Seeking limits of Document Intelligence while extracting data from complex documents

I spent several months working on data extraction from multi-page documents related to the insurance industry.

Goal was to leverage built-in features of Azure Document Intelligence only to:

• validate presence of various forms inside the document
• verify if forms are correctly and fully filled and signed
• extract values from forms and tables

I will show samples of original files, explain approaches we tried as a team and present outcome from most successful one, also sharing tips&tricks for those of you facing similar challenge.