Route every return to its highest-value recovery path.
ReturnRoute classifies return reasons, suggests condition grades from inspection notes and photos, and recommends restock, repair, resale or recycling under your business rules. A person approves every decision that matters.
Labour, parts, channel fees and logistics, compared side by side
Rule trace and rationale attached to each recommendation
No refund, disposal or stock adjustment happens without a person
Watch a damaged product get its route
Pick a fictional return and run the AI inspection. The engine won’t act on the AI grade alone: confirm it as the inspector to see the recommended route, the value of every option and the rules behind it.
- Intake & ID
- AI inspection
- Human verifies
- Route & explain
“Box was crushed when it arrived and the kettle has a dent in the base. The lid hinge is cracked so it won't close.”
Outer carton crushed. 12mm dent on base rim, heats normally, no leak. Lid hinge cracked — lid won't latch. All accessories present.
- dented lower rim
- cracked hinge on lid assembly
Five AI capabilities, one accountable workflow
Free-text reasons become structured categories, so recurring product and fulfilment issues surface on their own.
Inspector notes and product photos produce a suggested condition grade with cited evidence — for a person to verify.
Your configurable rules decide which routes are permitted; the engine ranks them by net value and shows its working.
Every permitted route is costed — labour, parts, channel fees, logistics — so the trade-off is explicit.
Spot damage hotspots, defective SKUs, carrier problems and stage bottlenecks before they become write-offs.
Amazon S3 for images and documents, Amazon Bedrock for classification and summaries, PostgreSQL for returns and inventory, and AWS Lambda for processing workflows.
Five recovery paths
Sealed, complete items back to A-stock at full price.
Repairable damage to the bench — only when the maths works.
Open-box and B-stock to secondary channels.
Certified recycling, always supervisor-approved.
Uncertain or hazardous items quarantined first.
AI that advises. People who decide.
Image classification is useful, and it can be wrong. ReturnRoute treats every AI output as a suggestion: it is labelled, carries a confidence score and evidence, and is recorded in the audit trail next to the person who confirmed or overrode it.
- AI condition grades must be verified by an inspector before any route can be approved
- Refund decisions are recorded by staff and never executed automatically
- Disposal and recycling, and any item over your high-value threshold, need supervisor approval
- Lithium-battery products need a recorded safety check before leaving quarantine
- Inventory adjustments are only sent to your WMS after a human approval
- Overrides need a reason, and every step is audit-logged
See it on your own returns mix
We’ll model recovery value on a sample of your SKUs and business rules.