A notary in Bern keeps forty years of deeds in a basement archive. A doctor in Lausanne has a wall of patient files going back a decade. A fiduciary in Zurich stores client folders in boxes that nobody has opened in years. The paper is safe — it is also, for all practical purposes, unfindable.
When a client calls to ask about a contract from 2019, the answer usually begins with a hunt through folders, drawers and boxes. That hunt can take ten minutes or a whole afternoon, and it almost always interrupts work that actually earns money. This article describes a calm routine for turning those piles into a searchable digital archive — the AI does the naming, sorting and indexing, while you stay in control of every decision.
You do not need a big document-management system or an expensive scanning project. You need a simple habit, a phone or a scanner, and an AI model to handle the tedious part.
Why paper piles cost more than you think
Paper is not the problem. Paper without an index is. The cost shows up in small, repeated ways:
- Lost time: every retrieval is a hunt, and the hunt grows longer as the archive grows.
- Repeated work: you redo a document because the earlier version is buried somewhere.
- Risk: boxes age, folders get mislaid, and one water leak can take a decade of records with it.
A calm three-step routine
Step 1 — Scan once, with the right settings
You do not need to digitise everything at once. Start with what you actually look for: current client files, contracts, invoices, and anything still inside its legal retention period. Use a scanner with a document feeder if you have one; a good phone scanning app works just as well for small batches. Make sure the scan produces searchable text (OCR), not just an image — that text is what the AI reads later.
Step 2 — Let the AI name, sort and tag
This is where the AI earns its keep. Instead of naming files by hand, hand the scanned pages to the AI and ask it to name them, sort them and assign a few tags. One consistent naming rule — client name, document type, date — is worth more than any elaborate folder structure.
I will paste several scanned documents below. For each one, give me a short, clear file name in the form "client — document type — date", assign up to three tags, and tell me which folder it belongs in. Keep the names calm and consistent, in the language of the document.
Step 3 — Store once, find forever
Put the files in one place — a local drive, a shared folder, or a cloud with a clear retention policy — and let the AI keep a simple index: a list of file names, tags and a one-line summary of each document. From then on, finding a document is a search instead of a hunt.
Here is my document index. When I ask "where is the contract with X from 2021?", answer with the file name and folder, and quote the one-line summary so I can confirm it is the right document. If you are not sure, say so.
What this looks like for your profession
- Notaries and law firms: deeds, contracts and certificates — decades of them, all with long retention periods.
- Fiduciaries: client folders, annual statements, payroll records, correspondence.
- Doctors: patient files, referral letters, lab results — sensitive and strictly regulated.
- Small offices: invoices, offers, insurance policies, guarantees.
The routine is the same everywhere. Only the documents change.
A note on data
Scanned documents are often confidential — client data, health data, financial records. Keep the usual rules in mind: run OCR on your own hardware or a service with a clear no-training, no-retention agreement, store the files where your professional obligations are met, and if you are unsure, index with anonymised names first, then add the real names back at the end.
The outcome
One calm weekend afternoon to set it up, ten minutes a week to keep it fed, and the archive stops being a source of dread. When a client calls, you find the document in seconds instead of promising to "get back to them". You stop redoing work, you stop worrying about the basement, and you free up the headspace for the advisory work that actually earns your rate. Start with one box — the one that annoys you most.