Case Study · AI Productivity Tool
How I built an AI meeting tool from scratch
A self-built productivity tool that taught me to think like a product owner, not just a designer.
Client
Personal Startup
Role
Founder & Product Designer
Tools
Cursor, Figma, Supabase
01
The Problem
Freelancers don’t need better transcripts. They need better outcomes.
I kept finishing calls with vague notes and missed action items. Otter and Fireflies gave me transcripts. Nobody wants to reread those.
02
Three decisions that shaped it
03
The Solution
One tool. Three steps. Zero busywork.
Phase 01 · Capture
Record, upload, or link calls from Zoom or Meet. Everything processes in the background.
Phase 02 · Summarize
A short executive summary and bullet takeaways, so you know what happened in under a minute.
Phase 03 · Extract tasks
Small talk filtered out, action items pulled with ownership tags, ready to act on.
04
How I built it
Designed and shipped solo, AI-natively.
Next.js
React
Supabase
Speech-to-text
OpenAI
Claude
How I work with AI
01
Brainstorm
02
Spec
03
Plan in phases
04
Build
05
Review
A security, optimization, and simplify pass after each phase. The spec and the review gates are where I decide what ships, not the AI.
The hard part
A raw transcript in one prompt gives you mush, so I designed capture as a staged flow: transcribe, then clean and segment, then separate passes for the summary and for action items with owners. I specced each stage and its guardrails, built it with AI, and reviewed the output. Reliable beats clever when people act on it.
05
The Final Product
The final product, up close.
06
The Impact
37,000+ demo views. 200+ on the waitlist. One validated problem.
TikTok demo response, from a solo build with no funding and no team.
37,000+
Demo views on TikTok
200+
Waitlist signups
1 person
Solo builder, zero to full-stack MVP