← AI Terminology

OCR / Document AI

OCR (optical character recognition) extracts text from images of documents; Document AI extends this to layout, tables, forms, and key-value understanding.

Essential infrastructure for enterprise automation and RAG over scans.
Why It Matters in AI
Most business knowledge sits in PDFs and scans. OCR/Document AI unlocks search, extraction, and LLM grounding on real paperwork — a high-ROI AI vertical.
Key Points
Aspect Description
OCR Detect and read characters/words/lines
Use Invoices, IDs, archives, RAG chunking
Layout Reading order, blocks, tables, forms
Models Classic Tesseract; modern layout transformers; VLMs
Quality Noise, handwriting, multi-column pitfalls
Related VLM, NER, information extraction
Simple Analogy
A clerk who can type up what’s on a photographed form and also know which box is “address” versus “total.”
Common Usage Examples
  • Tesseract/PaddleOCR pipelines
  • LayoutLMv3 document understanding
  • VLM OCR for complex pages
  • Feed OCR text into RAG
Summary
In short: OCR and Document AI turn images of paperwork into structured machine text — the gateway from scans to searchable, automatable knowledge.