How to Organize Household Documents With One AI Naming Template

Somewhere on your drive there is a folder called Scans, or To File, or Documents if you were feeling optimistic. Inside it: an electricity bill, two lab results, a credit card statement, the car insurance renewal, a receipt for a washing machine, and the mortgage annual statement. All of them named some variation of nothing at all.

Scans/
├── scan_0042.pdf
├── scan_0043.pdf
├── IMG_2231.jpg
├── statement.pdf
├── statement (1).pdf
└── ...394 more

Search the folder for "insurance" and you get nothing, because the word "insurance" is inside the PDF, not in its name. So you open files one at a time until you find the right one. That is the actual failure mode of a home filing system, and it has almost nothing to do with your folder structure.

Read enough guides on organizing household documents and you will notice they all end at the same place: a list of categories. Financial, Medical, Home, Auto, Insurance. Useful, and not the problem. The problem is that sorting 400 unnamed PDFs into those nine folders is 400 individual decisions, each one requiring you to open the file to find out what it is.

This guide covers the layer those guides skip: getting the date, the category, the sender and the amount into the filename itself, automatically, in one pass over a mixed folder. The setup takes about twenty minutes and then it keeps working.

Why Filenames Do the Work Folders Can't

A folder tells you where a document lives. A filename tells you what the document is, and it is the only part of that pair that survives contact with the real world.

Export your Drive to a zip and the folder context thins out. Email one statement to your accountant and the folder is gone entirely — they receive statement.pdf. Restore a backup onto a new machine, or move from Dropbox to iCloud, and the hierarchy you carefully built is the first casualty. The filename is the one piece of metadata that travels with the file everywhere, forever.

Here is the same set of documents with the naming layer in place:

Every name carries the date, what kind of document it is, who sent it, and the amount. Notice the second and third files: no person's name. That is deliberate, and it is the part of this recipe that takes the most explaining, so we will come back to it.

One Template, Not Nine

The instinct when setting this up is to build a template per document type: one for medical records, one for bank statements, one for utility bills. It feels tidier. It is a trap.

Household paperwork arrives mixed. You empty the letter tray, run everything through the scanner in one go, and end up with a folder containing eleven unrelated document types. If you have nine templates, you now have to sort that folder by hand and run nine separate batches — and sorting by hand means opening each file to see what it is. You have reinvented the original problem with extra steps.

One template with a category field pushes that decision to the AI, where it belongs. It reads the document, works out that this one is a lab result and that one is an electricity bill, and names both correctly in the same pass.

Build the Template in Five Steps

Step 1 — Create the template and pick a separator

Open Templates → New template and give it a name you will recognise in a dropdown six months from now — Household Documents does the job. Then pick your separator: dash, underscore or space.

This looks like a cosmetic choice and is not. A dash separator conflicts with hyphenated names, and you will meet at least one: Smith-Jones Law, Anne-Marie, Blue-Sky Energy. Step 5 fixes that, but deciding now is much cheaper than deciding after you have renamed 400 files. If you want to skip the extra rule entirely, choose a space or an underscore.

Step 2 — Add the six fields

Fields are the details FilesDesk looks for in each document. The description column is not documentation — it is the instruction the AI actually follows, so paste it in as written. Leave Required switched off on all six, so a field the AI genuinely cannot fill comes back blank instead of blocking the rename.

Swap James and Sarah for the names in your household, and edit the category list to match the folders you already use. Everything else can stay as it is.

Step 3 — Set the filename pattern

The pattern decides the order. Type it exactly:

Date first is not a stylistic preference. YYYY-MM-DD is the one date format where alphabetical order and chronological order are the same order, which means every folder sorts itself correctly in Finder, in File Explorer, in Drive, in a zip listing, and in any tool you migrate to in ten years. It is the whole reason ISO 8601 dates are worth the mild ugliness.

Step 4 — Preview one file before you touch the folder

Drop a single document into the Manual tab, select your new template, and press Preview. Nothing is renamed until you click Apply.

Check the suggested name against what you expected. If the category is wrong or the amount picked up the due total instead of the paid total, adjust that field's description and preview the same file again. This loop takes seconds, and it is dramatically faster than batching 400 files, discovering a systematic error, and undoing the lot. Get one file right, then two or three that are deliberately different from each other, then point it at the folder.

Step 5 — Add the separator clean-up rules

Skip this step if you chose a space or an underscore in Step 1. If you chose a dash, you need it.

The problem: Smith-Jones Law contains the same character the template uses to divide one field from the next. Composed naively, the filename becomes ambiguous — nothing reading it afterwards can tell which dash is structural and which is part of a name. In the template's Transformations section, add:

• Vendor — Replace - with (space)
• Person — Replace - with (space)

Smith-Jones Law now becomes Smith Jones Law inside the filename, while the dash separators between fields stay exactly where they are. Scope the rules to those two fields rather than applying them globally, or you will strip the dashes out of your dates as well.

Conditional Fields Without if/else

Now the interesting part, and the question that produced this recipe in the first place:

This is a completely reasonable thing to want. A lab result belongs to a specific person; an electricity bill belongs to the house. Stamping Shared onto every utility bill is noise, and stamping nothing onto a medical record loses information you need.

In most renaming tools this is where you would go looking for conditional syntax — some {if category=Health} construct — and not find it. FilesDesk does not have if/else in patterns, and does not need it, because of two behaviours that combine neatly.

1. The description carries the condition

Field descriptions are instructions to the AI, so the condition goes there in plain English:

That is the entire mechanism. The AI has already decided the category for this document, so it can apply the rule and return either a name or nothing.

2. Empty fields collapse, and so do their separators

After the name is composed, the sanitizer removes runs of adjacent separators — and strips any that end up at the start or end:

So a blank field does not leave a hole. It leaves nothing at all, and the fields either side close up. No doubled dashes, no filename that starts or ends with a separator.

The same pattern therefore produces both of these, from the same template, in the same batch:

Once you see it, the same trick covers most of what you would otherwise want conditional logic for. An amount that only appears on documents that have one. An invoice number that only appears on invoices. A policy number only on insurance papers. In each case the rule is: describe the condition, return blank when it is not met, let the separators close.

Writing Field Descriptions the AI Actually Follows

The quality of your filenames is almost entirely the quality of your descriptions. Four patterns that consistently work:

A useful test: read the description aloud and ask whether a competent new assistant, handed the document and nothing else, could follow it without asking a follow-up question. If they would need to ask, the AI needs the answer written down.

Routing Files Into Folders Automatically

Naming and filing are separate jobs, and naming is the one worth doing first — a well-named file in the wrong folder is still findable, while a perfectly filed scan_0042.pdf is not.

Once the names are right, an Organize Flow can move each file by matching the subcategory value against a folder name. The match is exact, so the two lists have to agree:

Documents/
├── Finance/
│   ├── Banking/          ← subcategory: Banking
│   └── Credit Card/       ← subcategory: Credit Card
├── Health/
│   ├── Medical/
│   └── Insurance/
└── Home/
    └── Bills/

Adding a subcategory later means editing two places, and they drift apart the moment you forget one:

1. The subcategory field description — append the new name to the comma-separated list.
2. The folder tree — create the matching folder under the right parent.

New top-level categories work identically: extend the category description and add the folder. If you want the whole pipeline running unattended on your scanner's output folder, watch folders handles the trigger.

Troubleshooting

When a name comes back the wrong shape, open the History tab first. Every row shows the exact field values the AI returned, which turns guesswork into a two-second diagnosis — you can see immediately whether the category was wrong or the pattern was.

What Not to Scan

Worth saying because the enthusiasm of a working system tends to run ahead of good sense. Scan freely, but keep the physical original of anything whose value depends on being an original:

For everyday paperwork — bills, statements, receipts, correspondence — a well-named scan is genuinely better than the paper, because you can find it. Retention requirements for tax and financial records vary by country, so check the rules where you live before shredding anything you might have to produce later.

Frequently Asked Questions

How should I name household documents?

Date first in YYYY-MM-DD, then category, then who sent it, then the amount if there is one. That order sorts chronologically by default and stays readable when the folder context is gone. Folders tell you where a document lives; the filename is the part that still makes sense after an export, an email or a backup restore.

Should I use one template for all household documents or one per category?

One. Household scanning is mixed by nature, so a template per category forces you to hand-sort the pile before you can rename it — which means opening every file to see what it is. A single template with a category field hands that decision to the AI instead.

Can I add if/else logic directly to a filename pattern?

Not as pattern syntax, and you rarely need to. Write the condition into the field's description — "only when the category is Health, Legal or Estate, otherwise leave blank" — and the empty result collapses cleanly out of the name. That covers most conditional cases without a scripting layer.

What happens when the AI leaves a field blank?

The separators either side close up. ...Alexander Electric--$100.00 sanitizes to ...Alexander Electric-$100.00, and no filename ever starts or ends with a separator. That collapse is what makes conditional fields work at all.

What if a company or person name contains a hyphen?

It collides with a dash separator. Either choose a space or underscore as your separator, in which case hyphenated names pass through untouched, or keep the dash and add a Replace transformation on the vendor and person fields turning - into a space.

Does a Replace transformation affect anything outside this template?

No. Transformations are scoped to the template that defines them, so other templates and watch folders are untouched. Within a template, though, do scope a dash-replacement to specific fields rather than globally — a global rule would eat the dashes in your dates.

What if the AI can't tell whether a document is Medical or Legal?

It picks the closest match from your list, falling back to Reference if nothing fits. The History tab shows which it chose, and you can override before applying. If it happens repeatedly, the fix is a more specific category description rather than more categories.

Do I have to use a dash as the separator?

No — space, underscore and dash all work. Underscore is the safest default: it never collides with hyphenated names and it survives URLs and command lines, which spaces do not.

The Bottom Line

Organizing household documents is usually framed as a folder problem, which is why so many attempts stall. You spend a Saturday designing a nine-category hierarchy, file forty documents into it by hand, and the other 360 stay in Scans forever. The structure was never the bottleneck.

The bottleneck is that opening a file to find out what it is costs about fifteen seconds, and 400 files is an afternoon. Automate that — the reading and the naming — and the filing becomes trivial, because a file called 2026-09-10-Finance-Credit Card-Chase Bank-$412.35.pdf is already sorted, already searchable, and already tells you everything before you open it.

Three things to get right: one template instead of nine, date-first names in YYYY-MM-DD, and conditions in the field descriptions rather than in pattern syntax that does not exist. Twenty minutes of setup, and the letter tray stops being a project.

Further reading: file naming conventions for the rules behind the pattern, folder structure best practices for the tree it files into, renaming PDFs by content for the scanning side, and organizing invoice PDFs if your pile is mostly billing.