Status
Prototype complete — every screen wired, 100+ realistic feedback items seeded, end-to-end flows from capture to loop-close working in the browser.
100+
Feedback items
<30s
AI clustering
12
Screens built
6
Source channels
The Challenge
Product teams are drowning in feedback that lives in silos. Customers tell you exactly what to build next — in Slack, in support tickets, in NPS verbatims, on sales calls — but there’s no systematic way to detect emerging issues before they become crises, quantify how widespread a complaint actually is, or close the loop with the customers who told you in the first place. PMs end up prioritizing from opinion instead of signal.
Impact & Results
Signal is a fully-wired product prototype: 100+ realistic feedback items, six source channels, three role-based views (Submitter / PM / Admin), and end-to-end flows from raw capture through Watch Mode regression detection and customer loop-close. Every AI classification *sentiment, theme, severity, RICE) is editable in every view. Every flow is honest about what’s stubbed.
What I Learned
- AI proposes, humans dispose. Trust in AI tagging only works when every classification is editable everywhere — and visually distinct from human-confirmed data until a PM signs off.
- Keyboard-first triage isn’t a nice-to-have. Single-key shortcuts to confirm, flag-as-bug, or route to engineering turn a daily slog into a few minutes. PMs won’t live in a tool that makes them click.
- The loop-close is the highest-leverage feature — and the first one teams skip. Notifying the customer who originally reported an issue when you ship the fix changes the relationship. It’s also what almost no feedback tool actually does.
The Outcome
A complete prototype that takes feedback from raw capture through AI clustering, keyboard triage, RICE prioritization, Jira handoff, and post-ship monitoring — with the customer loop-close baked in as a first-class flow, not an afterthought.
Project Deliverables
- Brand Kit
- Design System
- Product Prototype
- Triage Flow
- Prioritization Engine
- Loop-Close Digest
Key Features
- AI clustering with editable everything — sentiment, theme, severity, and bug detection within 30 seconds of intake; AI tags visually distinct from human-confirmed ones until a PM signs off.
- Keyboard-first Triage Queue — single-key confirm, flag-as-bug, mark noise, or create opportunity. Burn through a day of feedback in minutes.
- Six source channels in one inbox — Slack reactions, HubSpot tickets, NPS verbatims, email, Notion, and manual entry, normalized into one feed with source-of-truth chips.
- RICE scoring you actually trust — AI suggests Reach, Impact, and Confidence; you set Effort. Every score editable from Kanban card, detail page, or spreadsheet view.
- Bugs route directly to engineering — Flag-as-bug opens a modal with an AI-drafted Jira ticket (title, repro steps, severity, project) so bugs skip the opportunity lifecycle and don’t get stuck in discovery.
- Watch Mode catches regressions — ship a fix and Signal watches the theme for 60 days. New feedback lands as post-ship signal so you know within minutes if the fix didn’t actually solve it.
- Loop-close with the customers who told you — when a feature ships, Signal generates a digest of every customer who reported the underlying issue, ready to push to Slack for CX follow-up.
- Eisenhower priority matrix — a 2×2 view of opportunities by impact and effort, so leadership can see at a glance what’s worth doing first.
- Role-based views — Submitter, PM, and Admin each get a tailored surface; the role switcher makes it trivial to see the product from every seat.
- Charts and dashboards — source mix, theme volume over time, and severity breakdowns built in, so trends are visible without exporting to a BI tool.
- Severity badges and source chips as a system — a consistent visual language for severity (Critical → Excellent) and source channel that holds up across every screen.