← All cases
Industrial parts · supplier quotes

Supplier PDFs into the ERP, without retyping a line

Every supplier quote arrived as a PDF and left as manual data entry — about two hours a day per employee. An LLM service now does the transfer and writes the result into the ERP.

Client
Spare parts & industrial equipment · Germany · $15M revenue · 100 staff
Scope
Inbox → classification → extraction → ERP
Duration
1 month to rollout
Role
Solo — design, build, hand-over
PDF quote extraction service
The extraction service: a supplier PDF in, structured quote lines out.
Time saved
600h
of manual data entry
Throughput
1,500–2,000
PDFs processed per month
Rollout
1mo
from start to production
Per employee
2h
a day given back
§ 01 / Problem

Two hours a day retyping what a PDF already said

Suppliers send quotes as PDF files. Everything in them had to be moved into the ERP by hand.

An employee opened each file and transferred the prices, delivery times and line items into the ERP field by field. The document already contained every value — the work was purely mechanical transcription.

Across the team this routine consumed roughly two hours a day per person, and it grew with every new supplier added.

~2 hper employee, daily
1,500–2,000PDFs / month
§ 02 / Solution

A service that reads the quote and fills the ERP

The PDF never needs to be opened by a person — it is pulled, classified and extracted.

01

Email arrives

inbox

A message comes in from the supplier with the quote attached.

02

Pull and identify

detect

The bot takes the PDF out of the inbox and determines that it is a supplier quote rather than any other document.

03

Classify and extract

parse

An LLM pass extracts prices, delivery times and line items from the document layout.

04

Write to the ERP

erp

The structured values are written into the ERP, where the quote is ready to work with.

§ 03 / Result

Six hundred hours, and a month to get there

Measured after the service took over the full supplier-quote flow.

~600 h

Of manual data entry removed

1,500–2,000

PDFs processed every month

1 month

From first commit to production

§ Built with
PythonLLM extractionPDF / OCREmail ingestionERP integration