AI Document Understanding

Overview — AI Document Understanding uses a machine-learning model together with optical character recognition (OCR) to interpret, extract and process information from a dynamic application form in PDF format. It increases confidence when extracting data for subsequent recreation of the application form.

How it is used — Once the model has been applied, OCR reads the PDF text and the model intelligently identifies and extracts data points within the form regardless of variability, by learning the patterns and context of the data fields rather than relying solely on their positions. This means that even if the form's content or structure changes over time, the model can still recognise and process the relevant information, improving efficiency and reducing the error rate associated with manual data entry.

Scope and limitations — The system is designed specifically to extract predefined fields from a predefined digital document. It does not extract fields it has not been trained to extract, and it does not retrain itself.

Human oversight — Where the model's confidence threshold is not met in identifying data, the document is referred to the business area for processing. An audit trail of the automation's output is retained in line with platform retention policy, and business-process data output is retained in line with the applicable data-retention policy.

• Saves time and improves efficiency by identifying specific data within digital documents that would otherwise have to be read manually.
• Provides a consistent approach to identifying specific data within unstructured, variable inputs compared with people conducting the same activity, reducing human error.