Lyron
Operations

Automatic Document Naming and Filing

Every incoming file gets a name from the same formula, matching metadata and exactly one place to live. What cannot be matched with certainty is not guessed – it goes to a review folder with suggestions.

Context

The hard part is the name, not the folder

Dropping a file into a folder takes twenty seconds. The time goes into everything that comes first: opening scan_0042.pdf to work out what it even is, digging out the customer number, deciding whether the delivery note belongs to the customer or to the site, and typing a name that will still mean something in three months. Everybody makes those calls afresh every day, and no two people make them the same way.

A good file name needs knowledge the file itself does not carry. The document gives you a company name, not a customer number, and a date – usually three of them: document date, service date, print date. Whether “Reinhardt Stahlbau GmbH” is your customer 10428 or the similarly named holding company is settled by your CRM, not by the paper in front of you. Naming is a matching problem, not a typing problem – and a match nobody can retrace leaves an archive nobody trusts.

Then there is a question you only ask once you walk through the existing files: does one filing scheme actually apply here, or do three sit side by side? One by customer, one by year, one by project, each brought in by somebody who has since left. Nobody notices until somebody else has to look, during a holiday or after a resignation. Anyone who sets out to fix that soon finds the technology is the smaller half: a script will apply any rule you give it, but it cannot settle one.

Use cases

Which documents are worth it

We start with the intake that carries the most volume. Further document types share the same scheme and matching logic.

Most common starting point

Attachments from the shared mailbox

Anything arriving at info@ or accounts@ is matched to a customer by the sender domain and filed together with the message.

Sender domainSubject lineMultiple attachmentsThread link

Scans from the office multifunction device

No sender, no metadata, just text on paper – the matching has to come off the sheet itself. Separator sheets or a change of document number cut a batch into single documents.

Batch scanSeparator sheetText recognition

Supplier invoices and receipts

The scheme is at its strictest here, because your accountant reads along. Date and document number belong in the name, not only in the folder path.

Document dateSupplierDocument number

Quotes, orders and contracts

Documents that exist in several versions. Instead of “final_final”, each carries a date and a status, so the current one is obvious from the name.

VersionStatusCountersignature

Site photos and measurements from a phone

Photographs carry no text to read. Matching runs on the upload folder, the capture date and the project number typed into the comment.

Project numberCapture dateSite diary

Personnel documents

Same logic, separate area, separate permissions. What belongs in a personnel file must never turn up in a customer folder, not even by accident.

Own permissionsPersonnel fileRetention
Example

The scheme, four intakes and the folder behind them

The formula at the top, four files as they actually arrive, and the tree they land in. The right-hand column says what the rules made of them.

Naming scheme

YYYY-MM-DD_Customer no._Document type_Document no..pdf

YYYY-MM-DD
the date on the document, not the day it arrived
Customer no.
from the CRM, never from the company name on the page
Document type
from a fixed list, never free text
Document no.
taken from the document; if there is none, it reads no-number
How the file arrivesWhat it is called afterwardsResult
scan_0042.pdfReception scanner · 21 Jul 2026, 08:14 2026-07-17_10428_delivery-note_LS-88214.pdf filed
inv july.pdfAttachment from accounts@reinhardt-stahlbau.de 2026-07-21_10428_invoice_R-2026-4471.pdf filed
quote final final.docxUploaded from OneDrive, sales 2026-07-20_????_quote_no-number.docxtwo customers share the same name customer not unique
scan_0043.pdfReception scanner · 21 Jul 2026, 08:15 2026-07-17_10428_delivery-note_LS-88214.pdfsame checksum as row one duplicate, not filed
Where the files end up
  • Customers
    • 10428 Reinhardt Stahlbau GmbH
      • 2026
        • Delivery notes
          • 2026-07-17_10428_delivery-note_LS-88214.pdfnew
        • Invoices
          • 2026-07-21_10428_invoice_R-2026-4471.pdfnew
  • _Review
    • 2026-07-20_quote_no-number.docxwaiting to be matched, two suggestions attached

The tree stays shallow: customer, year, document type – three levels, not seven. Every level below that is one more decision when filing and one more place to look when searching.

named and filed review folder, a person decides duplicate, deliberately not filed
Two files are not filed. The quote carries a company name that the example firm holds twice: customer 10428 and customer 10429. The fourth file is the same delivery note, scanned a second time a minute later.

The software settles two of the four rows on its own. The third fails on a customer number that appears nowhere in the document.

One possible scheme, not a standard. Order and separator are yours to set – the date goes first so that files sort themselves chronologically.

How it works

What happens between intake and folder

  • Agree the scheme and test it on real files

    Before any code comes the formula: which fields, in which order, with which separator, with which of the three dates. It is then run against roughly fifty of your own files.

  • Bring every intake onto one path

    Mailbox, scanner folder, manual upload and the folder a phone photographs into all feed a single intake. Every file gets an intake number, and the original stays untouched.

  • Read and match

    Document type from layout and keywords, date and number out of the text – by OCR on a scan, straight from the file on a PDF. Which identifiers decide the customer, and in which order, we settle with you beforehand.

  • Name it, tag it, file it

    The same values sit in the name and in the library columns: the name is for people, the metadata is for search. Renaming happens at the moment of filing, so an interrupted run never leaves a half-named file behind.

  • Doubtful cases, aliases and the log

    Anything ambiguous is parked with two or three candidates to choose from, instead of quietly landing somewhere plausible. That decision can be stored as an alias so the same sender passes through in future, and every filing leaves a line in the log.

Impact

What changes day to day

Today

  • Attachments sit in four mailboxes and one shared folder
  • Names are invented on the spot, the date last or missing
  • Everyone files by the habit they picked up
  • The same invoice exists twice, in two places
  • Searching means asking whoever filed it

With a fixed scheme

  • Every intake runs down the same path and is logged
  • Every name opens with the document date, then the customer
  • The same scheme applies no matter who is filing
  • Duplicates are recognised and not filed again
  • You search by customer, year and document type
Limits

What filing automation does not do

Four points we settle before quoting. On the first one the honest answer may be that we tell you to wait:

  • Until the scheme has actually been decided there is nothing to automate. If sales, accounts and site management each hold a different order to be the correct one, that is a question of authority rather than technology, and no supplier workshop settles it. You need one person on your side who says how it will be done, if necessary against two departments. Without them, postpone the project – otherwise you are paying us to encode the argument instead of the work.
  • Existing files are not renamed. A retrospective pass breaks every link sitting in an email, a quote or a shortcut, so we start at the intake rather than in the archive. That leaves you with two sets of files for a long time, and people search both. Anyone expecting one consistent name everywhere once the project ends will be disappointed.
  • Where a document carries no distinguishing mark, the automation stops: a delivery note with no customer number, no VAT number and a sender whose name you hold twice goes to the review folder. If that is not the exception in your post, you have not bought a filing system, you have bought a second queue for somebody to work through every day. Only your own paperwork shows how large that share is.
  • Order holds only as long as nobody files around it. The moment someone drags files into a folder of their own, or keeps drafts in OneDrive, search is incomplete again – and it goes unnoticed, because everything that is in the system is still found. We can make the right path the convenient one; we cannot make the wrong one impossible, which makes this a management question and one for before the start.
Systems

Fits the storage you already use

SharePointOneDriveMicrosoft 365OutlookGoogle DriveNextcloudDocuWareDATEV Unternehmen onlinen8n
Scope and price

Scope and price

The entry price covers one scheme, one intake path and one filing structure. What moves it, we say before quoting.

from €990 one-off
  • One naming scheme, tested against around fifty real files
  • One intake path, mailbox or scanner folder, plus manual upload
  • Up to five document types with date and document number
  • Matching to a customer or project through one leading identifier
  • Filing into a library or folder structure with metadata columns
  • Review folder for doubtful cases, duplicate detection before filing
  • Documented rules, a handover session and 30 days of support

What increases the price

  • Further intake paths: extra mailboxes, portals, supplier logins
  • More than five document types or a scheme per department
  • A second source system for matching: ERP, CRM or project software
  • Renaming and re-sorting an existing archive
  • Filing into a DMS or audit-proof archive rather than a library

Several intake paths, a second source system for matching and a renamed archive typically land in our Workflow Advanced package from €2,490. The binding fixed price follows the intro call.

All prices excl. VAT · operation and further development optionally via a support package

Included

What you get

  • Production filing flow

    Set up from intake to folder and signed off with your own files

  • The scheme as a document

    Formula, source of every field, the document types and the agreed special cases

  • Review folder and filing log

    The review folder with its stored suggestions, the log filterable by date, customer and rule

  • Handover session for your team

    How the review folder is worked through and how a document type is added

Questions & answers

Frequently asked questions about automatic filing

They stay where they are and keep their names. For the go-live date we propose a clean break: the old area goes read-only, the new one takes everything that arrives – whether you cut that sharply is your call. Anyone who opens an old file anyway can lift it into the new scheme when saving. Aligning the whole archive is a separate job, with a checklist and a redirect for the old links.
Through an order of identifiers we agree with you. What we suggest: a customer number first, if the document prints one, then the VAT number or the IBAN, and the sender domain last. The first identifier that returns exactly one CRM record wins, and two records count as none. We leave the company name out on purpose, because spelling and legal form differ far too often. Anyone not held in the CRM cannot be found, so that post waits in the review folder until the record exists.
In a session of one to two hours, with the people who search every day and with whoever makes the final call. We bring a proposal and adapt it against your real paperwork rather than argue on a blank page. Technology sets two limits: folder path plus file name has a ceiling in both SharePoint and Windows, which is why we use customer numbers rather than spelled-out company names, and characters such as slash and colon are not permitted.
In exactly one place, and you decide once which that is – usually the customer, because a customer outlives a project. The second location holds a reference instead of a copy: in SharePoint a link, or a view filtered on the project number. The difference shows at project closure: the file does not move into the archive, it stays with the customer.
No. We work with the storage you already run – SharePoint, OneDrive, a network drive or Google Drive. You need a document management system once tax-relevant records have to be kept unalterable and traceable; a folder on a drive does not achieve that. If you already have one, we connect to it rather than replace it.
In two ways. The file checksum catches everything that literally arrives twice: the same scan, the same forwarded message. Customer, document type and document number together catch the second case, the invoice that comes once as an attachment and once as a paper scan. A corrected re-issue with the same number but different content counts as a new version rather than a duplicate, and nothing is deleted automatically.
Related solutions

These solutions fit alongside

Email Attachment Automation

The most common intake: attachments pulled out of the shared mailbox and passed on.

AI Document Processing

When the name is not enough and the values inside the document are needed as data.

SharePoint Document Workflows

What happens after filing: approvals, notifications and status changes.

Contract Management

Filed contracts turn into deadlines that announce themselves in time.

What was the file you hunted for yesterday called?

In the free intro call we go through twenty real files from your intake and write down the formula that fits them. You leave knowing how many of them can be matched without guesswork – and which identifier you are missing today.

Book a free intro call
Practical guide

Where automatic document filing creates value in everyday work

The automation detects document type and relevant details, creates a defined file name, applies metadata and files the document in the correct area.

Three concrete operating scenarios to compare with your own process.
01

File email attachments

Attachment and message are assigned to the relevant customer or project folder.

02

Rename scanned files

Unclear scan names are replaced with date, sender and document type.

03

Keep project records complete

New documents receive metadata and a link in the project or CRM record.

A strong fit when …

Recurring office tasks follow clear rules, consume small blocks of time every day and should run reliably without replacing existing systems.

  • You handle recurring documents using repeatable rules.
  • The intake, target system and accountable business role can be named clearly.
  • Exceptions are allowed to remain visible and move to people deliberately.
Transparent potential estimate

Estimate time savings with your own volume

The calculator uses 4 minutes today and 1 minutes after automation as fixed example assumptions. It does not replace process analysis.

Illustrative estimate based on the visible assumptions — not a guarantee.

35Hours per month
420Hours per year
Additional measures after launch Handling time Open exceptions Manual transfers
Frequently asked questions

What decision-makers should know before starting

How does automatic document filing work in practice?
A file arrives by email, scan, upload or approved folder. The workflow then validates the required data, runs approved steps and routes exceptions to the responsible person with context.
Which systems can be connected?
Typical integrations include SharePoint, OneDrive, Microsoft 365, Google Drive, DMS, E-Mail. The decisive factors are a stable interface and clearly defined ownership of each data field, not a specific tool.
Which tasks deliberately stay with the team?
Unclear document types, conflicting data and legally relevant classifications are presented for review rather than guessed.
How is the automation introduced?
We choose one frequent, tightly scoped task, document its intake, destination and exceptions, and test it with a small user group. A tightly scoped first process typically takes 2–4 weeks; scope, interfaces and approvals determine the actual plan.
How can the benefit be measured?
Before implementation we record volume and current handling time. After launch we also compare Handling time, Open exceptions, Manual transfers. The calculator on this page is a transparent estimate, not a promise.
Content reviewed on 26 July 2026 About Lyron AI
arrow_back Back to all products