Cloudmersive Document AI is here, and it’s ready to revolutionize your document processing workflows.

Document AI empowers enterprises to intelligently summarize or categorize document contents, retrieve form responses, pull data from tables, and extract fields of data from invoices, receipts, tax forms, and other standard documents.

Cloudmersive Document AI Field and Table Extraction in Power Automate

The Document AI API can be imported to Power Automate as a Custom Connector, and from there it can be plugged into any new or existing document processing workflow with ease.

In this article, we walk through an example Power Automate flow which takes advantage of the Document AI API’s field and table extraction capabilities to process items from an example “employee receipt” and return that data as a JSON object array. This extremely simple flow allows us to ready receipt data for storage in an ERP, CRM, or database of any kind, and it also drastically simplifies our ability to run approval workflows in Power Automate.

Uploading a Custom Document AI Connector to Power Automate

Before we begin walking through our example field and table extraction workflow in Power Automate, we’ll first address that we’re using the Document AI API as a custom connector in this case. Uploading any Cloudmersive API to Power Automate as a custom connector is as simple as 1) uploading the API spec to the power platform via JSON file or URL and 2) setting a host endpoint and base path in the General information tab of the custom connector editor.

If we’ve deployed the Document AI API to a private or managed instance endpoint, we can set that endpoint as the host in our custom connector to take full advantage of our infrastructure in Power Automate.

Extracting Fields and Tables from a Receipt in a Power Automate Flow

After we’ve created our Document AI custom connector, we can immediately begin designing our AI field and table extraction flow.

In this example, we’ll build an Instant Cloud Flow with a manual file input trigger. This will simplify testing with one or multiple example receipt files.

We’ll configure our Manually trigger a flow step to accept a single file input. This step should specify that our flow can accept PDF, DOCX, PNG or JPEG files.

We’ll now place our Document AI field and table extraction action directly after our manual trigger or input.

To find this action, we’ll first navigate to the Custom connector filter in the action search window and select our custom uploaded Document AI connector.

On the Document AI actions list, we’ll find and select the action titled Extract All Fields and Tables of Data from a Document using AI.

We’ll now structure our field and table extraction request.

This part is extremely simple. We only have 3 request parameters to worry about, and 2 of them are related to our file input.

The RecognitionMode parameter is the only unique option amongst these, and it can be set to either Basic or Advanced recognition mode (default is Advanced). These recognition modes offer a tradeoff between API processing speed and accuracy. We want the most accurate possible results in this flow, so we’ll leave this parameter on its Advanced setting.

After just two quick steps, our flow is already functional. All that’s missing are subsequent steps which process or share our data in some way.

In this example, we’ll create a basic Approval workflow which directly follows our field and table extraction workflow. We’ll send the receipt fields (which include the total cost of the receipt) to a stakeholder and wait until they approve the receipt before taking a follow-up action in our flow.

First, we’ll create a String variable called Receipt Fields using the Initialize variable action. We’ll loop through the body/FieldResults array from our Document AI response to populate our variable with each FieldName and FieldStringValue.

Next, we’ll add a Start and wait for an approval action to our flow, and we’ll set the Approval type to Approve/Reject – First to Respond. In the Details section, we’ll add the Receipt Fields variable we just generated.

We’ll include the uploaded file as an attachment in this approval as well.

We’ll now save our flow and test it with an example receipt PNG.

The file used in this test is displayed below for reference:

Reviewing Our Results

When our flow finishes running, we’ll receive a detailed approval request in our email inbox outlining every field name and field value from our input receipt. We can compare this with our example receipt to understand how our data is parsed.

As we can see, each field was correctly parsed and shared within the approval message.

Of course, receipt fields are only half of what the Extract All Fields and Tables of Data from a Document using AI action extracted from our receipt. We’ll find the parsed table of receipt items by opening and reviewing the entire response object in our flow.

In the above screenshot, we’ve highlighted the TableResults array. This array returns data for each item in our receipt relative to the original table headers and values.

Document AI fully parses all the relevant information provided on any receipt and structures that data cleanly and accurately in its response object. This drastically reduces the amount of work we need to do to correctly parse information from receipts in Power Automate approval workflows.

Conclusion

In this article, we learned how to easily extract fields and tables from a receipt using the Document AI Extract All Fields and Tables of Data from a Document using AI action. Our flow correctly returned each relevant field from our example receipt, along with the description, quantity, rate, and amount of each receipt item.

Including Document AI in our Power Automate flows immediately empowers our document processing capabilities. We can easily extract and subsequently share or store the data we need from receipts without needing to know anything about the structure of the receipt ahead of time.

For expert advice on using the Cloudmersive Document AI API in Power Automate, please reach out to a Cloudmersive representative.