Use "Start with PDF" AI/OCR capabilities at runtime to extract data from PDF documents

The idea here is to extend the engine behind the "Start with PDF" feature to manage business cases at runtime. For example, we could have a workflow action named "Extract data from PDF", that would open a "mapper" like the one we have in the "Start with PDF" UI, but instead of mapping PDF areas to form controls, we could map them to workflow variables.

E.g.: in the case of an Invoice Management workflow, it could start with a PDF document added to a SharePoint library, then this "Extract data from PDF" action could get the data from the top left corner of the document and put that in the "vendor details" variable, the data from the bottom right could be saved to the "total amount" variable", and then have that invoice approved through an "Assign a task" action surfacing the variables.

There are multiple use cases that could benefit from this feature, like automatic mail sorting, resume processing, document control (checking that a generated PO matches the expected numbers), etc.