How to extract invoice data to Excel automatically with AI Builder and Power Automate

In Power Automate, go to AI hub and select Prompts, then Build your own prompt. (Under AI models there is a pre-built invoice model, but it needs training with sample documents. A custom prompt works right away.)

In the instructions, ask the model to extract Vendor Name, Invoice Number, Invoice Date, Due Date, Item Name, Quantity, Unit Price and Amount, and to return a JSON array where each element is one line item. The prompt in the video also includes:

• Normalization rules: a fixed date format, and plain numbers with no currency symbols or commas
• Missing values: an empty string for text and 0 for numbers
• Calculation rules: if the unit price is missing, divide amount by quantity. If the amount is missing, multiply unit price by quantity

Delete the document placeholder text in the instructions, then use Add content and choose Image or document so the prompt accepts a file.

The default model is GPT-4.1 mini, which is fast and cost effective for this task. Under the model response settings, change the output format to JSON so the result can be mapped to Excel columns.

Upload a sample invoice to the document input and click Test. Within a few seconds you should see the vendor name, invoice number, dates and each line item, formatted as you asked. Name the prompt (in the video: Invoice Data Extractor) and save it.

Create an Excel file in OneDrive (in the video: Invoice Data Log) with the headers Vendor Name, Invoice Number, Invoice Date, Due Date, Item Name, Quantity, Unit Price and Amount.

Select the headers, choose Format as Table, pick a style, tick My table has headers and click OK. The table name shows under Table Design (the default is Table1).

Select Create, then Instant cloud flow. Name it (Extract Invoice Data), choose Manually trigger a flow and click Create.

Add the OneDrive for Business action List files in folder and use the folder picker to select the folder that holds your invoices.

Add Apply to each from Control and set its input to the output (value) of List files in folder, so each file is processed one at a time.

Inside the loop, add the OneDrive for Business action Get file content and set File to the Id from List files in folder. The Id uniquely identifies each file.

Still inside the loop, add Run a prompt from AI capabilities. Select your Invoice Data Extractor prompt. For the document input, pick the File Content output of Get file content.

Add the Excel Online (Business) action Add a row into a table. Set Location to OneDrive for Business, Document Library to OneDrive, pick the Invoice Data Log file and Table1.

Expand Advanced parameters, select all columns, and map each one to the matching field from Run a prompt (use See more to find them): Vendor Name, Invoice Number, Invoice Date, Due Date, Item Name, Quantity, Unit Price and Amount. Because the prompt returns an array of line items, Power Automate places the Excel action in its own loop so one row is added per line item. Save the flow.

Put a few invoices in the folder, click Test, choose Manually, then Continue and Run flow. Watch the Excel file: rows appear for each line item of each invoice.

Use Save As to copy the flow (Extract Invoice Data Mobile). Open the copy, turn it on and edit it.

A phone capture handles one invoice at a time, so you do not need the folder listing or the file loop. Drag Run a prompt directly under the trigger, drag the line item loop with the Excel action under it, then delete List files in folder, the outer Apply to each and Get file content.

Select the Manually trigger a flow trigger, click Add an input and choose File. This lets you upload a file or take a photo when you run the flow from your phone.

In Run a prompt, clear the old document input and set it to the file content (contentBytes) from the trigger. The Excel mapping stays the same.

Open the Power Automate mobile app, find the mobile flow, tap the three dots and choose Run flow. Tap the file input, choose Launch the camera, photograph the invoice, tap Use photo, then Run flow. The data lands in the Excel table within seconds.