daily
Loses track of information sources
Manually downloads/converts files
Can't use AI with company data
Quotes
“I spend more time managing information than thinking about it”
“Half my research lives in my browser, the other half in three different systems”
“By the time I organize everything, I've lost my train of thought”
“I end up doing the same search twice because I forgot where I saved something”
“We have amazing internal data but can't safely analyze it with the best tools”
“I know there are connections I'm missing, but I don't know how to find them”
“Search results don't show me the bigger picture”
"I bookmark things and then never find them again"
We discovered that researchers rely on linear search patterns (keywords → list → filter) that completely break down when they need to integrate trusted external sources with confidential company data, forcing them into chaotic workflows of juggling multiple tabs and losing track of information sources.
Design Principles
Bridging familiar search patterns with network thinking
Four design decisions that help users transition from linear search to relationship discovery without losing their cognitive comfort zone.
Design Principles:
Keep reading flow while introducing lateral connections that reveal relationships.
Let users control information depth: summary first, details on demand.
Make data sensitivity visible through clear visual indicators
Enable simultaneous multi-source exploration that reveals hidden patterns instead of one-at-a-time searching.
KEy functions
Visual Relationship Discovery for Secure Enterprise Research
Moving beyond traditional keyword search to enable relationship discovery across secure data boundaries.
Flexible Source Selection
Choose any combination of internal systems (Drive, Slack) and external sources in a single query
Thematic Insight Cards
Transform scattered research into organized topic themes that visually reveal how ideas connect
Rapid AutoFetch
Parallel processing automatically gathers related information from multiple sources simultaneously
Visual Security Controls
Color-coded indicators (red/orange/green) make data sensitivity instantly recognizable during exploration
Workflow Pain Points
How Claira Should Transform User Experience
This journey map represents the design vision for how Privacy Guard AI should integrate into users' natural workflow.
Current workflow
Pain Points: fragmented search, manual file handling, and repetitive prompt engineering.
Pain Points: fragmented search, manual file handling, and repetitive prompt engineering.
Current workflow
Researchers cycle through 8+ disconnected steps—visiting websites, downloading files, converting formats, uploading to systems, writing prompts, and iterating. This fragmented process consumes more time managing information than analyzing it.
Four connected steps replace the manual cycle: enter topic → select sources → review insights → create report. By handling source complexity behind the scenes, researchers focus on discovering patterns instead of juggling files.
Ideation & Concept Development
Exploring Relationship Visualization Patterns
I explored various formats for organizing insight cards and visualizing their relationships—testing radial graphs, linear lists with branches, and hierarchical trees. The breakthrough came from combining familiar vertical stacking with horizontal connection lines, allowing users to maintain their natural reading flow while discovering relationships between topics that would otherwise remain hidden.
Solution
A unified platform that reveals hidden connections across secure data sources
Claira securely connects internal and external data sources, processes them in parallel to generate hierarchical topic clusters, and outputs structured insight cards with clear source attribution.
Internal Data
External Data
LLM:
Security: Access Control, Secure Processing, and Compliance
Thematic Insight Generator
Hierarchical Topic Clustering
Cross-Source Relationship Mapping
Progressive Insight Summarization
AutoFetch Engine
Semantic Connection
Parallel Retrieval
Source Attribution Classifier
Insight Cards
Source Map
Flexible Source Selection
Users choose any combination of internal systems and external platforms in a single search, eliminating the need to query multiple tools separately.
Parallel Information Gathering
The system automatically retrieves related information from all selected sources simultaneously, transforming sequential searching into concurrent discovery.
Hierarchical Relationship View with Security Labels
Topics stack vertically with lines connecting related information. Each connection shows a color label—red for confidential, orange for restricted, green for public—so researchers always know what's safe to share while exploring connections.
Reflection
Impact Metrics
These numbers represent more than efficiency gains—researchers discovered connections they previously missed and spent more time on actual innovation rather than information management.
34%
Reduction in time spent
searching for information.
29%
Decrease in time
needed to process
and understand data.
17%
Higher utilization rate of
insights in final reports.
Key Learnings
This experience not only confronted me with the technical and interactive challenges brought by AI systems but also forced me to rethink the essence of
human-AI collaboration. Particularly in two key challenges: Capability Uncertainty and Output Complexity[1]. I experienced the limitations of traditional
UX methods when facing AI, as well as the strategic role of designers in this context.
The unpredictability of AI tools, especially in early feature development and testing stages, often leads to gaps between design and implementation goals.
Designers struggle to predict the boundaries of AI systems, making accurate judgments during requirement definition and user modeling stages difficult.
This uncertainty made me realize that the designer's task is not merely to optimize the user interface, but to actively participate in "how to define what is
designable", a practice of meta-design. AI capabilities are not fixed; they evolve together with designers during the design process, making design an
action of constantly exploring boundaries with technology and reconstructing possibilities.
On the other hand, the large amount of unstructured and highly variable outputs generated by AI also brings an unprecedented burden to design decisions.
As I observed during the design process, output complexity not only increases users' cognitive load but also places new interpretive demands on the
design interface: we no longer design to simplify options, but to design structures that help users "organize uncertainty", a new framework for data
stewardship. In such contexts, design is no longer about control, but harmony; not about reducing possibilities, but guiding understanding.
The most important thing I learned in the process of designing Claira was not how to operate a technical tool, but how to create possibilities for
human users to participate, intervene, and understand within unstable knowledge and capability boundaries.
[1] Qian Yang, Aaron Steinfeld, Carolyn Rosé, and John Zimmerman. 2020. Re-examining Whether, Why, and How Human-AI Interaction Is Uniquely
Difficult to Design. In Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems (CHI '20). Association for Computing Machinery,
New York, NY, USA, 1–13. https://doi.org/10.1145/3313831.3376301
Li-Yuan Chiou
All rights reserved
Li-Yuan Chiou
Interaction Designer
Based in Seattle
Published researcher on Designer-AI Collaboration.
Former UI/UX Intern worked on AI product design at Logitech and ASUS.
Master of Interaction Design at the University of Washington.
2025
Claira
Visualizing Complex Data Relationships for Enterprise Innovation Teams
Context
Problem
USER INTERVIEWS
Workflow Pain Points
Design Principles
KEy functions
Customer Journey Map
Ideation
Solution
Reflection
Role
Product Designer
Developer
Skills
UIUX Design
Prototyping
Coding
Collaborators
Me
Myself
and I
Duration
14 Days
Context
Why do innovation teams struggle to leverage powerful AI tools safely?
Innovation teams need breakthrough insights that come from connecting internal company data with external research, but security policies prevent them from using powerful AI tools with proprietary information. This creates a fundamental tension between staying secure and accessing the most advanced analytical capabilities.
Problem
Research efficiency drops when information becomes fragmented.
Enterprise researchers spend 28% of their time hunting for information instead of actually innovating with it.
Fragmented Information
Critical insights scattered across internal databases, external websites, and research papers with no unified access
Security Barriers
Teams cannot safely use external AI tools with internal company data, limiting analytical capabilities
Manual Research Burden
Hours wasted manually copying, converting, and organizing data before analysis can even begin
Connection Blindness
Difficulty seeing relationships between different pieces of information leads to missed insights and opportunities
USER INTERVIEWS
What we discovered by watching innovation teams work
In the flow of chatting with AI, sharing sensitive details feels natural and harmless.
Observations
Spends 2+ hours on data gathering
Juggles 20+ browser tabs daily
Loses track of information sources
Manually downloads/converts files
Can't use AI with company data
Quotes
“I spend more time managing information than thinking about it”
“Half my research lives in my browser, the other half in three different systems”
“By the time I organize everything, I've lost my train of thought”
“I end up doing the same search twice because I forgot where I saved something”
“We have amazing internal data but can't safely analyze it with the best tools”
“I know there are connections I'm missing, but I don't know how to find them”
“Search results don't show me the bigger picture”
"I bookmark things and then never find them again"
We discovered that researchers rely on linear search patterns (keywords → list → filter) that completely break down when they need to integrate trusted external sources with confidential company data, forcing them into chaotic workflows of juggling multiple tabs and losing track of information sources.
Design Principles
Bridging familiar search patterns with network thinking
Four design decisions that help users transition from linear search to relationship discovery without losing their cognitive comfort zone.
Design Principles:
Keep reading flow while introducing lateral connections that reveal relationships.
Let users control information depth: summary first, details on demand.
Make data sensitivity visible through clear visual indicators
Enable simultaneous multi-source exploration that reveals hidden patterns instead of one-at-a-time searching.
KEy functions
Visual Relationship Discovery for Secure Enterprise Research
Moving beyond traditional keyword search to enable relationship discovery across secure data boundaries.
Flexible Source Selection
Choose any combination of internal systems (Drive, Slack) and external sources in a single query
Thematic Insight Cards
Transform scattered research into organized topic themes that visually reveal how ideas connect
Rapid AutoFetch
Parallel processing automatically gathers related information from multiple sources simultaneously
Visual Security Controls
Color-coded indicators (red/orange/green) make data sensitivity instantly recognizable during exploration
Workflow Pain Points
How Claira Should Transform User Experience
This journey map represents the design vision for how Privacy Guard AI should integrate into users' natural workflow.
Current workflow
Pain Points: fragmented search, manual file handling, and repetitive prompt engineering.
Pain Points: fragmented search, manual file handling, and repetitive prompt engineering.
Current workflow
Researchers cycle through 8+ disconnected steps—visiting websites, downloading files, converting formats, uploading to systems, writing prompts, and iterating. This fragmented process consumes more time managing information than analyzing it.
Four connected steps replace the manual cycle: enter topic → select sources → review insights → create report. By handling source complexity behind the scenes, researchers focus on discovering patterns instead of juggling files.
Ideation & Concept Development
Exploring Relationship Visualization Patterns
I explored various formats for organizing insight cards and visualizing their relationships—testing radial graphs, linear lists with branches, and hierarchical trees. The breakthrough came from combining familiar vertical stacking with horizontal connection lines, allowing users to maintain their natural reading flow while discovering relationships between topics that would otherwise remain hidden.
Solution
A unified platform that reveals hidden connections across secure data sources
Claira securely connects internal and external data sources, processes them in parallel to generate hierarchical topic clusters, and outputs structured insight cards with clear source attribution.
Internal Data
External Data
LLM:
Security: Access Control, Secure Processing, and Compliance
Thematic Insight Generator
Hierarchical Topic Clustering
Cross-Source Relationship Mapping
Progressive Insight Summarization
AutoFetch Engine
Semantic Connection
Parallel Retrieval
Source Attribution Classifier
Insight Cards
Source Map
Flexible Source Selection
Users choose any combination of internal systems and external platforms in a single search, eliminating the need to query multiple tools separately.
Parallel Information Gathering
The system automatically retrieves related information from all selected sources simultaneously, transforming sequential searching into concurrent discovery.
Hierarchical Relationship View with Security Labels
Topics stack vertically with lines connecting related information. Each connection shows a color label—red for confidential, orange for restricted, green for public—so researchers always know what's safe to share while exploring connections.
Reflection
Impact Metrics
These numbers represent more than efficiency gains—researchers discovered connections they previously missed and spent more time on actual innovation rather than information management.
34%
Reduction in time spent
searching for information.
29%
Decrease in time
needed to process
and understand data.
17%
Higher utilization rate of
insights in final reports.
Key Learnings
This experience not only confronted me with the technical and interactive challenges brought by AI systems but also forced me to rethink the essence of
human-AI collaboration. Particularly in two key challenges: Capability Uncertainty and Output Complexity[1]. I experienced the limitations of traditional
UX methods when facing AI, as well as the strategic role of designers in this context.
The unpredictability of AI tools, especially in early feature development and testing stages, often leads to gaps between design and implementation goals.
Designers struggle to predict the boundaries of AI systems, making accurate judgments during requirement definition and user modeling stages difficult.
This uncertainty made me realize that the designer's task is not merely to optimize the user interface, but to actively participate in "how to define what is
designable", a practice of meta-design. AI capabilities are not fixed; they evolve together with designers during the design process, making design an
action of constantly exploring boundaries with technology and reconstructing possibilities.
On the other hand, the large amount of unstructured and highly variable outputs generated by AI also brings an unprecedented burden to design decisions.
As I observed during the design process, output complexity not only increases users' cognitive load but also places new interpretive demands on the
design interface: we no longer design to simplify options, but to design structures that help users "organize uncertainty", a new framework for data
stewardship. In such contexts, design is no longer about control, but harmony; not about reducing possibilities, but guiding understanding.
The most important thing I learned in the process of designing Claira was not how to operate a technical tool, but how to create possibilities for
human users to participate, intervene, and understand within unstable knowledge and capability boundaries.
[1] Qian Yang, Aaron Steinfeld, Carolyn Rosé, and John Zimmerman. 2020. Re-examining Whether, Why, and How Human-AI Interaction Is Uniquely
Difficult to Design. In Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems (CHI '20). Association for Computing Machinery,
New York, NY, USA, 1–13. https://doi.org/10.1145/3313831.3376301
Li-Yuan Chiou
All rights reserved
Li-Yuan Chiou
Interaction Designer
Based in Seattle
Published researcher on Designer-AI Collaboration.
Former UI/UX Intern worked on AI product design at Logitech and ASUS.
Master of Interaction Design at the University of Washington.
2025
Claira
Visualizing Complex Data Relationships for Enterprise Innovation Teams
Context
Problem
USER INTERVIEWS
Workflow Pain Points
Design Principles
KEy functions
Ideation
Solution
Reflection
Role
Product Designer
Skills
Research
UIUX Design
Prototyping
Collaborators
Product Manager
Engineer
Duration
6 months
Context
Why do innovation teams struggle to leverage powerful AI tools safely?
Innovation teams need breakthrough insights that come from connecting internal company data with external research, but security policies prevent them from using powerful AI tools with proprietary information. This creates a fundamental tension between staying secure and accessing the most advanced analytical capabilities.
Problem
Research efficiency drops when information becomes fragmented.
Enterprise researchers spend 28% of their time hunting for information instead of actually innovating with it.
Fragmented Information
Critical insights scattered across internal databases, external websites, and research papers with no unified access
Security Barriers
Teams cannot safely use external AI tools with internal company data, limiting analytical capabilities
Manual Research Burden
Hours wasted manually copying, converting, and organizing data before analysis can even begin
Connection Blindness
Difficulty seeing relationships between different pieces of information leads to missed insights and opportunities
USER INTERVIEWS
What we discovered by watching innovation teams work
We embedded ourselves with enterprise innovation teams for in-depth workflow observation.
Observations
Spends 2+ hours on data gathering
Juggles 20+ browser tabs daily
Loses track of information sources
Manually downloads/converts files
Can't use AI with company data
Quotes
“I spend more time managing information than thinking about it”
“Half my research lives in my browser, the other half in three different systems”
“By the time I organize everything, I've lost my train of thought”
“I end up doing the same search twice because I forgot where I saved something”
“We have amazing internal data but can't safely analyze it with the best tools”
“I know there are connections I'm missing, but I don't know how to find them”
“Search results don't show me the bigger picture”
"I bookmark things and then never find them again"
We discovered that researchers rely on linear search patterns (keywords → list → filter) that completely break down when they need to integrate trusted external sources with confidential company data, forcing them into chaotic workflows of juggling multiple tabs and losing track of information sources.
Design Principles
Bridging familiar search patterns with network thinking
Four design decisions that help users transition from linear search to relationship discovery without losing their cognitive comfort zone.
Design Principles:
Keep reading flow while introducing lateral connections that reveal relationships.
Let users control information depth: summary first, details on demand.
Make data sensitivity visible through clear visual indicators
Enable simultaneous multi-source exploration that reveals hidden patterns instead of one-at-a-time searching.
KEy functions
Visual Relationship Discovery for Secure Enterprise Research
Moving beyond traditional keyword search to enable relationship discovery across secure data boundaries.
Flexible Source Selection
Choose any combination of internal systems (Drive, Slack) and external sources in a single query
Thematic Insight Cards
Transform scattered research into organized topic themes that visually reveal how ideas connect
Rapid AutoFetch
Parallel processing automatically gathers related information from multiple sources simultaneously
Visual Security Controls
Color-coded indicators (red/orange/green) make data sensitivity instantly recognizable during exploration
Workflow Pain Points
How Claira Should Transform User Experience
Researchers cycle through 8+ disconnected steps—visiting websites, downloading files, converting formats, uploading to systems, writing prompts, and iterating. This fragmented process consumes more time managing information than analyzing it.
Four connected steps replace the manual cycle: enter topic → select sources → review insights → create report. By handling source complexity behind the scenes, researchers focus on discovering patterns instead of juggling files.
Ideation & Concept Development
Exploring Relationship Visualization Patterns
I explored various formats for organizing insight cards and visualizing their relationships—testing radial graphs, linear lists with branches, and hierarchical trees. The breakthrough came from combining familiar vertical stacking with horizontal connection lines, allowing users to maintain their natural reading flow while discovering relationships between topics that would otherwise remain hidden.
Solution
A unified platform that reveals hidden connections across secure data sources
Claira securely connects internal and external data sources, processes them in parallel to generate hierarchical topic clusters, and outputs structured insight cards with clear source attribution.
Internal Data
External Data
LLM:
Security: Access Control, Secure Processing, and Compliance
Thematic Insight Generator
Hierarchical Topic Clustering
Cross-Source Relationship Mapping
Progressive Insight Summarization
AutoFetch Engine
Semantic Connection
Parallel Retrieval
Source Attribution Classifier
Insight Cards
Source Map
Flexible Source Selection
Users choose any combination of internal systems and external platforms in a single search, eliminating the need to query multiple tools separately.
Parallel Information Gathering
The system automatically retrieves related information from all selected sources simultaneously, transforming sequential searching into concurrent discovery.
Hierarchical Relationship View with Security Labels
Topics stack vertically with lines connecting related information. Each connection shows a color label—red for confidential, orange for restricted, green for public—so researchers always know what's safe to share while exploring connections.
Reflection
Impact Metrics
These numbers represent more than efficiency gains—researchers discovered connections they previously missed and spent more time on actual innovation rather than information management.
34%
Reduction in time spent
searching for information.
29%
Decrease in time
needed to process
and understand data.
17%
Higher utilization rate of
insights in final reports.
Key Learnings
This experience not only confronted me with the technical and interactive challenges brought by AI systems but also forced me to rethink the essence of
human-AI collaboration. Particularly in two key challenges: Capability Uncertainty and Output Complexity[1]. I experienced the limitations of traditional
UX methods when facing AI, as well as the strategic role of designers in this context.
The unpredictability of AI tools, especially in early feature development and testing stages, often leads to gaps between design and implementation goals.
Designers struggle to predict the boundaries of AI systems, making accurate judgments during requirement definition and user modeling stages difficult.
This uncertainty made me realize that the designer's task is not merely to optimize the user interface, but to actively participate in "how to define what is
designable", a practice of meta-design. AI capabilities are not fixed; they evolve together with designers during the design process, making design an
action of constantly exploring boundaries with technology and reconstructing possibilities.
On the other hand, the large amount of unstructured and highly variable outputs generated by AI also brings an unprecedented burden to design decisions.
As I observed during the design process, output complexity not only increases users' cognitive load but also places new interpretive demands on the
design interface: we no longer design to simplify options, but to design structures that help users "organize uncertainty", a new framework for data
stewardship. In such contexts, design is no longer about control, but harmony; not about reducing possibilities, but guiding understanding.
The most important thing I learned in the process of designing Claira was not how to operate a technical tool, but how to create possibilities for
human users to participate, intervene, and understand within unstable knowledge and capability boundaries.
[1] Qian Yang, Aaron Steinfeld, Carolyn Rosé, and John Zimmerman. 2020. Re-examining Whether, Why, and How Human-AI Interaction Is Uniquely
Difficult to Design. In Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems (CHI '20). Association for Computing Machinery,
New York, NY, USA, 1–13. https://doi.org/10.1145/3313831.3376301
Li-Yuan Chiou
Interaction Designer
Based in Seattle
Published researcher on Designer-AI Collaboration.
Former UI/UX Intern worked on AI product design at Logitech and ASUS.
Master of Interaction Design at the University of Washington.
2025
Claira
Visualizing Complex Data Relationships for Enterprise Innovation Teams
Context
Problem
USER INTERVIEWS
Workflow Pain Points
Design Principles
KEy functions
Customer Journey Map
Ideation
Solution
Reflection
Role
Product Designer
Developer
Skills
UIUX Design
Prototyping
Coding
Collaborators
Me
Myself
and I
Duration
14 Days
Context
Why do innovation teams struggle to leverage powerful AI tools safely?
Innovation teams need breakthrough insights that come from connecting internal company data with external research, but security policies prevent them from using powerful AI tools with proprietary information. This creates a fundamental tension between staying secure and accessing the most advanced analytical capabilities.
Li-Yuan Chiou
All rights reserved
Back to top
Li-Yuan Chiou
All rights reserved