Invoice Processing Automation with Power Automate

Overview

• Client: Burger company
• Goal: Automate invoice processing from email attachments to Excel.
• Tools: Microsoft Power Automate, AI Builder.

Input Data

• Invoices saved locally (6 invoices provided).
• Structure: Each invoice contains:
  • Invoice number
  • Date
  • Addresses
  • Total amount
  • Taxes
  • Table with items (varying number of rows)

Process Overview

1. Download Invoices: Recommended to download all six invoices for practice.
2. Structured vs Unstructured Data: Today's focus is on structured documents (same format). Unstructured data will be covered in future videos.

Creating the AI Model

1. AI Builder: Navigate to AI Builder in Power Automate.
2. Form Processing: Click "Build" and select "Form Processing".
3. Model Setup:
   • Name the model (e.g., "Invoice Processing").
   • Specify fields to extract:
     • Invoice Number
     • Date
     • Total
     • Table items (single page table selection).
4. Tagging the Data:
   • For each invoice, identify and tag the invoice number, date, total, and table items.
   • Ensure to correctly tag data on all five training invoices.
5. Training the Model:
   • Upload the first five invoices.
   • Tag the necessary fields and table headers.
   • Train the AI model.

Creating the Power Automate Flow

1. Create a Flow:
   • Choose an "Automated Cloud Flow".
   • Trigger: New email with attachments containing "invoice" in the subject.
2. Set Up Loop:
   • Loop through each attachment in the email.
3. Output to Excel:
   • Create an Excel workbook in OneDrive.
   • Define a table structure with headers: Invoice Number, Date, Total, Items.
   • Use the AI model to extract data from the invoices.
   • Add extracted data to the Excel table.

Testing the Flow

1. Testing with Invoice Number Six:
   • Send an email with invoice number six attached to test the flow.
2. Inspecting Results:
   • Upon successful run, data should appear in OneDrive Excel workbook.
   • Check and format the data as needed (e.g., currency formatting).

Final Steps

• Review run history to check for errors and ensure data integrity.
• Adjust JSON output if necessary to refine data formatting.
• Ensure proper column mapping if rows need to be separated.

Additional Resources

• For further learning, consider taking a full Power Automate course.