How to Organize Files You Already Have, Automatically

TL;DR: Almost every filing automation only acts on files that arrive after you switch it on, which leaves the eleven years already in the folder untouched. A scheduled workflow set to go through every file on every run reads what is already there, renames it, and moves it, in place. That is the difference between a tidy future and a tidy drive.

You finally set up automatic filing. New invoices land in the right vendor folder, new contracts get named properly, and it works. Then you open the folder that started the whole project and it is exactly as it was: 14,000 files called scan_0043.pdf, Document (7).docx, and FINAL_v2_actually final.xlsx, going back to 2015.

This is the most common disappointment in file automation, and it is structural rather than a bug. A trigger fires on an event. A file that arrived in 2019 is not an event, so nothing fires, and the backlog sits there being the actual problem you were trying to solve.

Organizing existing files means applying the same reading, naming, and filing logic to documents that are already stored, rather than only to the next one that arrives.

Why is the backlog the hard half?

New files are easy because something happens when they appear. Old files are hard for three reasons worth naming before reaching for any tool.

There is no event to hang the work on. Nothing happens to a file sitting in a folder, so the automation needs to go looking rather than wait.

The backlog is inconsistent in a way new files are not. A folder built up over a decade holds four naming conventions, three people's ideas of a folder tree, scans with no text layer, duplicates that differ by one page, and files whose only clue to their contents is inside them.

Getting it wrong at volume is worse. A mistake on one incoming invoice is a small annoyance. The same mistake applied to 14,000 files at once is a bad afternoon and possibly a restore.

That third point should shape how you approach it. Anything that touches a backlog needs to be reviewable, reversible, and run on a slice before it runs on everything.

What to do before you automate anything

Some of this work is not automatable, and doing it first makes the automation far more accurate. This part applies whatever tool you use.

Decide the destination structure before you sort into it. A workflow files into folders that already exist rather than inventing a tree. If you do not know whether contracts live under Clients/Northwind/Contracts or Contracts/Northwind, no tool can decide that for you, and one that picks for you will pick inconsistently.

Agree one naming convention and write it down. Date first if you browse chronologically, client first if you browse by relationship. Whichever you choose, it should be a sentence you can hand to a person or a workflow and get the same result.

Deal with duplicates separately, and first. Deduplication is its own job and it changes how much there is to file. Duplicate file detection covers that pass.

Decide what should not be organized but archived. A lot of a ten-year backlog does not need a home, it needs to be out of the way. Age is the cheapest way to separate the two.

How do you organize files that are already filed?

A workflow in The Drive AI starts from a sentence, and the steps appear on a canvas. For a backlog, the source is On a Schedule, which runs on a repeating schedule, hourly, daily, weekly, or monthly, over a folder in The Drive AI or a connected storage account.

The setting that matters is on the source and is called "Each run goes through". It has two choices:

• New and changed files only. Each file is handled once, then left alone until it changes. This is the right setting for ongoing upkeep.
• Every file, every run. The run goes through everything in the folder, including files that were already filed. This is the setting that touches the backlog, and it is also what you need for any condition about age, such as older than 90 days, because whether a file qualifies changes over time.

From there the steps do the work:

• Check File Details (free) branches on type, age, location, name, or size, without opening anything. Each rule you add gets its own path on the canvas, plus a fall-through for everything else. On a backlog this is the cheapest and most valuable step, because it separates the images from the spreadsheets and the 2016 files from the current ones before any AI is spent.
• Ask AI About the File (ai) reads the file and sends it down a named path. The paths are the answers you allow, so one step can sort into invoice, contract, report, and other.
• Read Details from File (ai) pulls out named values such as client, vendor, or date, which later steps can use in names and folder paths.
• Rename (ai) and Move (ai) apply the convention you described. Move has an "If the same file is already there" setting with three choices: move it anyway (renamed if the name is taken), leave it where it is, or set it aside for review in a "Possible Duplicates" folder. On a backlog, the third option is usually the right one.
• Ask a Person to Approve (human) pauses until someone decides, and can be set to Only when unsure, so confident filing into existing folders goes ahead and is logged, while new folders, unsure matches, and anything destructive still ask.
• Move to Trash (free) soft-deletes, recoverably, at the end of a branch.

Files in a connected Google Drive, OneDrive, SharePoint, or Dropbox folder are renamed and moved in place, inside that storage. Nothing is imported unless you add a Save a Copy step.

Example workflows for a backlog

Each of these is the sentence you would type.

The first pass, cheap and non-destructive, separating rather than filing:

Every night, go through every file in Shared/Unsorted, including ones already there. If it is older than three years, move it to Archive/Pre-2024. If it is an image, move it to Media. Leave everything else where it is.

The real sort, with a person checking the uncertain ones:

Every night, go through every file in Shared/Unsorted, including ones already there. Work out whether it is an invoice, a contract, or a report. File invoices under Vendors by vendor and year, contracts under Clients by client, and reports under Reports by year. Rename each one with the client or vendor, the document type, and the date. Ask me only when you are unsure.

A vendor folder brought up to one convention:

Every Sunday, go through every file in Finance/Vendors, including ones already filed. Read the vendor, invoice number, and date from each one and rename it "<vendor> - Invoice <number> - <date>". Leave the folders as they are.

Retention, where the destructive branch always asks:

Every month, go through every file in Projects/Completed. If it is older than seven years, ask Dana to approve before moving it to trash.

Notice that the last one asks a person even though the condition is exact. Age is a free check and the workflow could act on it alone. Deletion is the one place where the cost of being right quickly is lower than the cost of being wrong quietly.

Which surface reorganizes the files you already have?

This is the question people get wrong, so it is worth being exact. There are three ways, and they end in different places.

The scheduled workflow is the only one of the three that covers SharePoint, and the only one that keeps running after the first cleanup. The extension is the fastest way to tidy a drive you are already staring at. For the import route, importing from Google Drive, Dropbox, and OneDrive covers what comes across.

What it will not do

• AI steps can be wrong, and volume multiplies it. Every step that opens a file carries the ai badge. Run on one subfolder first, read the activity feed, then widen.
• Every file counts. A run set to every file, every run processes each file each time, and every file a workflow processes counts once against a monthly allowance. Use Check File Details to narrow the set before an AI step sees it.
• It files into folders that exist. Move uses the structure you have rather than inventing one, which is why deciding the structure first matters.
• Approvals expire. An approval waits 24 hours, 3 days, or 1 week, with 3 days the default. Then the file takes the Expired exit, and if nothing is connected there, nothing happens to the file and the expiry is recorded.
• Nothing leaves the drive. Files in connected storage are renamed and moved in place. Nothing is imported unless a Save a Copy step says so, and nothing outside the folder you named is touched.
• Connected accounts need a paid plan. Workflows on files uploaded to The Drive AI work on every plan. Connecting a cloud account, mailbox, or chat workspace needs the Max plan on a personal workspace, or any Team plan.

Setting it up on a backlog

1. Pick one subfolder, not the whole drive. Choose a folder with a few hundred files that you understand well enough to check the results.
2. Create the destination folders first, so the workflow has somewhere to file into.
3. Build the cheap pass first. Use Check File Details on type and age to separate the obvious before spending an AI step on anything.
4. Add the sort, with Ask a Person to Approve set to Only when unsure. Choose Careful, Balanced, or Relaxed depending on how much you want it to act on its own.
5. Run it once and read the activity feed. Every run can be replayed on the canvas exactly as it happened and retried from the step that failed.
6. Widen the folder, then switch the schedule to new and changed files only once the backlog is done, so it becomes upkeep rather than a sweep.

That last step is the one people forget. Once the backlog is handled, "every file, every run" is doing work that does not need doing.

For the full set of sources, decisions, and actions, see the document workflow steps guide, and for the source-by-source setup, the guides for Google Drive, OneDrive, SharePoint, and Dropbox.

Frequently Asked Questions

Can AI organize files I already have, or only new ones?

Both, but they need different setups. A workflow triggered by a file arriving only ever sees new files. To handle a backlog, use a scheduled source and set it to go through every file on every run, which includes documents that were already filed. The same reading, renaming, and filing steps then apply to what is already there.

Will it move files out of my Google Drive or SharePoint?

No. A workflow watching a folder in a connected drive renames and moves files in place, inside that storage. Nothing is imported unless the workflow includes a Save a Copy step, and then only into the place you name. Nothing outside the folder you chose is touched, and your existing permissions still apply.

How do I stop it making a mess of 14,000 files at once?

Run it on one subfolder first and read the activity feed before widening. Use the free Check File Details step to separate obvious cases before any AI step sees them, set the approval step to ask when unsure, and choose the duplicate setting that sets files aside for review. Any run can be replayed and retried from the step that failed.

What happens to duplicates during a cleanup?

The Move step has a setting for when the same file is already in the destination, with three choices: move it anyway and rename if the name is taken, leave it where it is, or set it aside in a "Possible Duplicates" folder for review. Sources also have a setting to skip duplicate files. For a real deduplication pass, handle that before the filing pass.

Can a scheduled workflow delete old files?

It can move them to trash, which is a soft delete and recoverable. Because deletion is the one action where a wrong decision is expensive, put an approval step in front of that branch so a person confirms before anything is trashed. The approved, rejected, and expired outcomes each get their own path on the canvas.

How often should the schedule run?

For a backlog sweep, nightly is usually enough, and you can narrow it to one folder at a time. For upkeep after the backlog is done, switch the source to new and changed files only so each file is handled once and then left alone. Conditions about age, such as older than 90 days, need every file, every run, because whether a file qualifies changes over time.

Start with one folder

Pick the folder that made you search for this in the first place, decide where its files ought to live, and let one scheduled run show you what it would do. Build your first workflow, or see how Workflows handle files wherever they already are.