Automations/Everyone
When the Same Document Arrives Twice
Two people forward the same invoice. One replies to the thread with it attached, the other uploads it. Nothing is wrong with any of them and you now have three.
The short answer
Duplicates in a filing system are a collision problem, not a search problem: the moment that matters is when a file is about to land somewhere a file already is. Every Move and Save a Copy step carries a policy for exactly that moment, and it is a choice rather than a default. Leave it where it is, and the workflow carries on without filing a second copy. Set it aside for review, and it goes to a Possible Duplicates folder for a person to glance at. Move it anyway, and it is filed with a name that does not clash. Triggers can also skip duplicate files before a workflow ever sees them.
Steps this uses
Before and after
What actually changes
As they arrive
- invoice_8823.pdf (emailed by AP)
- invoice_8823.pdf (uploaded by the controller)
- Scan_0043.pdf
- Scan_0043.pdf (a different document)
After the workflow
- Vendors/Northwind/2026-08-02-invoice-northwind-4821.pdf
- Possible Duplicates/2026-08-02-invoice-northwind-4821.pdf
- Vendors/Acme/2026-08-04-invoice-acme-1180.pdf
- Vendors/Dataflow/2026-08-04-receipt-dataflow.pdf
The genuinely duplicated invoice is set aside for a glance; the two identically named scans turn out to be different documents and are filed apart.
Setting it up
Described once, then it runs
This is the sentence. Send it to the builder and the steps below appear on a canvas, wired and named, for you to change before anything runs.
When a file arrives that would land on top of one already filed, set it aside in a Possible Duplicates folder for review instead of saving a second copy.
Decide at the step that files
The policy lives on the Move step and the Save a Copy step, because those are the moments a collision can happen. It is set per workflow, so an invoice flow can be strict and a photo flow can keep everything, which is usually what you want.
Pick the behaviour deliberately
Leave it where it is, and nothing is filed twice. Set it aside for review, and it lands in Possible Duplicates instead. Move it anyway, and it is filed under a name that does not clash. The last one is the default and is the right answer more often than it looks, because two files with the same name are not always the same document.
Name from contents, so collisions mean something
This only works if names are derived rather than inherited. Two scans called Scan_0043.pdf collide and are unrelated; two files both named 2026-08-02-invoice-northwind-4821.pdf almost certainly are the same invoice. Renaming from the document is what turns a name collision into useful evidence.
Or stop them at the source
Connected sources can skip duplicate files before a workflow runs at all. That is the cheapest option and the least visible one — if you would rather see what was skipped, leave it off and let the Possible Duplicates folder collect them.
The same name is not the same document
Every scanner in the building produces Scan_0043.pdf eventually, and two files sharing that name have nothing else in common. This is why a filing system that deduplicates on filename quietly loses documents, and why reading the contents first matters: once a name is derived from the document, a collision is a real signal rather than a coincidence of counters.
Why review beats silent skipping
Skipping is tidier and it is the option that hides mistakes. A near-duplicate — the revised contract, the corrected invoice, the same statement with a page added — looks like a duplicate at the moment it lands and is not one. Setting them aside costs a folder somebody glances at occasionally and preserves the difference between a document you already have and a document you nearly have.
What it does not do
It decides what happens when an arriving file collides with one already filed. It does not scan your drive for duplicates that are already there, compare file contents to find copies under different names, or merge two versions of a document. This is about the moment of filing, which is the moment the problem is cheapest to handle.
Then build
Variations worth building next
When a file lands in my Possible Duplicates folder, send me a Slack message with its name so I can check it.
Makes sure set-aside files get looked at instead of quietly piling up.
Every month, check my Possible Duplicates folder for files older than 60 days, ask me to approve, and move what I approve to the trash.
Clears reviewed duplicates on a schedule, with a person signing off first.
When a scan lands in my Dropbox scans folder, rename it from its contents and file it by document type. If a file with that name is already there, file it anyway under a name that does not clash.
For folders where losing a revised version would cost more than keeping a copy.
Limits
What to know before you build it
It runs on new files, not your backlog
This workflow fires when a file arrives. It will not go back over the documents already sitting in the folder, so switching it on leaves the existing pile exactly where it is. Working through what is already there is a scheduled run instead, which is a different workflow.
Organize the files you already haveIt needs a connected account
This workflow reads files out of an email inbox, so it cannot run until that account is connected. Connecting one is on the Max plan, or on any team workspace. There is no version of this particular workflow that runs on the free plan, because the files it organizes never arrive in The Drive AI on their own.
FAQ
Common questions
What happens when the same file is uploaded twice?
Whatever you chose for that workflow. It can be left where it is and not filed again, set aside in a Possible Duplicates folder for review, or filed anyway under a name that does not clash. The choice sits on the step that does the filing.
Can it find duplicates already sitting in my drive?
No. This handles collisions at the moment a file is filed, not a sweep of what is already there. Naming every document from its contents does make existing duplicates visible in a sorted folder, since genuine copies end up with identical names side by side.
Two different scans have the same filename. Will one be lost?
No, and this is the reason to name files from their contents. The workflow reads each document and names it from what is inside, so two files that arrived as Scan_0043.pdf become two different names and are filed separately. Deduplicating on the name a scanner produced is what loses documents.
Should I skip duplicates at the source or review them?
Skipping at the source is cheaper and hides what it skipped. Reviewing costs a folder somebody checks occasionally and catches the near-duplicate — a revised contract or a corrected invoice — that looks identical at the moment it lands and is not.
Decide once what a second copy means.
Every workflow knows what to do when a document it has seen before turns up again.
5 GB free · No credit card required