Returnberry reads every return request that lands in your inbox or helpdesk, catches the duplicates, gets the missing details from the customer and creates the return in your order system — then brings the outcome back and closes the ticket.
TRK-88213047 posted to the customerWorks with the order system you already run — by connector, API, or the import file it already accepts.
A wholesale buyer emails their account manager; a retail partner sends a spreadsheet; a marketplace opens a ticket; a customer replies to a shipping confirmation. None of it touches your returns portal — so each one is read by a person and re-keyed into your order system, field by field.
Then, days later, someone goes back to the ticket to tell the customer what happened. Returnberry automates both halves of that reverse-logistics work.
Someone reads the email, opens the order system and retypes it. The duplicate turns up next week — or never.
The return is approved, labeled and received. Nobody tells the ticket. The customer writes in again: any update?
The order is created from the request. The status comes back from your system. The ticket is updated and resolved — once, with an audit trail.
If it has an API, Returnberry connects to it. If it only takes a file, Returnberry speaks the file. Your helpdesk stays your helpdesk, and your order system doesn’t have to change.
Connects to your helpdesk and order system — by API, webhook, or the file it already accepts.
Mailboxes, helpdesk queues, web forms and the spreadsheets partners attach — all into one queue. It never reaches behind the day you switch it on, so your backlog stays history.
Your templates and forms are read without AI, at zero cost. Freeform email goes to AI, with a rules-based fallback. No return number on the request? It stays empty and goes to a person.
Re-sends, chase-ups and follow-ups are linked to the original on identifiers before names. A follow-up fills a blank; it never overwrites a value you already hold.
A one-time link asks the customer only what’s missing — dimensions, packaging, a collection address. The logic is deterministic; AI only rephrases, and answers questions from your own knowledge base.
By native connector, through your OMS’s API, or as the import file it already accepts — mapped to your fields. Every batch is recorded, so you can always see exactly what was sent.
Status comes back by webhook, API or file, matched on the references Returnberry sent. The update lands on the original ticket and the ticket resolves — exactly once.
Connected systems send status back automatically. When it comes back as a file, it’s never a blind upload: the preview writes nothing, maps columns from the file’s own headers and gives every row a verdict. Only rows a person ticks are applied — and the whole import can be undone.
Nothing is guessed, nothing is overwritten, and the whole import can be undone.
Not rules someone has to remember. Properties of the system.
No return number on the request? It stays empty and the order goes to a person. Copying the order number across because it looks close enough is exactly how the wrong return gets approved.
Two independent dedupe layers, plus a fingerprint re-check at the moment an order is sent. Status alone is never trusted — a manual override and a stale flag are how duplicates slip through.
The ticket update is built only from what’s recorded on the order. If your system hasn’t reported it, the customer isn’t told it.
Every returns operation is different — different order systems, different return locations, different handling rules, a different definition of “complete”.
Returnberry is configured, not customized. Your workspace holds your field mapping, your routing, your rules and your knowledge base — and you change them yourself, without a release.
Returnberry connects to your live helpdesk and your live order system, so it’s read-only until you say otherwise. Creating orders, emailing customers, replying on tickets, posting status and resolving tickets are separate switches with separate consequences.
Turn one on. Watch a batch. Then widen.
Anything with an API, and anything that imports a file. Native connectors cover our initial launch partners; the rest connect through the API, webhooks, or your system’s own import format.
Any mailbox, by forwarding address or direct connection, plus native helpdesk connectors for our initial launch partners. Adding a channel never changes how requests are processed.
No. Returnberry handles the requests that never reach it. If you want one place to work every return, portal returns can flow into the same queue.
Structured requests are read without AI at all. Everything the AI reads has a rules-based fallback, a confidence score, a permanent record of how it was read, and a person reviewing what it wasn’t sure of.
Not until you switch it on, and there’s a test-recipient mode first. A whole workspace can run with every outbound write forced off.
Every request, order, file and audit row is scoped to your workspace, and your connection credentials are yours alone and encrypted.
A walkthrough takes 20 minutes and uses your own return emails.