Skip to main content

Optical Character Recognition

Optical Character Recognition (OCR) Wrk Actions use AI and OCR to read, classify, and extract structured data from files, without manual data entry.

What it is​

These Wrk Actions accept documents (PDF, PNG, JPEG, TIFF) and return text, tables, labeled fields, or specialized data from pre-built templates (invoices, government IDs, billing documents, and more).

Good for​

  • Accounts payable and invoice processing
  • Identity verification from government-issued documents
  • Extracting tables or labeled fields from contracts, forms, and reports
  • Classifying document content, identifying entities, or analyzing sentiment in text
  • Querying a document with natural language when field labels are inconsistent

What it does​

  • Retrieves all text or specific labeled fields from a document
  • Extracts tables and line items from billing documents
  • Identifies entities, sentiment, and specific pages within multi-page files
  • Runs pre-built extraction templates for common document types (Quebec health cards, US government IDs, birth certificates, and more)
  • Accepts custom extraction when fields are clearly labeled on the document

Go deeper​

Reference​

  • AI in Wrkflows: document extraction patterns and when to add Human in the Loop review