Ripple

2025

Feedback signal extraction

An internal AI agent that pulls product feedback from team conversations and creates enriched cards with severity, MRR context, quotes, and evidence.

~3-4x

lift in feedback submissions (40-55 -> 150-200/mo)

sources unified (Gong, Intercom, Slack, Email)

unified feedback inbox replacing scattered channels

The bet

Product feedback was scattered across four channels: Gong calls, Intercom chats, Slack threads, and email. Built an AI agent to extract it into a single inbox of enriched cards, so PMs and designers find signal instead of digging.

Role & collaborators

Solo design and agent-assisted build

Stack

Figma

TypeScript

n8n

Notion

Slack

Codex / OpenAI

Claude Code

The Outcome

Ripple lifted feedback submissions from roughly 40-55 to 150-200 per month and unified four feedback sources.