Extract information from expense receipts, invoices, Canvas app for submission- approval

Azure AI's prebuilt model for invoice processing leverages advanced machine learning and artificial intelligence to streamline the extraction and processing of key information from invoices. This sophisticated tool is designed to handle a wide range of invoice formats and structures, automatically identifying and extracting crucial data such as vendor details, invoice numbers, dates, and line items. By automating these tasks, businesses can significantly reduce the time and effort required to process invoices, minimize errors, and ensure timely payments.

In this blog, we will explore the transformative impact of Azure AI's prebuilt invoice processing model. We'll dive into its features, how it works.


So lets start the process by creating canvas app, I created one table "invoices" and built a simple canvas app to submit the expense receipts. 


When the user submit the new expense the power automate flow triggers and read the attached file and write back the values to the data base. 

Trigger the flow when the new expense is submitted/added

To get the attached file from the submission use the action download the image

Now select the azure AI prebuilt model extract from the invoice

If we want to show the extract information  in a format approver can read before approves it, to do this we need to do couple of steps. 

Initialize the variable with the fields you want to populate


Now do the Append to string variable action


To put the information about the expenses back to user by email, I am creating HTML table. 

Getting the user details by using get my profile 


Now initiate the approval process to approve the expenses.



Now based on the outcome, we will update the record and send the confirmation about approval/rejection. 


As we are getting the values from AI in a certain format, we can write these values to other system with ease. 


This is the sample document I have used to test the model which is working nicely


Based on your business requirements if the AI prebuilt model is not working as expected we can also built our model pretty easily. 

Just follow the wizard setup to create a new model, some of the fields you may need to adjust based on the document

You can add many collection to train the model based on the type of requirement yo have, Each collection should have min 5 documents, Microsoft recommending 20 examples could yield better results. 

Thank you for visiting this blog 🙏. 

Comments

Popular posts from this blog

How to update 100000(hundred thousand) records with power automate flow by using chunk expression, Split the list and pass onto child flows.

How to keep flows remain on if the flows don't run for 90 days

D365 CRM-Change Entity translations(Sub area) in Modern driven app.