ENterprise LLM chatbot: Rocket Sammy
Led the end-to-end design of custom Rocket LLM, Rocket Sammy, a proprietary AI tool that reduced the conversation design cycle from 10 days to 2. Internal hackathon project turned into intern project.
Role:
Product Design Intern
Product Design
Intern
Tools:
Figma + FigJam
Lucid + LLM
Team:
1 Product Designer
1 Conversational Designer
1 Content Designer
Project Duration :
May 2025 - August 2025
Overview:
Automating Conversation Design: From Product Docs to Prototypes in Seconds
Automating Conversation Design: From Product Docs to Prototypes in Seconds
Automating Conversation Design: From Product Docs to Prototypes in Seconds
The AI Generator Sample Flow application automates the generation of sample conversation flows – including happy paths, edge cases, error conditions, and digressions – based on structured product documentation inputs such as PRDs, flowcharts, brand voice and tone guidelines, legal requirements, and other technical specs. This tool helps conversation AI designers, product managers, and engineers accelerate prototyping, testing, and documentation of chatbot/voice assistant flows.
PROBLEM:
How Might We Cut Down the 10-Day Process of Designing Conversational Flows?
How Might We Cut Down the 10-Day Process of Designing Conversational Flows?
How Might We Cut Down the 10-Day Process of Designing Conversational Flows?
Manual conversation design takes 7–10 days per flow. Teams must interpret dense docs, extract logic, and map branches, often leading to:
Inconsistent flows across teams
Delayed dev cycles
Gaps in edge case coverage
Repetitive work across intents
Misalignment with voice & tone
SOLUTION:
Introducing Sammy, the Sample Flow Generator
Introducing Sammy, the Sample Flow Generator
Introducing Sammy, the Sample Flow Generator
This custom application uses AI to streamline the creation of sample flows, cutting the design process time down significantly from 7-10 days to 1 business day.
SOLUTION OVERVIEW:
Input Parsing Module
Input Parsing Module
Input Parsing Module
Supports inputs from PRDs, Lucidcharts, Figma, voice and tone guidelines, and legal requirements, accepted via Doc, PDF, Text, or direct file integrations from Lucid and Figma.
Path Customization Layer
Path Customization Layer
Path Customization Layer
Allowing users to tweak sample flows (modify, extend, or exclude paths) without having to rerun prompt.
Desired Output Formats
Desired Output Formats
Desired Output Formats
Ability to export flow as a Word Doc or PDF.
IMPACT:
Impact at a Glance: Cutting Flow Design Time from 10 Days to Under 2
Impact at a Glance: Cutting Flow Design Time from 10 Days to Under 2
Impact at a Glance: Cutting Flow Design Time from 10 Days to Under 2
Cut flow design time from 10 days to under 2
Secured stakeholder buy-in with MVP prototypes
Improved flow quality without formal content review
Boosted consistency across IVR platforms
Adapted to reorgs by scoping features for limited teams
RESEARCH GOALS & QUESTIONS:
How Can I Evaluate Conversation Flow Prototyping, Process Pain Points, and Potential for Automation?
How Can I Evaluate Conversation Flow Prototyping, Process Pain Points, and Potential for Automation?
How Can I Evaluate Conversation Flow Prototyping, Process Pain Points, and Potential for Automation?
STAKEHOLDER INTERVIEWS:
"Our Flows Miss Edge Cases and Sound Inconsistent"
"Our Flows Miss Edge Cases and Sound Inconsistent"
"Our Flows Miss Edge Cases and Sound Inconsistent"
Teams mainly struggled to maintain quality and coverage without a standardized process, resulting in incomplete or off-brand experiences.
Contextual inquries takeaway:
Not Enough Time to Test
Not Enough Time to Test
I observed that teams often rush testing due to time lost on manual flow creation and transfer.
No Single Source of Truth
No Single Source of Truth
No consistent way to track how flows were created or which inputs were used.
Time-Consuming Reviews
Time-Consuming Reviews
Flow approvals from content and legal took days due to manual review.
SOLUTION IDEATIONS:
Automation with AI
Automating flow generation can reclaim time and improve UAT efficiency. Supports Rocket's mission of leveraging AI within their workspace.
Centralized Workplace
Generator should act as a centralized, structured workspace that ties generated flows directly to their source inputs.
Supports Flow Updates
Generator must support easy regeneration and versioning of flows based on updated inputs to reduce maintenance overhead.
Solution ideations:
Automation with AI
I observed that teams often rush testing due to time lost on manual flow creation and transfer.
Centralized Workplace
Generator should act as a centralized, structured workspace that ties generated flows directly to their source inputs.
Supports Flow Updates
Generator must support easy regeneration and versioning of flows based on updated inputs to reduce maintenance overhead.
SOLUTION IDEATIONS:
Automation with AI
Automating flow generation can reclaim time and improve UAT efficiency. Supports Rocket's mission of leveraging AI within their workspace.
Centralized Workplace
Generator should act as a centralized, structured workspace that ties generated flows directly to their source inputs.
Supports Flow Updates
Generator must support easy regeneration and versioning of flows based on updated inputs to reduce maintenance overhead.
USE CASES:
Aligning Stakeholders with Use Cases
Aligning Stakeholders with Use Cases
Aligning Stakeholders with Use Cases
This use case shows how automating conversation flow generation helped cross-functional teams rapidly prototype retention strategies, aligning design, legal, and business goals at a critical client touchpoint.
MOCKUPS & Design EXPLORATIONS:
Simplistic Mockup to Test
Simplistic Mockup to Test
Simplistic Mockup to Test
Users felt the mockup was "too plain" and "not personal" enough.
Users didn't know what they could attach as a document.
Users didn't know how to start their conversation with the chatbot
Offering Clearer Cues to Support Natural User Progression
Offering Clearer Cues to Support Natural User Progression
Offering Clearer Cues to Support Natural User Progression
Smart Starting Prompts:
Help users confidently initiate conversations and explore key actions faster.
Replaced "Attach" button with Direct Plug-ins:
Direct platform integrations (e.g., Figma, Lucid) to streamline workflow continuity.
Reference of Human Name for Personalization:
Create a more conversational and tailored experience.
DIAGRAM FLOW:
With the UI Set, Refining the Chatbot Conversation Flow
With the UI Set, Refining the Chatbot Conversation Flow
Built structured flow diagrams, mapping both happy and error paths, to communicate complex conversation logic across design, engineering, and product teams.
FINAL DESIGN:
Rocket Sammy Says Hello!
Rocket Sammy Says Hello!
Built structured flow diagrams, mapping both happy and error paths, to communicate complex conversation logic across design, engineering, and product teams.
Takeaways:
Exploring AI Design and Navigating Layoffs
Exploring AI Design and Navigating Layoffs
Leveraging Figma Make - A great way to create quick prototypes with my designs to persuade stakeholders and engineers on the features and pathways I was striving for.
Shifted Priorities – Midway through this project, my team got laid off which forced me to re-scope the project to fit limited resources within engineering, breaking down features into an MVP.
Maintaining Quality with Limited Content Review – Without a dedicated content team, I proactively applied voice heuristics, ran internal usability tests, and gathered cross-functional feedback to uphold a high standard of quality for across Sammy's flow and design.
Next steps:
Catching Performance Bugs and Updating Content
Catching Performance Bugs and Updating Content
Testing & QA: Conduct thorough testing, including functional, usability, and edge-case testing, to catch bugs and improve user experience.
Performance Monitoring: Set up analytics to track chatbot interactions, user satisfaction, drop-off points, and error rates.
Content Updates & Expansion: Continuously update the chatbot’s knowledge base and add new intents or conversational paths as needed.
ENterprise LLM chatbot: Rocket Sammy
Led the end-to-end design of custom Rocket LLM, Rocket Sammy, a proprietary AI tool that reduced the conversation design cycle from 10 days to 2. Internal hackathon project turned into intern project.
Role:
Product Design Intern
Product Design
Intern
Tools:
Figma + FigJam
Lucid + LLM
Team:
1 Product Designer
1 Conversational Designer
1 Content Designer
Project Duration :
May 2025 - August 2025
Overview:
Automating Conversation Design: From Product Docs to Prototypes in Seconds
Automating Conversation Design: From Product Docs to Prototypes in Seconds
Automating Conversation Design: From Product Docs to Prototypes in Seconds
The AI Generator Sample Flow application automates the generation of sample conversation flows – including happy paths, edge cases, error conditions, and digressions – based on structured product documentation inputs such as PRDs, flowcharts, brand voice and tone guidelines, legal requirements, and other technical specs. This tool helps conversation AI designers, product managers, and engineers accelerate prototyping, testing, and documentation of chatbot/voice assistant flows.
PROBLEM:
How Might We Cut Down the 10-Day Process of Designing Conversational Flows?
How Might We Cut Down the 10-Day Process of Designing Conversational Flows?
How Might We Cut Down the 10-Day Process of Designing Conversational Flows?
Manual conversation design takes 7–10 days per flow. Teams must interpret dense docs, extract logic, and map branches, often leading to:
Inconsistent flows across teams
Delayed dev cycles
Gaps in edge case coverage
Repetitive work across intents
Misalignment with voice & tone
SOLUTION:
Introducing Sammy, the Sample Flow Generator
Introducing Sammy, the Sample Flow Generator
Introducing Sammy, the Sample Flow Generator
This custom application uses AI to streamline the creation of sample flows, cutting the design process time down significantly from 7-10 days to 1 business day.
SOLUTION OVERVIEW:
Input Parsing Module
Input Parsing Module
Input Parsing Module
Supports inputs from PRDs, Lucidcharts, Figma, voice and tone guidelines, and legal requirements, accepted via Doc, PDF, Text, or direct file integrations from Lucid and Figma.
Path Customization Layer
Path Customization Layer
Path Customization Layer
Allowing users to tweak sample flows (modify, extend, or exclude paths) without having to rerun prompt.
Desired Output Formats
Desired Output Formats
Desired Output Formats
Ability to export flow as a Word Doc or PDF.
IMPACT:
Impact at a Glance: Cutting Flow Design Time from 10 Days to Under 2
Impact at a Glance: Cutting Flow Design Time from 10 Days to Under 2
Impact at a Glance: Cutting Flow Design Time from 10 Days to Under 2
Cut flow design time from 10 days to under 2
Secured stakeholder buy-in with MVP prototypes
Improved flow quality without formal content review
Boosted consistency across IVR platforms
Adapted to reorgs by scoping features for limited teams
RESEARCH GOALS & QUESTIONS:
How Can I Evaluate Conversation Flow Prototyping, Process Pain Points, and Potential for Automation?
How Can I Evaluate Conversation Flow Prototyping, Process Pain Points, and Potential for Automation?
How Can I Evaluate Conversation Flow Prototyping, Process Pain Points, and Potential for Automation?
STAKEHOLDER INTERVIEWS:
"Our Flows Miss Edge Cases and Sound Inconsistent"
"Our Flows Miss Edge Cases and Sound Inconsistent"
"Our Flows Miss Edge Cases and Sound Inconsistent"
Teams mainly struggled to maintain quality and coverage without a standardized process, resulting in incomplete or off-brand experiences.
Contextual inquries takeaway:
Not Enough Time to Test
Not Enough Time to Test
I observed that teams often rush testing due to time lost on manual flow creation and transfer.
No Single Source of Truth
No Single Source of Truth
No consistent way to track how flows were created or which inputs were used.
Time-Consuming Reviews
Time-Consuming Reviews
Flow approvals from content and legal took days due to manual review.
SOLUTION IDEATIONS:
Automation with AI
Automating flow generation can reclaim time and improve UAT efficiency. Supports Rocket's mission of leveraging AI within their workspace.
Centralized Workplace
Generator should act as a centralized, structured workspace that ties generated flows directly to their source inputs.
Supports Flow Updates
Generator must support easy regeneration and versioning of flows based on updated inputs to reduce maintenance overhead.
Solution ideations:
Automation with AI
I observed that teams often rush testing due to time lost on manual flow creation and transfer.
Centralized Workplace
Generator should act as a centralized, structured workspace that ties generated flows directly to their source inputs.
Supports Flow Updates
Generator must support easy regeneration and versioning of flows based on updated inputs to reduce maintenance overhead.
SOLUTION IDEATIONS:
Automation with AI
Automating flow generation can reclaim time and improve UAT efficiency. Supports Rocket's mission of leveraging AI within their workspace.
Centralized Workplace
Generator should act as a centralized, structured workspace that ties generated flows directly to their source inputs.
Supports Flow Updates
Generator must support easy regeneration and versioning of flows based on updated inputs to reduce maintenance overhead.
USE CASES:
Aligning Stakeholders with Use Cases
Aligning Stakeholders with Use Cases
Aligning Stakeholders with Use Cases
This use case shows how automating conversation flow generation helped cross-functional teams rapidly prototype retention strategies, aligning design, legal, and business goals at a critical client touchpoint.
MOCKUPS & Design EXPLORATIONS:
Simplistic Mockup to Test
Simplistic Mockup to Test
Simplistic Mockup to Test
Users felt the mockup was "too plain" and "not personal" enough.
Users didn't know what they could attach as a document.
Users didn't know how to start their conversation with the chatbot
Offering Clearer Cues to Support Natural User Progression
Offering Clearer Cues to Support Natural User Progression
Offering Clearer Cues to Support Natural User Progression
Smart Starting Prompts:
Help users confidently initiate conversations and explore key actions faster.
Replaced "Attach" button with Direct Plug-ins:
Direct platform integrations (e.g., Figma, Lucid) to streamline workflow continuity.
Reference of Human Name for Personalization:
Create a more conversational and tailored experience.
DIAGRAM FLOW:
With the UI Set, Refining the Chatbot Conversation Flow
With the UI Set, Refining the Chatbot Conversation Flow
Built structured flow diagrams, mapping both happy and error paths, to communicate complex conversation logic across design, engineering, and product teams.
FINAL DESIGN:
Rocket Sammy Says Hello!
Rocket Sammy Says Hello!
Built structured flow diagrams, mapping both happy and error paths, to communicate complex conversation logic across design, engineering, and product teams.
Takeaways:
Exploring AI Design and Navigating Layoffs
Exploring AI Design and Navigating Layoffs
Leveraging Figma Make - A great way to create quick prototypes with my designs to persuade stakeholders and engineers on the features and pathways I was striving for.
Shifted Priorities – Midway through this project, my team got laid off which forced me to re-scope the project to fit limited resources within engineering, breaking down features into an MVP.
Maintaining Quality with Limited Content Review – Without a dedicated content team, I proactively applied voice heuristics, ran internal usability tests, and gathered cross-functional feedback to uphold a high standard of quality for across Sammy's flow and design.
Next steps:
Catching Performance Bugs and Updating Content
Catching Performance Bugs and Updating Content
Testing & QA: Conduct thorough testing, including functional, usability, and edge-case testing, to catch bugs and improve user experience.
Performance Monitoring: Set up analytics to track chatbot interactions, user satisfaction, drop-off points, and error rates.
Content Updates & Expansion: Continuously update the chatbot’s knowledge base and add new intents or conversational paths as needed.
ENterprise LLM chatbot: Rocket Sammy
Led the end-to-end design of custom Rocket LLM, Rocket Sammy, a proprietary AI tool that reduced the conversation design cycle from 10 days to 2. Internal hackathon project turned into intern project.
Role:
Product Design Intern
Product Design
Intern
Tools:
Figma + FigJam
Lucid + LLM
Team:
1 Product Designer
1 Conversational Designer
1 Content Designer
Project Duration :
May 2025 - August 2025
Overview:
Automating Conversation Design: From Product Docs to Prototypes in Seconds
Automating Conversation Design: From Product Docs to Prototypes in Seconds
Automating Conversation Design: From Product Docs to Prototypes in Seconds
The AI Generator Sample Flow application automates the generation of sample conversation flows – including happy paths, edge cases, error conditions, and digressions – based on structured product documentation inputs such as PRDs, flowcharts, brand voice and tone guidelines, legal requirements, and other technical specs. This tool helps conversation AI designers, product managers, and engineers accelerate prototyping, testing, and documentation of chatbot/voice assistant flows.
PROBLEM:
How Might We Cut Down the 10-Day Process of Designing Conversational Flows?
How Might We Cut Down the 10-Day Process of Designing Conversational Flows?
How Might We Cut Down the 10-Day Process of Designing Conversational Flows?
Manual conversation design takes 7–10 days per flow. Teams must interpret dense docs, extract logic, and map branches, often leading to:
Inconsistent flows across teams
Delayed dev cycles
Gaps in edge case coverage
Repetitive work across intents
Misalignment with voice & tone
SOLUTION:
Introducing Sammy, the Sample Flow Generator
Introducing Sammy, the Sample Flow Generator
Introducing Sammy, the Sample Flow Generator
This custom application uses AI to streamline the creation of sample flows, cutting the design process time down significantly from 7-10 days to 1 business day.
SOLUTION OVERVIEW:
Input Parsing Module
Input Parsing Module
Input Parsing Module
Supports inputs from PRDs, Lucidcharts, Figma, voice and tone guidelines, and legal requirements, accepted via Doc, PDF, Text, or direct file integrations from Lucid and Figma.
Path Customization Layer
Path Customization Layer
Path Customization Layer
Allowing users to tweak sample flows (modify, extend, or exclude paths) without having to rerun prompt.
Desired Output Formats
Desired Output Formats
Desired Output Formats
Ability to export flow as a Word Doc or PDF.
IMPACT:
Impact at a Glance: Cutting Flow Design Time from 10 Days to Under 2
Impact at a Glance: Cutting Flow Design Time from 10 Days to Under 2
Impact at a Glance: Cutting Flow Design Time from 10 Days to Under 2
Cut flow design time from 10 days to under 2
Secured stakeholder buy-in with MVP prototypes
Improved flow quality without formal content review
Boosted consistency across IVR platforms
Adapted to reorgs by scoping features for limited teams
RESEARCH GOALS & QUESTIONS:
How Can I Evaluate Conversation Flow Prototyping, Process Pain Points, and Potential for Automation?
How Can I Evaluate Conversation Flow Prototyping, Process Pain Points, and Potential for Automation?
How Can I Evaluate Conversation Flow Prototyping, Process Pain Points, and Potential for Automation?
STAKEHOLDER INTERVIEWS:
"Our Flows Miss Edge Cases and Sound Inconsistent"
"Our Flows Miss Edge Cases and Sound Inconsistent"
"Our Flows Miss Edge Cases and Sound Inconsistent"
Teams mainly struggled to maintain quality and coverage without a standardized process, resulting in incomplete or off-brand experiences.
Contextual inquries takeaway:
Not Enough Time to Test
Not Enough Time to Test
I observed that teams often rush testing due to time lost on manual flow creation and transfer.
No Single Source of Truth
No Single Source of Truth
No consistent way to track how flows were created or which inputs were used.
Time-Consuming Reviews
Time-Consuming Reviews
Flow approvals from content and legal took days due to manual review.
SOLUTION IDEATIONS:
Automation with AI
Automating flow generation can reclaim time and improve UAT efficiency. Supports Rocket's mission of leveraging AI within their workspace.
Centralized Workplace
Generator should act as a centralized, structured workspace that ties generated flows directly to their source inputs.
Supports Flow Updates
Generator must support easy regeneration and versioning of flows based on updated inputs to reduce maintenance overhead.
Solution ideations:
Automation with AI
I observed that teams often rush testing due to time lost on manual flow creation and transfer.
Centralized Workplace
Generator should act as a centralized, structured workspace that ties generated flows directly to their source inputs.
Supports Flow Updates
Generator must support easy regeneration and versioning of flows based on updated inputs to reduce maintenance overhead.
SOLUTION IDEATIONS:
Automation with AI
Automating flow generation can reclaim time and improve UAT efficiency. Supports Rocket's mission of leveraging AI within their workspace.
Centralized Workplace
Generator should act as a centralized, structured workspace that ties generated flows directly to their source inputs.
Supports Flow Updates
Generator must support easy regeneration and versioning of flows based on updated inputs to reduce maintenance overhead.
USE CASES:
Aligning Stakeholders with Use Cases
Aligning Stakeholders with Use Cases
Aligning Stakeholders with Use Cases
This use case shows how automating conversation flow generation helped cross-functional teams rapidly prototype retention strategies, aligning design, legal, and business goals at a critical client touchpoint.
MOCKUPS & Design EXPLORATIONS:
Simplistic Mockup to Test
Simplistic Mockup to Test
Simplistic Mockup to Test
Users felt the mockup was "too plain" and "not personal" enough.
Users didn't know what they could attach as a document.
Users didn't know how to start their conversation with the chatbot
Offering Clearer Cues to Support Natural User Progression
Offering Clearer Cues to Support Natural User Progression
Offering Clearer Cues to Support Natural User Progression
Smart Starting Prompts:
Help users confidently initiate conversations and explore key actions faster.
Replaced "Attach" button with Direct Plug-ins:
Direct platform integrations (e.g., Figma, Lucid) to streamline workflow continuity.
Reference of Human Name for Personalization:
Create a more conversational and tailored experience.
DIAGRAM FLOW:
With the UI Set, Refining the Chatbot Conversation Flow
With the UI Set, Refining the Chatbot Conversation Flow
Built structured flow diagrams, mapping both happy and error paths, to communicate complex conversation logic across design, engineering, and product teams.
FINAL DESIGN:
Rocket Sammy Says Hello!
Rocket Sammy Says Hello!
Built structured flow diagrams, mapping both happy and error paths, to communicate complex conversation logic across design, engineering, and product teams.
Takeaways:
Exploring AI Design and Navigating Layoffs
Exploring AI Design and Navigating Layoffs
Leveraging Figma Make - A great way to create quick prototypes with my designs to persuade stakeholders and engineers on the features and pathways I was striving for.
Shifted Priorities – Midway through this project, my team got laid off which forced me to re-scope the project to fit limited resources within engineering, breaking down features into an MVP.
Maintaining Quality with Limited Content Review – Without a dedicated content team, I proactively applied voice heuristics, ran internal usability tests, and gathered cross-functional feedback to uphold a high standard of quality for across Sammy's flow and design.
Next steps:
Catching Performance Bugs and Updating Content
Catching Performance Bugs and Updating Content
Testing & QA: Conduct thorough testing, including functional, usability, and edge-case testing, to catch bugs and improve user experience.
Performance Monitoring: Set up analytics to track chatbot interactions, user satisfaction, drop-off points, and error rates.
Content Updates & Expansion: Continuously update the chatbot’s knowledge base and add new intents or conversational paths as needed.