What is OCR?

To a computer, a scanned PDF or a photographed invoice is just an image — pixels without meaning. OCR, optical character recognition, is the technology that reads those pixels and turns them into text and data software can actually work with: searchable, copyable, importable into your accounting system. The current AI-driven generation goes far beyond reading letters: modern systems understand document structure — they know the number next to "Total due" is the invoice amount, they extract the supplier, tax ID, due date and line items, and they cope reasonably with crooked photos, crumpled receipts and legible handwriting. The flagship business case is the inbound invoice pipeline: instead of someone retyping every supplier invoice into the books, incoming documents pass through OCR, extracted fields flow into the system automatically, and a human validates only the items flagged as uncertain — at medium volumes the saving runs to dozens of hours a month, with retyping errors eliminated as a bonus. Other fertile ground: expense receipts, transport documents in logistics, HR files, identity documents at onboarding, and making a historical archive searchable by content rather than by file name. The calibration note worth keeping: accuracy is never 100 percent, so a well-designed flow promises triage rather than magic — the certain items pass automatically, the uncertain ones request a click — and the number that matters commercially is net hours saved after validation, not the raw recognition rate.