Order Process Automation
From an Email with a PDF to an Order in SAP
What will you find in this article?
In this article, we show how automation can improve the processing of orders received by email. Using an implementation based on Microsoft Power Automate, Outlook, SharePoint, Azure Document Intelligence, and SAP, we walk through the entire process: from receiving a PDF attachment, through data extraction and validation, to automatically creating an order in an ERP system. We also show where human oversight remains part of the process and how Confidence Score is used to route records either for automatic processing or manual review.
Automation
Automation replaces manual, routine tasks with an organized sequence of actions performed automatically or semi-automatically, with human involvement where needed. No-code and low-code tools such as Microsoft Power Automate make it possible to connect applications and systems and build workflows without having to develop an entire solution from scratch.
Flow
An automation process consists of a sequence of actions that are initiated when a specific event, called a trigger, occurs. This organized sequence of actions is referred to as a workflow or flow.
Traditional Order Processing
why is it worth automating?
Order processing is an important part of many business processes. In a traditional model, PDF documents are reviewed one by one, and order information is manually transferred to an ERP system such as SAP. With a large volume of documents, this way of working becomes time-consuming and costly.
Manual processing requires sustained attention and repeated execution of the same tasks over long periods of time. This increases the risk of errors, and an incorrectly created order can lead to costly consequences later in the process. Automation helps reduce the amount of work involved in manually reading and re-entering data, leaving people to focus mainly on cases that require verification or a decision.
Order Automation
from an email attachment to SAP-ready data.
Information contained in an email attachment does not have to be manually re-entered into an ERP system. Combining Power Automate with OCR (Optical Character Recognition) and intelligent document processing technologies makes it possible to read the contents of a PDF file, transform unstructured information into structured data, and pass it on to the next stages of the process.
In the implementation described here, the document goes through a controlled process of data extraction, quality assessment, and storage in SharePoint. If it meets the defined criteria, the data is then sent to SAP. As a result, some orders can be processed entirely automatically, while cases with a lower recognition confidence are routed for human review.
How Does Order Process Automation Work?
OCR – What Is It and What Is It Used For?
OCR enables automatic text recognition in images and documents. In Azure Document Intelligence, dedicated data extraction models are available, including the Read model and the Layout model. The Read model recognizes printed and handwritten text in PDF files and images, including JPEG/JPG and PNG. The service also supports text extraction from Microsoft Word, Excel, PowerPoint, and HTML documents; in these formats, it reads text embedded in the document.
In practice, intelligent document processing goes beyond character recognition alone. With the Layout model, it is also possible to identify the full structure of a document and extract elements such as fields, tables, selection marks, headings, and key-value pairs. This makes it possible to transform a source document into structured data that can then be processed by an automation workflow.
This approach speeds up document digitization and analysis, reduces manual data entry, and lowers the risk of errors caused by manual processing.
Document Processing Tools in the Microsoft Ecosystem
The choice of solution depends on the complexity of the process, customization requirements, and the way it needs to integrate with other systems. Within the Microsoft ecosystem, organizations can use both low-code tools available in Power Platform, such as AI Builder, and Azure Document Intelligence in Foundry Tools, which is part of Microsoft Foundry.
AI Builder Document Processing is built on Azure Document Intelligence technology, formerly known as Azure Form Recognizer. This means that both approaches use the same family of Microsoft technologies for intelligent document processing. The main differences lie in the level of abstraction and the degree of control available to the person building the solution.
1. AI Builder – Low-Code Approach
AI Builder is part of Microsoft Power Platform. It enables the use of prebuilt models as well as the creation of custom models for document processing without having to build the entire integration layer in code. Power Automate provides actions for tasks such as text recognition and document processing, while the models can support scenarios involving invoices, forms, orders, and other business documents. The AI Builder ecosystem also includes content classification models that can be used as part of a process for classifying documents based on their content.
2. Azure Document Intelligence – Pro-Code Approach
Azure Document Intelligence in Foundry Tools is a cloud service within Microsoft Foundry designed for automated document processing using machine learning models. It can be used with prebuilt models or with custom models tailored to specific document types and business processes.
SAP Integration and Automated Order Creation
Case Study
In the project described here, the goal was to automate the processing of orders received as PDF attachments. The solution architecture was built on the Microsoft platform using Power Automate Cloud, Outlook, SharePoint, and Azure Document Intelligence. The final stage of the process is to send the verified data to SAP and automatically create the order.
Key Process Assumptions
1. Trigger: Power Automate monitors a designated email inbox. When a message with a PDF attachment is received, the automation process is triggered.
2. Data extraction and structuring: data from the PDF file is extracted using Azure Document Intelligence, evaluated for quality, and stored in a SharePoint list before being sent to SAP.
3. SAP integration: a record approved for further processing triggers the integration stage. The data is retrieved, mapped to the format required by SAP, and sent to the system through BAPI (Business Application Programming Interface).
The process was divided into two separate Power Automate flows. The first handles document receipt, data extraction, evaluation, and storage, while the second prepares approved records and sends them to SAP. This separation makes it possible to verify data quality before performing an operation in the ERP system.
Data Verification and Approval in Power Automate
A key element of the solution is combining automated processing with human oversight. Records with a sufficiently high recognition confidence can proceed to the next flow automatically, while records below the defined threshold are held for manual review and approval by a designated person.
Flow 1: Data Extraction and Processing
The first flow is triggered when an email with a PDF attachment is received in the monitored mailbox. The attachment is sent to Azure Document Intelligence through an API and then analyzed to extract the data required for the subsequent stages of the process.
The project also uses Document Intelligence Studio, a Microsoft environment for working with Document Intelligence models. The flow itself communicates with the service programmatically through an API, allowing the document to be submitted for analysis and the structured results to be returned.
API
An API (Application Programming Interface) is an intermediary that enables communication between different applications, systems, or devices operating in the cloud. It acts like a “translator” by defining how requests are sent and responses are received. This allows systems to exchange data securely regardless of differences in their architecture or underlying technologies.
The extracted data is then evaluated for quality. For recognized elements, the model returns a Confidence Score (CS), a numerical value representing how confident the model is that the result is correct. The score ranges from 0 to 1: the higher the value, the greater the model’s confidence in a given element.
In this specific implementation, a threshold of 0.85 was defined as a business rule controlling the next stage of the process. Data with a score of 0.85 or higher is added to the list of records approved for SAP and can automatically trigger the second flow. Data with a score below 0.85 is added to a list of records requiring manual review through a dedicated Power Apps application. Only after the data has been reviewed and approved by a designated person can it proceed to the next stage.
This mechanism keeps the process automated for results with a high confidence level while maintaining human oversight in cases where the model does not reach the defined threshold.
Flow 2: Creating Orders in SAP
The second flow handles the final stage of the process. It retrieves data approved for further processing from SharePoint and maps it to the structure required by SAP.
The data is then sent to SAP through BAPI (Business Application Programming Interface). BAPIs are standardized SAP interfaces that allow external systems to call specific business functions. It is also worth noting that Power Platform supports calls to both BAPI functions and SAP RFCs (Remote Function Calls) through the SAP ERP connector, allowing this type of integration to be implemented without manually entering the same data into SAP.
As a result, an approved order is created automatically in SAP. Employees do not need to re-enter data from the source document; their involvement is primarily required for cases that were previously routed for manual review.
What Business Benefits Did This Solution Deliver?
Summary
Business process automation makes it possible to transform repetitive tasks that require consistency and accuracy into controlled workflows that reduce the amount of routine manual work. In the case of order processing, this means automatically receiving a document, extracting and structuring the data, assessing the quality of the result, and sending approved information to SAP.
However, the goal of the solution is not automation at any cost. In the implementation described here, Confidence Score is used to separate cases that can be handled automatically from those that require human review. This allows the organization to improve efficiency and reduce repetitive work without giving up control over data quality.
This approach is one practical element of digital transformation: it helps organizations make better use of resources, increase operational flexibility, and focus employees’ time on tasks where their knowledge, experience, and attention provide the greatest value.
See Where Automation Can Support Your Organization
Contact us.
Let’s talk about the processes worth automating, the documents your organization works with, and the systems they need to connect to. We can help assess where Power Automate is sufficient, when Azure Document Intelligence is the better fit, and how to plan a secure integration with an ERP system such as SAP.
Choose a date for a phone consultation.
