Solutions / Document classifier

Document & email classifier

Inbound mail and documents are read, classified and filed to the right team, folder and system — so nothing sits in a shared inbox waiting to be noticed.

Operations Document processing E-commerce · Logistics
3–5 hrs
a week returned per operations lead
0
documents lost in the shared inbox
24/7
sorting, including out of hours
Who it is for

Operations leads whose shared inbox is the de facto routing layer between customers, suppliers and several internal teams.

The problem

A shared inbox is a queue nobody owns.

01

Routing is a person’s morning

Someone opens each message, works out which team it belongs to, forwards it and files the attachment. It is pure overhead and it happens every day.

02

Things fall through when that person is away

The routing rules exist only in one head. Holidays and sick days show up as a backlog and a missed deadline.

03

Nobody can prove what arrived

Asking whether a document was received means searching a mailbox. There is no register, so there is no answer.

How it works

From trigger to result, step by step.

01

Ingest from every channel

Shared mailboxes, scanner output folders, cloud drives and portal downloads are watched in one place.

02

Identify the document

The agent determines type — order, invoice, complaint, contract, delivery note, CV — along with sender, reference numbers and dates.

03

Apply your routing rules

Classification drives the destination: a team, a folder, a ticket, a CRM record, or all four. Rules are written down and versioned instead of remembered.

04

File and register

The document is stored where it belongs, renamed to your convention, and written to a register that answers "did it arrive, and when".

05

Escalate what does not fit

Unknown types and ambiguous cases go to a person with the agent’s best guess attached, and the resolution feeds back into the rules.

Before / after

What changes on the ground.

Today, by hand
×A person sorts the shared inbox every morning
×Routing knowledge lives in one head
×Attachments filed by hand, named inconsistently
×No answer to "did we receive it?"
With the automation running
Sorting happens on arrival, around the clock
Routing rules are explicit and versioned
Files named and stored to convention automatically
A searchable register of everything received
What you get

Delivered, not demoed.

A classifier tuned to your document types and your routing rules
Connections to your mailboxes, drives and downstream systems
A register with search, so receipt questions get answered without a mailbox hunt
An exception queue for anything ambiguous
Documentation and handover, including how to add a new document type
Built with

We build in your stack rather than moving you onto ours. The list below is what this solution most often connects to — other systems are a scoping question, not a blocker.

Gmail API Microsoft 365 Google Drive SharePoint OpenAI Anthropic AWS Textract Postgres n8n Slack
Time to production3–5 weeks
Build priceFixed quote
First stepFree mini-audit
Honest limits

When this is not the right solution.

·If two people can handle the volume comfortably, this is a nice-to-have rather than a payback case.
·If your document types are genuinely unbounded and each needs judgement, classification will route but not decide — expect assistance, not replacement.
·If the downstream systems have no API and no import path, the filing half of the value is not reachable; we will say so before quoting.

Questions we get about this one

In practice, as many as you can describe with examples. Accuracy depends on how distinguishable the types are, not on the count — twelve clearly different types are easier than three that look almost identical.

Bring us the process that hurts.

The mini-audit is free: we take your version of this process apart and tell you plainly whether automating it pays. If it does, you get a scope and a fixed price.