Invoice Information Extraction

Problem Statement

An accounting and finance company, with an accounts payable team handling over 50000 documents per month had a major challenge addressing the huge volume of invoices and purchase orders of various inputs and formats resulting in a decline in the overall productivity of their business. Manual data entry and traditional OCR methods had failed to meet the expectations of these businesses in processing such documents due to the following reasons:

    sample invoice
    purchase order
  • Variable formats and layouts.
  • Information to be sourced from multiple documents.
  • Watermarks/Inkmarks & signatures on invoices.
  • Inadequacy of existing OCR based solutions in extracting accurate information at a fast speed.

The solution

  • Cerescope D-tect uses the power of artificial intelligence (AI) to automate invoice processing.
  • D-tect uses cognitive methods to read from varied layouts and multiple documents. Semantic labelling of the invoice frees the team from the hassles of template definition.
  • D-tect is great at handling uncertainty and variation, thus helping in the accurate extraction of information from invoices.


  • After implementing D-tect, accuracy of more than 95% was recorded.
  • Reduced handling time per document leading to an increase in productivity.