RMA Request Portal & Eligibility Checker
A return-request intake that matches each request to its order, checks it against your real return policy (window, condition, exclusions, proof of purchase), and proposes approve / deny / needs-info with a reason - so eligibility is consistent, not mood-based, and no RMA goes out without a human approving it.
A logged-in portal where a return request gets matched to its order, scored against your policy, and given a suggested decision and reason - then an agent reviews, approves, and an RMA number plus return instructions are emailed automatically and recorded with a full audit trail.
Before you start
- A free Vercel account
- A free Supabase account
- A free Resend account (and a sender address you can verify)
- Your written return policy (window, condition rules, final-sale / hygiene / category exclusions)
- An order reference export (Google Sheet or CSV with order ID, SKU, purchase date)
The problem this kills
Returns decisions drift. One agent waves through a return that's two days past the window; another denies the same case the next morning. Final-sale and hygiene items slip through. Someone issues an RMA number for an order that doesn't exist, or for a line that already has an open RMA. Customers get inconsistent answers, your refund leakage creeps up, and nobody can later explain why a particular return was approved.
The root cause isn't bad people - it's that the policy lives in a PDF and in everyone's head, applied by hand, under time pressure, with no record of the reasoning.
What you'll build
A small internal web app where:
- A return request comes in (web form or CSV import) with order ID, SKU, purchase date, reason, and optional photos.
- The tool matches it to the real order and checks it against your policy: return window measured from the actual purchase date, condition rules, final-sale / hygiene / category exclusions, and proof-of-purchase requirements.
- It proposes approve / deny / needs-info with a plain-language reason for every request.
- A human agent reviews the suggested decision, can override it, and approves - only then is an RMA number issued and the customer emailed return instructions via Resend.
- Everything is logged: who decided what, when, and why.
What's inside the Implementation Plan
- It starts by interviewing you about your business. Before any code, the plan has the AI agent ask you about your actual return policy, your order data, your SKU and category naming, your volumes, and your messy edge cases - then it tailors the data model, the eligibility rules, and every later step to your answers. You are not getting a generic template; you're getting a tool shaped to how you really process RMAs.
- A clear definition of done so you know exactly when you're finished.
- Step-by-step build instructions, each ending with a ready-to-paste prompt for your AI coding agent.
- A simple architecture diagram so you understand what you're building.
- A verification checklist to prove it works.
- A "No API yet?" fallback: import your orders from a Google Sheet / CSV and export approved RMAs as a CSV in your ERP's exact import columns - so you can ship this today even with no integration.
The governance it includes (this is the point)
- Login so only your team can use the tool.
- Row-level security so each organization only ever sees its own data.
- A complete audit trail - every match, suggestion, override, and approval is recorded with who and when.
- A hard human-in-the-loop gate - the AI drafts the decision; a person reviews and approves; only then is an RMA issued or a refund/replacement set in motion.
- Duplicate guards - one open RMA per order line (dedupe key = order ID + SKU), so the same return can't be processed twice.
Who it's for
Returns and support agents, and the ops people who own the RMA process - anyone who has to apply a return policy consistently, defensibly, and fast, without reinventing the call every time.
You've got this - paste the first prompt and let the agent interview you.