Priority & Severity Scoring Engine
Score every support ticket's priority from impact (how many users, how much money) and urgency (how time-sensitive) — blending your team's hard rules with AI judgment — so priority stops depending on who happens to read the ticket first, with a lead approving or overriding before it's committed.
A ticket comes in, your hard business rules plus AI compute an impact × urgency score and a proposed priority with plain-language rationale, your lead approves or overrides with a logged reason, and the final priority is written with a full audit trail of every factor used.
Before you start
- A source of incoming tickets (CSV, Google Sheet, or a simple intake form)
- Your team's priority matrix / tier rules and your exact priority labels (e.g., P1–P4)
- A customer-tier reference list (which accounts are VIP / enterprise / standard)
- A free Supabase account
- A free Vercel account
- A free Resend account (only if you want emailed alerts on high-priority tickets)
The problem this kills
Right now, a ticket's priority depends on who picks it up. A calm agent marks a real fire as "normal." A stressed one flags a password reset as "urgent." The same incident gets a different priority on Monday than it would on Friday. Your most valuable customer's outage sits behind a chatty low-value request because nobody connected the ticket to the account behind it.
The result is predictable: SLAs get missed on the things that actually matter, your team argues about what's "really" P1, and you can't defend your prioritization to a customer or an exec because there's no record of why anything was ranked the way it was.
This tool makes priority consistent and defensible. It blends your hard business rules (the lines that are never up for debate — "anything mentioning data loss is P1," "a full outage for an enterprise account is P1") with AI judgment on the fuzzy parts (reading the ticket to estimate how many users are affected and how time-sensitive it is). Rules always win over the AI. And a human lead reviews the proposed priority — and the factors behind it — before it's committed.
What you'll build
A small internal web app where your team:
- Loads incoming tickets from a CSV, a Google Sheet, or a simple intake form.
- Lets the engine compute impact and urgency for each ticket — your hard rules first, then AI judgment on what's left — and produce a proposed priority (using your labels) with a clear, plain-language rationale listing every factor used.
- Connects each ticket to your customer-tier reference list, so a VIP or enterprise account automatically lifts impact.
- Shows the support lead a review queue: the proposed priority, the impact × urgency breakdown, which hard rules fired, and the AI's reasoning.
- Lets the lead approve, or override with a logged reason, before the priority is committed.
- Writes the final priority back with a full audit trail of the factors used — ready to push into your help desk or export as a clean CSV.
What's inside the Implementation Plan
The plan is a complete, paste-and-go runbook for an AI coding agent (Claude Code). You don't write code — you paste, answer questions, and approve.
It opens by interviewing you about your business — your exact priority labels, your impact and urgency definitions, your hard non-negotiable rules, how your customer tiers work, your real ticket fields and ID conventions, your typical and peak ticket volumes, and your messy edge cases — so the scoring fits your matrix, not a generic one. The agent reflects a short tailored spec back to you and waits for your thumbs-up before it builds anything.
From there it walks step by step through standing up the database, importing tickets and the customer-tier list, encoding your hard rules so they always beat the AI, wiring up the AI impact/urgency estimate, building the lead review-and-override queue, committing the final priority, the audit log, and the CSV export — each step ending in a ready-to-paste prompt.
The governance it includes (this is the point)
- Login so only your team can open the tool.
- Row-level security so each organization only ever sees its own tickets and rules.
- A complete audit trail — the proposed priority, which hard rules fired, the AI's rationale, who approved or overrode it, the override reason, and when.
- A hard human-in-the-loop approval gate — the engine proposes a priority; the lead reviews and approves (or overrides with a logged reason); only then is the priority committed to the record of truth.
- Hard rules beat AI guesses — your non-negotiable rules (like "data loss = P1") always override the AI's estimate, and the tool shows you exactly when a rule, not the AI, set the priority.
- Duplicate guards — the dedupe key is ticket ID, so the same ticket can't be scored or committed twice.
Who it's for
Support leads and operations managers who own ticket triage and need prioritization to be consistent, repeatable, and defensible — to their team, their customers, and their leadership — instead of depending on who happened to read the ticket first.
You've got this. Open the plan, paste the first prompt, and let the agent interview you.