Overview
Overview
Overview
As a Product Designer, I spend a significant amount of time conducting user interviews, stakeholder discussions, workshops, and product discovery sessions. Every conversation contains valuable insights that influence product decisions.
As a Product Designer, I spend a significant amount of time conducting user interviews, stakeholder discussions, workshops, and product discovery sessions. Every conversation contains valuable insights that influence product decisions.
But I realized something surprising.
But I realized something surprising.
The more focused I was on understanding users, the less I was able to document what they were saying.
The more focused I was on understanding users, the less I was able to document what they were saying.
That realization eventually became TransVoice — an AI-powered conversation intelligence platform that doesn’t just record meetings, but remembers them.
That realization eventually became TransVoice — an AI-powered conversation intelligence platform that doesn’t just record meetings, but remembers them.
Problem Discovery
Problem Discovery
Problem Discovery
Almost every week, I found myself conducting user interviews, stakeholder discussions, feedback sessions, surveys, and research calls.
Almost every week, I found myself conducting user interviews, stakeholder discussions, feedback sessions, surveys, and research calls.
These conversations are where the most valuable insights emerge. A single comment from a user can completely change the direction of a product.
These conversations are where the most valuable insights emerge. A single comment from a user can completely change the direction of a product.
Because of this, I always tried to stay fully present during conversations — actively listening, asking follow-up questions, and observing user behavior.
Because of this, I always tried to stay fully present during conversations — actively listening, asking follow-up questions, and observing user behavior.
But there was one recurring problem.
But there was one recurring problem.
The more engaged I became in the conversation, the more difficult it became to capture important notes.
The more engaged I became in the conversation, the more difficult it became to capture important notes.
Interesting quotes were forgotten.
Important observations were missed.
Action items disappeared after the meeting ended.
Interesting quotes were forgotten.
Important observations were missed.
Action items disappeared after the meeting ended.
Recording meetings seemed like the obvious solution.
Recording meetings seemed like the obvious solution.
However, recordings quickly became archives that I never wanted to revisit. Listening to a one-hour meeting just to find one important insight wasn’t practical.
However, recordings quickly became archives that I never wanted to revisit. Listening to a one-hour meeting just to find one important insight wasn’t practical.
That led me to a simple question:
That led me to a simple question:
What if I had an AI companion sitting beside me in every meeting — listening, understanding, taking notes, and remembering everything for me?
What if I had an AI companion sitting beside me in every meeting — listening, understanding, taking notes, and remembering everything for me?
Hiring someone for every meeting wasn’t realistic.
Building one with AI was.
Hiring someone for every meeting wasn’t realistic.
Building one with AI was.
That idea became the foundation of TransVoice.
That idea became the foundation of TransVoice.
Defining the MVP
Defining the MVP
Defining the MVP
Before writing a single line of code, I focused on defining the minimum experience that would genuinely solve the problem.
Before writing a single line of code, I focused on defining the minimum experience that would genuinely solve the problem.
Instead of trying to build a feature-heavy application, I identified the core journey I wanted every user to experience.
Instead of trying to build a feature-heavy application, I identified the core journey I wanted every user to experience.
01
Record Conversation
Record Conversation
→
02
Generate Transcript
Generate Transcript
→
03
Create AI Summary
Create AI Summary
→
04
Extract Action Items
Extract Action Items
→
05
Save Conversation
Save Conversation
Once this workflow felt complete, I moved into rapid prototyping using AI-assisted development.
Once this workflow felt complete, I moved into rapid prototyping using AI-assisted development.
Building with AI
Building with AI
Building with AI
Rather than following a traditional development process, I chose to experiment with AI-assisted product development using Google AI Studio and Cursor.
Rather than following a traditional development process, I chose to experiment with AI-assisted product development using Google AI Studio and Cursor.
My objective wasn’t simply to generate code.
My objective wasn’t simply to generate code.
How can structured prompts be used to build, iterate, and refine an entire product while maintaining complete control over the user experience?
How can structured prompts be used to build, iterate, and refine an entire product while maintaining complete control over the user experience?
Instead of asking AI to “build an app,” I broke the product into small, well-defined capabilities.
Instead of asking AI to “build an app,” I broke the product into small, well-defined capabilities.
Each prompt focused on solving one problem at a time.
Each prompt focused on solving one problem at a time.
Some examples included:
Some examples included:
•
Building the audio recording experience
Building the audio recording experience
•
Generating AI-powered transcripts
Generating AI-powered transcripts
•
Creating searchable conversation history
Creating searchable conversation history
•
Synchronizing timestamps with audio playback
Synchronizing timestamps with audio playback
•
Designing a conversational AI interface
Designing a conversational AI interface
•
Adding voice interactions
Adding voice interactions
•
Supporting collaborative workspaces
Supporting collaborative workspaces
•
Enabling multi-microphone meetings
Enabling multi-microphone meetings
This iterative approach allowed me to validate ideas quickly while continuously improving the product experience.
This iterative approach allowed me to validate ideas quickly while continuously improving the product experience.
Designing the Experience
Designing the Experience
Designing the Experience
As the product evolved, I focused less on adding features and more on creating an experience that felt natural.
As the product evolved, I focused less on adding features and more on creating an experience that felt natural.
Every new capability answered one simple question:
Every new capability answered one simple question:
How can conversations become easier to revisit?
How can conversations become easier to revisit?
AI Meeting Intelligence
AI Meeting Intelligence
Every recorded meeting is automatically transformed into structured information.
Every recorded meeting is automatically transformed into structured information.
Instead of ending with an audio file, each conversation generates a transcript, an AI summary, action items, and a searchable conversation history.
Instead of ending with an audio file, each conversation generates a transcript, an AI summary, action items, and a searchable conversation history.
01
Record
Record
→
02
Transcript
Transcript
→
03
Summary
Summary
→
04
Action Items
Action Items
→
05
Conversation Library
Conversation Library
AI Memory Chat
AI Memory Chat
One of the biggest shifts in the product happened when I stopped thinking about meetings as recordings and started thinking about them as knowledge.
One of the biggest shifts in the product happened when I stopped thinking about meetings as recordings and started thinking about them as knowledge.
Users can simply ask:
Users can simply ask:
•
“What did the client say about pricing?”
“What did the client say about pricing?”
•
“Who was assigned the dashboard?”
“Who was assigned the dashboard?”
•
“Summarize yesterday’s discussion.”
“Summarize yesterday’s discussion.”
The AI understands the meeting context and responds instantly.
The AI understands the meeting context and responds instantly.
01
Past Meeting
Past Meeting
→
02
User Question
User Question
→
03
Context Retrieval
Context Retrieval
→
04
AI Response
AI Response
Voice-Based AI Interaction
Voice-Based AI Interaction
If users could chat with meetings, why should they have to type?
If users could chat with meetings, why should they have to type?
I extended the experience by introducing voice conversations with AI.
I extended the experience by introducing voice conversations with AI.
Users can ask questions naturally using their microphone, and the AI responds back in voice — creating an experience that feels more like talking to a teammate than searching through documents.
Users can ask questions naturally using their microphone, and the AI responds back in voice — creating an experience that feels more like talking to a teammate than searching through documents.
01
Voice Question
Voice Question
→
02
Meeting Context
Meeting Context
→
03
AI Understanding
AI Understanding
→
04
Voice Response
Voice Response
Team Workspace & Multi-Microphone Meetings
Team Workspace & Multi-Microphone Meetings
As the product matured, collaboration became the next focus.
As the product matured, collaboration became the next focus.
Teams can create shared workspaces where members join meetings using a meeting code from their own devices.
Teams can create shared workspaces where members join meetings using a meeting code from their own devices.
Instead of relying entirely on speaker diarization, every participant is associated with their own microphone stream, improving speaker identification and making AI insights significantly more accurate.
Instead of relying entirely on speaker diarization, every participant is associated with their own microphone stream, improving speaker identification and making AI insights significantly more accurate.
01
Create Meeting
Create Meeting
→
02
Share Meeting Code
Share Meeting Code
→
03
Participants Join
Participants Join
→
04
Individual Audio Streams
Individual Audio Streams
→
05
Named Speakers
Named Speakers
→
06
AI Insights
AI Insights
Design Decisions
Design Decisions
Design Decisions
Throughout the project, I focused on making the interface feel approachable rather than technical.
Throughout the project, I focused on making the interface feel approachable rather than technical.
Some of the key design decisions included:
Some of the key design decisions included:
•
Glassmorphism-inspired interface
Glassmorphism-inspired interface
•
AI-first interaction model
AI-first interaction model
•
Voice-first conversational experience
Voice-first conversational experience
•
ChatGPT-inspired chat interface
ChatGPT-inspired chat interface
•
Timeline-based transcript navigation
Timeline-based transcript navigation
•
Clean workspace organization
Clean workspace organization
•
Minimal cognitive load throughout the journey
Minimal cognitive load throughout the journey
Challenges
Challenges
Challenges
This project wasn’t just about building features.
This project wasn’t just about building features.
It involved making dozens of product decisions along the way.
It involved making dozens of product decisions along the way.
Some of the biggest challenges included:
Some of the biggest challenges included:
•
Designing prompts that generated reliable results
Designing prompts that generated reliable results
•
Structuring AI responses consistently
Structuring AI responses consistently
•
Keeping conversations searchable as history grew
Keeping conversations searchable as history grew
•
Creating natural voice interactions
Creating natural voice interactions
•
Supporting collaborative meetings without adding complexity
Supporting collaborative meetings without adding complexity
•
Balancing AI capabilities with a simple user experience
Balancing AI capabilities with a simple user experience
Final Outcome
Final Outcome
Final Outcome
What started as an AI note-taking idea gradually evolved into something much larger.
What started as an AI note-taking idea gradually evolved into something much larger.
TransVoice became an AI conversation intelligence platform capable of:
TransVoice became an AI conversation intelligence platform capable of:
•
Recording conversations
Recording conversations
•
Generating transcripts
Generating transcripts
•
Creating AI summaries
Creating AI summaries
•
Extracting action items
Extracting action items
•
Organizing conversation history
Organizing conversation history
•
Supporting team collaboration
Supporting team collaboration
•
Enabling multi-microphone meetings
Enabling multi-microphone meetings
•
Letting users talk to their past meetings using voice or chat
Letting users talk to their past meetings using voice or chat
Reflection
Reflection
Reflection
The biggest lesson from this project wasn’t about AI.
The biggest lesson from this project wasn’t about AI.
It was about product thinking.
It was about product thinking.
I learned that users don’t necessarily want more recordings or longer transcripts.
I learned that users don’t necessarily want more recordings or longer transcripts.
They want answers.
They want answers.
By shifting the experience from “recording meetings” to “building an AI memory for conversations,” the product moved beyond documentation and became a tool that helps people remember, retrieve, and interact with knowledge naturally.
By shifting the experience from “recording meetings” to “building an AI memory for conversations,” the product moved beyond documentation and became a tool that helps people remember, retrieve, and interact with knowledge naturally.
