OCR (Optical Character Recognition)
Technology that converts images of text — scanned documents, photos, PDFs with embedded images — into machine-readable characters. OCR is the foundational layer of document processing, but alone it only produces raw text. AI is required to understand what that text means.
What is OCR?
Optical Character Recognition (OCR) is the process of detecting and converting text in an image into digital characters that a computer can process. A scanned invoice, a photo of a delivery note, a fax saved as a PDF — OCR reads the pixels and outputs the characters. Without OCR, these documents are just images. With OCR, they become text that can be searched, extracted, and processed.
OCR technology has been available since the 1970s and has matured significantly. Modern OCR engines — including Google Vision, AWS Textract, and Microsoft Azure Computer Vision — achieve high accuracy on clean, printed text and handle most standard business document formats reliably.
What OCR Does — and What It Does Not
OCR answers one question: "What characters appear in this image?" It does not answer: "What does this document mean?" or "Which of these numbers is the invoice total?" or "Does this delivery note match the purchase order?"
OCR output: A raw text dump of everything visible on the page, often without meaningful structure
What AI adds: Understanding of document type, field identification, relationship between values, validation against business rules
OCR limitations: Handwritten text (lower accuracy), poor scan quality, overlapping elements, complex table structures, non-Latin scripts
Modern document AI: Combines OCR with layout analysis (understanding where text appears on the page) to produce structured output, not just raw characters
OCR in Operations
For any operation that receives paper or scanned documents — delivery notes from drivers, signed CMR documents, faxed orders, inspection certificates — OCR is step one. Without it, document automation is impossible. But OCR alone is not document intelligence. A scanned invoice run through basic OCR produces a wall of text: line items, headers, totals, addresses, and footer boilerplate all mixed together. The AI layer built on top of OCR is what identifies which numbers are quantities, which are prices, which is the VAT amount, and which is the total — and extracts them into a structured record that can be validated and posted to the ERP. At Lleverage, OCR is infrastructure, not a product. The value is in what happens after the characters are read.