home / use cases / documents and vision / document-classification-filing

Every document filed right, every time

An Aivell picks up every incoming document — emailed, scanned or pushed by webhook — reads it inside your office, works out what it is and who it belongs to, renames it to your convention and files it in the right folder on your share or Drive. Unclassifiable ones land in a triage folder for a human.

Watch the demo

starts when: A document arrives in the intake mailbox or lands from the office scanner

The problem

Every office has the folder. It is called “Scans” or “To file”, and inside it live three hundred documents named scan_0217.pdf. The naming convention exists — it is written down somewhere, probably in the same folder — but filing is nobody’s favourite job, so it happens in bursts, weeks late, by whoever loses the coin toss. Then a client calls, and finding one delivery note costs twenty minutes.

The catch with fixing this in the cloud is that the pile is precisely the paperwork you least want uploaded: signed contracts, identity documents, payslips, medical certificates. Sorting it means reading it, and reading it on someone else’s servers is a hard sell to any client — or any data-protection officer.

How your Aivell does it

  1. Everything funnels to one place. Documents arrive by email at the intake address, or straight from the office scanner and your portals via webhook. No human has to move them first.
  2. Each page is read on the box. Typed PDFs, skewed scans, photographed pages — a vision model running onboard reads them all. No page ever leaves your network to be classified.
  3. The document is identified. Invoice, contract, delivery note, certificate, ID — and, from names, numbers and letterheads, which client or matter it belongs to.
  4. Renamed and filed, by your rules. Your convention, applied every single time, into the right folder on the share or Drive. Personal documents go only to the restricted folder — that route is written into the chore.
  5. Doubt goes to a human. Anything ambiguous is parked in triage with a note to the person you named. An index sheet records every filing, and a Friday log shows the week at a glance.

The convention finally enforces itself, because the one doing the filing never gets bored and never improvises. And the guardrails hold the archive steady: your Aivell can add documents to it, but what is already filed cannot be renamed, moved or deleted by anyone but you.

The chore, as you'd write it.

No flowcharts, no code — a chore is just a message to your Aivell, in your own words. It shapes it into a solid, guarded procedure and follows it to the letter. This one:

Giulia Aivell · sort the incoming paperwork ● active

pick up everything sent to docs@ or dropped by the office scanner

read it and decide what it is — invoice, contract, delivery note, certificate, id

rename it — date, sender, type — following the office convention

file it in the matching client folder on the share

ids and anything personal go to the restricted folder, nowhere else

can't tell which client it belongs to? park it in triage and ask Chiara

keep an index sheet of what was filed where

friday afternoon, send the week's filing log

You
Got it — I'll start right away, and check with you whenever something needs a human.
Giulia Aivell

The tools it uses.

Connected with single-click passkeys — no copied tokens, no OAuth hell.

Email receives documents, asks when unsure
Documents & Vision reads and classifies every page onboard
Network folders files under your naming convention
Webhooks accepts drops from scanners and portals

Sensible guardrails.

Every tool can be limited to exactly what this chore needs. For this one, you might set:

Network folders May create files inside the archive tree. Renaming, moving or deleting documents already filed is blocked.
Email Writes only to colleagues on your own domain — this chore never mails anyone outside the company.

Questions, answered.

How does it know our naming convention and folder structure? +

You describe both in the chore, exactly as you would explain them to a new office manager — "date first, then the client, then what it is; certificates go under compliance". If the convention changes, you edit the sentence and every document from then on follows it.

We receive IDs, payslips and signed contracts. Is it safe to let software read those? +

This software, yes — because it reads them behind your firewall. Nothing is uploaded to a classification service; the pages are read by a model running on the box. That keeps sensitive paperwork under your roof, which is [privacy by design in the GDPR sense](https://seresa.io/blog/eu-ai-act-for-marketers/gdpr-article-25-and-local-ai-why-on-premise-llm-inference-is-privacy-by-design), not just a policy promise.

What happens to a document it cannot place? +

It goes to a triage folder, and the person you named gets a short message saying what arrived and why it could not be placed. Nothing is filed on a guess, and nothing sits unread in a mailbox either.

Can it file into Google Drive instead of a network share? +

Yes — the destination is whatever your office already uses, Drive or a plain network share, same convention, same restricted-folder rules.

Related.

What is an Aivell?

An Aivell is an on-premise AI colleague: a small box that plugs into your network, runs its own AI onboard and takes the repetitive work off your desk. Prompts, documents, data — nothing ever leaves your office. Unplug the internet: it keeps working.

You describe each task as a chore, in plain language. Your Aivell turns it into a solid, guarded procedure and runs it in the background — with guardrails on every tool, a complete audit trail, and approvals in your hand when you want them. One-time setup, fixed monthly fee. No tokens, no overages.

Watch the demo

onboard inference · air-gapped · fixed monthly fee