Automation
AI Document Processing for Logistics
Built two production-grade document AI pipelines for a logistics firm handling RFQs and invoices at volume. The system extracts data, drafts quote responses, validates invoice fields against the ERP, and routes anything low-confidence to a human before it touches financial records.
The Problem
A logistics firm was receiving RFQs and invoices by email and PDF at a volume that overwhelmed manual processing. Staff had to read each document, key the data into the ERP by hand, and draft replies, a slow process with a real error rate on every entry. At scale, this wasn't just inefficient; it was a recurring cost centre with real financial risk sitting inside every keyed number.
The Solution
I built two n8n workflows to handle this end-to-end. The first ingests RFQ emails, extracts shipment details, drafts an AI-generated quote response, and logs it to the CRM for review. The second takes invoice PDFs through OCR, extracts and validates the fields, then posts directly to the ERP over its REST API, with retry logic to handle transient failures without manual intervention.
The Result
The core design decision was a confidence gate on every extracted field, not just a single score for the whole document. High-confidence extractions post automatically, while anything uncertain gets flagged to a human reviewer instead of risking a wrong number in the ERP. This meant the client got the speed of automation without gambling on accuracy where it mattered most, proving that reliability, not cleverness, is what makes document AI production-ready.