Loading patterns…
Route incoming documents — PDFs, scanned images, email attachments — by extracting text and classifying them automatically. Replaces human eyeballs at the front door.
Rendering diagram…
Use this pattern whenever documents flow into your process and a human has to look at each one just to decide where it belongs:
The triage is the bottleneck, not the work that follows. Automating the triage frees the humans for the judgment calls.
Three stages:
invoice_number, total_amount, counterparty_name, effective_date. Most commercial OCR services bundle this.Route each document based on its class plus any extracted fields (e.g. invoice with total > $10K goes to tier-2 AP).
Confidence scores exist for a reason. Never auto-route a classification below your confidence threshold — send it to a human review queue and learn from the corrections.
Before automating, know what you're replacing. If humans are right 97% of the time and the model is 93%, you've made things worse. Sample a week of decisions; label them; compare.
Forms benefit from structured OCR (AWS Textract's forms, Google's Form Parser). Free-text documents want plain extraction. Don't pay for structured OCR where you don't need it.
Rules: fast to start, brittle when formats change. Model: slower to start (needs labelled data), robust to variation. Hybrid is common — rules handle the obvious cases, model handles the rest, human handles low-confidence.
No production doc pipeline runs at 100% accuracy. Plan a review queue for low-confidence items. Route human corrections back as labelled training data.
Periodically sample outcomes and manually verify. Drift happens — a supplier changes their invoice layout and your model quietly starts misclassifying. Monitoring is the safety net.
/dev/null is a customer-impact bug waiting to happen.Collect data from multiple sources, merge it into a formatted document, route it for review and approval, then store the final version. The backbone of contract, proposal, and report workflows.