Discover Document Actions and Document Processing Automation

Learning Objectives

After completing this unit, you’ll be able to:

Explain how document actions streamline data extraction.
Utilize prompts to refine the extraction process.
Identify the steps necessary to automate document processing.
Explain when human interaction is needed in the IDP process.

Document Actions

As explained in the previous unit, document actions are a key component in extracting data from documents using IDP. Let’s dig into more details of a document action and how it can benefit your workflow.

A document action is a multistep process that enables the conversion of unstructured and semistructured documents into a structured format. The main components of a document action are a data model definition that specifies the relevant fields to extract from the document, an AI model for field extraction, and an API that enables programmatic execution of the action.

By creating document actions, you can define which fields are mandatory, exclude certain fields from the JSON response, and set minimum confidence scores for each field.

A document showing information being extracted and organized.

The purpose of a document action is to streamline the process of extracting data from documents, making it easier and more efficient for your applications to access the information they need. Whether you are looking to extract specific numbers, dates, or other data points from a document, document actions provide a structured approach to achieve this goal.

Example use cases for document actions include extracting the subtotal amount, grand total, due date, or highest price from a document. These specific data points can be easily identified and extracted using document actions, saving time and effort in manual data extraction processes.

Prompts

In addition to defining fields and confidence scores, document actions also support prompts in natural language. Prompts enable you to ask specific questions about the document, such as, “What is the subtotal amount?” or “When is the due date?” These prompts help refine the extraction process and ensure that the desired data is accurately extracted.

The purpose of prompts is to provide a more user-friendly and intuitive way to interact with document actions. By asking questions in natural language, you can quickly retrieve the information you need without having to navigate complex schemas or configurations.

For example, you can analyze invoices and purchase orders by creating prompts such as:

What is the total amount?
When was the document created?
Show me the customer name.
What is the payment status?

The IDP Journey

Now that you are familiar with the main IDP concepts, it’s time to explore the journey of using IDP to automate document processing. The IDP journey consists of five key steps: create, publish, execute, review, and retrieve.

Let’s dive into each step to understand how you can effectively utilize the IDP solution to automate the extraction of data from invoices and receipts so that the accounts payable analysts in your company don’t have to manually read and enter this data into the system.

Step	Detail
Create	The first step in the IDP journey is to create document actions, which contain the configuration needed to process documents effectively. You can define the output schema, prompts, and reviewers to ensure accurate results.
Publish	Once the document actions are created, you can publish them to Anypoint Exchange. By publishing the document actions, they become available for consumption through the IDP API, which enables you to process documents programmatically from MuleSoft RPA, Mule applications, or other systems.
Execute	After document actions are created and published, you can start consuming them by calling the IDP API. During this step, an RPA process or an application sends the documents to process to the IDP API and triggers the necessary document action to extract valuable information from the documents efficiently.
Review	In some cases, after document execution, a document may require human intervention for review. If a document is missing any required field or the extracted values have confidence scores lower than the minimum allowed, it goes into a review queue. A user needs to manually review the document in IDP and submit it after verification to ensure accuracy.
Retrieve	Finally, after a document extraction finishes successfully, or an extraction queued for review is reviewed and submitted, the results become available for consumption. You can retrieve the results by calling the IDP API, which retrieves the JSON response with the extracted information that you can use for further analysis or processing by integrating with other applications.

Wrap It Up

In this module, you learned the benefits of IDP, how it works, and the steps of the IDP journey: Create, publish, execute, review, and retrieve. Now you can effectively use the capabilities of the IDP platform to streamline document processing workflows and extract valuable insights from documents. Start your IDP journey today and unlock the power of intelligent document processing.