Showing results for 
Search instead for 
Did you mean: 
Frequent Visitor

Need help with analyzing PDF docs

I'm trying to use AI Builder to read PDF files and extract fields from it.  I've gone through the sample PDFs and that works OK, but when I upload my own PDFs (and they are all very similar and very clean), I keep getting the error "Couldn't analyze documents.  Try again later".


I was at one time able to get them to be analyzed, but it didn't capture all the fields in the PDFs.  And I didn't see a way to "help" it know what I'm trying to get from the PDFs. 


Any suggestions?

Resolver I
Resolver I


Hi @dibb ,


I experienced the same problem, do you have any solution?

@JoeF-MSFT Could you please help us solve this problem? Any assistance is greatly appreciated!

Frequent Visitor

No solution yet.

Power Apps
Power Apps



To train your AI Builder Form Processing model :

  • documents used for training the Form Processing model must use the Latin alphabet (English characters),
  • the total size of the dataset used must not exceed 4MB,
  • the total number of pages of the dataset used must not exceed 50 pages,
  • make sure your administrator has assigned you a security role with all organization privileges over the entity Note from Core Records, and read privilege over the entity you are using to select object names
  • other requirements are available in the documentation :


If a training fails, you can try :

  • recreating a new model and deleting the draft that failed,
  • for PDF files, reducing the number of pages per document (you can use the "print to PDF" printer to select only some pages from a document),
  • test with the sample material from .

Hope that helps.


If the issue persists and if it’s not confidential, would you be ready to share the data set you are using so we can check further ?


Thanks !

Community Support
Community Support


Hi @dibb ,


Have you had an opportunity to apply @jk_one 's recommendations to adapt your Flow?

Please take a try.


Best Regards,

Community Support Team _ Barry
If this post helps, then please consider Accept it as the solution to help the other members find it more quickly.
Frequent Visitor

Yes my PDFs are all in English

The total size of all the PDFs is almost 1MB

The total pages 25

I have the correct security role


I actually also opened a case with Microsoft and when they looked at it I could tell they had never used AI builder before so they called me back with someone who was a little more familiar with it but they couldn't help yet either. They said they'd call me back but haven't yet.  This was on Friday 12/13 so they are probably still trying to figure it out.

Hello @dibb,


We do have manual tagging of unrecognized fields on our roadmap for the future (see, but in the meanwhile, would you be ready to share the data set you are using / want to improve ?


Also, if you can send me your support incident id in a private message, I can check it out.


Thanks in advance !


J from the AI Builder team



Hi @jk_one 

I have a similar issue whereby the pdf document seems to be structured and legible but the AI model is failing to recognize all the fields as well as a table in the body of the document.  I tried creating the model with up to 50 documents and it still fails to recognize the fields.

Will the manual tagging of unrecognized fields feature also enable the tagging of tables?

Let me know if you would like to see some examples via PM.


Hello @Andre,


Yes, absolutely, please send me a sample dataset via PM so that we can check it out ! Tables should be recognized "out of the box" but the samples will allow us to confirm (and improve).


Thanks ! 


J from the AI Builder team


Helpful resources

UG GA Amplification 768x460.png

Launching new user group features

Learn how to create your own user groups today!

Community Connections 768x460.jpg

Community & How To Videos

Check out the new Power Platform Community Connections gallery!

M365 768x460.jpg

Microsoft 365 Collaboration Conference | December 7–9, 2021

Join us, in-person, December 7–9 in Las Vegas, for the largest gathering of the Microsoft community in the world.

Top Kudoed Authors
Users online (1,683)