Unanswered

Invoice attachments from email - extract Invoice Info

(0) Share

Report

Posted on by rowan_preston

I'm building a flow that takes invoice attachments from email, saves them to Onedrive and then extracts the invoice information and uploads to my dataverse table. All working great - except when I tested an invoice from a phone company which happened to have 452 pages...... It just hung for a couple of hours before I cancelled it.

I guess i need to be able to limit the number of pages that are saved from the attachment to just the front page basically as this will generally have all the information required on it.

Anyone know if/how this is possible???

Thanks

Categories:

AI Builder

All responses (4)

Answers (0)

Sort by

CedrickB Microsoft Employee on at

Like
a
(0)

Report
Copy link

Link copied!

It is, see https://learn.microsoft.com/ai-builder/form-processing-model-in-flow#page-range

Generally, it is a best practice optimizing costs to limit the number of pages analyzed to what is strickly needed.

Though, if the document is really too large, it may not even be properly uploaded to our service.
Linn Zaw Win 2,996 on at

Like
a
(1)

Report
Copy link

Link copied!

@rowan_preston
Extracting the first page of the PDF document is possible but only with third-party connector from Encodian which will have an additional cost.
Check out the video below.
https://www.youtube.com/watch?v=WBHIXW_QaBk
RoyIssa 2 on at

Like
a
(0)

Report
Copy link

Link copied!

Hi Rowan that's a very interesting flow to me. Is it possible to share some details about how you went about it so that I get inspiration out of it? thanks
takolota1 4,980 Moderator on at

Like
a
(0)

Report
Copy link

Link copied!

You could also try this new GPT data extraction method: https://powerusers.microsoft.com/t5/Power-Automate-Cookbook/Extract-Data-From-PDFs-and-Images-With-GPT/td-p/2201345

It doesn’t require specifying certain document areas, wordings, styles, data, etc. It just OCRs the file, converts it to a replica text (txt), and passes it to a GPT prompt where you can ask GPT to do whatever you want with the document data.

You can easily control the number of pages it uses by using the Take( ) expression on the Results output used in the 1st outer Apply to Each loop after the OCR step.