Since the Flowr integration in PowerAutomate knows which fields should be dates or not, this should be fairly straight forward.
I am seeing dates coming in from the recognised documents in 3 or 4 different formats – and its evident the Encodian Flowr is merely returning the text value it finds in the document and its not preprocessing or recognising or standardising the date as a true date or returning in ISO format. This should be perhaps as it would be very useful especially as the “field” returned by Encodian is known as a date by virtue of the label InvoiceDate / DueDate etc. Otherwise secondary / additional steps are needed to recognise the dates and process further.
All other AI recognition tools return a valid ISO date, for example Microsoft AI invoice recognition action returns three JSON fields for each OCR'd field that one would usually expect to be a date on such documents:
- the date value of the relevant field as a Power Automate date (i.e. InvoiceDate / DueDate) which can be handled as such
- it also returns the literal text value of the date field exactly as its written; and
- a confidence score on the recognition “success” for want of a better description.
Thanks.
