Request Prioritization & Scoring Engine
Score every incoming request with a consistent model (RICE, WSJF, or a value-vs-effort grid) so your team prioritizes by impact instead of by who escalated the loudest - with a human able to override the rank and document why.
A login-protected app that imports your open requests, computes a score per your chosen model, ranks them, lets a prioritization owner adjust inputs or override the rank with a written reason, and publishes an approved priority list plus a clean CSV export for your PM tool.
Before you start
- A list of open requests (CSV or Google Sheet)
- Your scoring model and the weights you want to use
- Effort estimates for each request (rough is fine)
- Free accounts on Vercel, Supabase, and Resend
The problem this kills
Right now your priority list is decided by volume - not the volume of work, the volume of the person shouting. The loudest stakeholder, the most senior escalation, the request that landed in your inbox last: those jump the queue. Quiet, high-impact work waits. Nobody can explain why item A beats item B, so every planning meeting reopens the same argument.
You probably already have a scoring model in your head or on a whiteboard - RICE, WSJF, or a simple value-vs-effort grid. But it lives in a spreadsheet that one person maintains, the formula drifts, and there's no record of who changed a rank or why. So the model gets ignored the moment someone important pushes back.
This tool makes the model the default and the override the exception. Every request gets a consistent, transparent score. A real person can still overrule it - but only with a reason that's captured forever. The argument moves from "who deserves it more" to "is this score right," which is a much shorter argument.
What you'll build
A small web app, just for your team, that:
- Imports your open requests from a CSV or Google Sheet - request id, title, requester, and the inputs your model needs.
- Computes a score for every request using the model you pick: RICE (Reach x Impact x Confidence / Effort), WSJF (Cost of Delay / Job Size), or a custom value-vs-effort grid - with the formula and weights stored as editable config you control.
- Ranks the list highest-priority first, and shows the math so the score is never a black box.
- Lets the prioritization owner review each one - tweak an input, or override the computed rank entirely, with a required written reason.
- Holds an approval gate - nothing is "the priority list" until the owner approves it. The AI computes; a human commits.
- Exports a clean CSV in the exact columns your PM tool (Jira, Asana, Monday, a sheet) expects, so the ranked list flows straight into where work actually gets scheduled.
What's inside the Implementation Plan
The plan is a runbook you paste into an AI coding agent (Claude Code), which then builds the tool with you step by step. No prior coding needed.
Crucially, the plan opens by interviewing you about your business - your current intake process, the systems and sheets you use, the real field names and id conventions in your data, your typical and peak request volumes, which scoring model and weights you use, and your messy edge cases. It reflects a short tailored spec back to you and waits for your thumbs-up before it builds anything. You get a tool shaped to how your team actually prioritizes - not a generic template you have to bend yourself around.
Inside you'll find: the discovery interview, the full stack, an architecture diagram, and copy-paste prompts for every build step - import, scoring engine with editable weights, ranked view, the override-with-reason gate, the approval flow, and the CSV export. It closes with a "how to know it works" checklist and a no-API fallback so you can ship today even with zero integration to your current tools.
The governance it includes (this is the point)
This isn't a toy spreadsheet replacement - it has the controls a PMO or prioritization committee actually needs:
- Login so only your team can open it.
- Row-level security so each organization only ever sees its own requests.
- A complete audit trail - who imported, who changed an input, who overrode a rank, what reason they gave, and when.
- A human-in-the-loop approval gate - the computed ranking is a draft until the prioritization owner approves it. Overrides require a written reason, every time.
- Duplicate guards - dedupe on request id and on (requester + normalized title), so the same request can't sneak into the list twice and skew the ranking.
Who it's for
Product owners, PMO leads, prioritization committees, and team leads - anyone who owns a backlog or an intake queue and is tired of priorities being set by escalation volume instead of impact. If you can describe your scoring model in a sentence, you can build this.
You've got this - paste the first prompt and let the agent interview you.