Technology Blogs by SAP
Learn how to extend and personalize SAP applications. Follow the SAP technology blog for insights into SAP BTP, ABAP, SAP Analytics Cloud, SAP HANA, and more.
Showing results for 
Search instead for 
Did you mean: 
Product and Topic Expert
Product and Topic Expert
Machine Learning has seen major advances over the last few years, as demonstrated by the recent success in redefining our day-to-day lives by using intelligent algorithms. Whether we are talking about Digital Assistants, Social Media Services or Commuting Systems, machine learning has done a pretty good job of improving our lives by detecting and recommending things that truly matter. Having the ability to extract valuable insights from large amounts of data can be very beneficial for companies in order to optimize their business processes that involve a lot of unstructured information.

Recently, SAP launched an entire portfolio of intelligent applications that provide generic machine learning capabilities which can be applied to several business processes. Those services are available on the SAP Business Technology Platform as SAP AI Business Services. For more information regarding SAP’s AI Business Services, please check out this blog from SAP Product Management.

Since these AI services are mostly reusable stand-alone capabilities, you may ask yourself, “How can I connect machine learning capabilities into my day-to-day processes to achieve actual business outcomes?” If you don’t have the ability to connect these outcomes to the productive systems of your business through Integration Layers, Workflows and Data Pipelines, then machine learning algorithms mean nothing.

In this blog, I will introduce you to a proven and tested scenario of integrating machine learning capabilities with core business processes to manage incoming customer orders.


What is the Business Problem?

The main problem this scenario addresses is that companies often receive customer orders in various types of formats. Customers tend to use different templates or don’t have a unified document structure, which makes it difficult for the receiving company to detect relevant data from the order. Therefore, extracting the necessary information is mostly done by humans, which is a root cause for process inefficiencies and errors.



How does it work?

The solution relies on an integrated architecture of separate services on the SAP Business Technology Platform, in which each service is responsible for a different task. The scenario starts with a customer that wants to place an order. The customer creates a sales order and sends it as an e-mail attachment to the company. SAP Business Technology Platform Integration Suite is the leading system for orchestrating this entire process, integrating to different systems and connecting to the necessary services. The out-of-the-box adapters scan for incoming e-mail attachments on the company’s mail-server and make them available for processing.




Step 1: SAP Cloud Platform detects the format of the e-mail attachment. If the format is other than PDF, CPI will automatically send an e-mail to the customer asking to re-send the order in the correct format.

Step 2: If the format is correct, CPI will call the SAP Document Information Extraction Service. This AI Business Service is pre-trained, pre-configured and can be called via RESTful API. After extracting all the necessary entities in a JSON-Format, CPI receives all entities for further processing.

Step 3: In order to process the order, the system first needs to know if the detected material number is available, as many customers tend to use their own unique material numbers. By leveraging the API’s available on the API Business Hub, CPI can now use this information and check if the material is available in the master data.

Step 4: In order for customer support to put focus on the most critical customers, the company may want to have priority orders be doublechecked by a human agent. . These thresholds and decision tables are not programmed within CPI but are instead externalized with the SAP Business Rules Service.

Step 5: Each order that was declared by the rule service to be doublechecked triggers a workflow to the internal sales staff to review the order. This workflow is developed and maintained with the SAP Workflow Service for user-centric process steps. The workflow can be integrated natively with the centralized inbox applications for the sales staff to review. The sales staff can then decide whether or not this order is approved.

Step 6: If the sales order is approved, CPI will automatically trigger a sales order confirmation in the underlying SAP S/4 HANA System and will send the order confirmation back to the customer via e-mail.


What’s the technology?

SAP Cloud Platform is an enterprise platform-as-a-service (PaaS) that provides comprehensive application development services and capabilities, which let you build, extend and integrate business applications in the cloud. As a key component of SAP’s vision for the Intelligent Enterprise, SAP Cloud Platform provides flexibility and agility for developing applications, for both customers and employees, that have high impact on processes to ultimately deliver superior business outcomes.

The underlying foundation of SAP’s proposed architecture is managed and orchestrated by SAP Cloud Platform. The platform provides multiple services, such as AI Business Services, which are necessary to realize the scenario.




Document Information Extraction extracts structured information from unstructured documents. The extracted information can be automatically enriched with your existing structured master data and transactional data. Document Information Extraction helps to drastically reduce manual efforts by automatically extracting structured information from unstructured business documents. By reducing manual effort, the overall document processing efficiency can be increased, while also reducing the error rate.

SAP Cloud Platform Integration (CPI) allows you to simplify integration by connecting people, processes, data, and devices, while supporting a variety of integration approaches and out-of-the-box features. Since the scenario requires a connection to various systems and services, CPI is the ideal layer for integrating and orchestrating the entire end-to-end process.

SAP Cloud Platform Business Rules is a service that enables a cloud application developer to embed decisions into cloud extension and workflow applications. The developer can expose the decision logic to different business users and knowledge experts, enabling them to directly author and influence the logic. Companies can reuse these business rules for any process to provide a centralized enterprise rule framework.

SAP Cloud Platform Workflow lets you build, run and manage workflows from simple approvals to end-to-end processes that span across organizations and applications. With an inbox app and custom-built UIs, companies can involve end users into business processes for decision making and data entry. The service comes with web-based tools for workflow modeling, APIs for consumption in custom apps, monitoring tools, and Fiori-based apps for end-user access.



Now you know how to build your own intelligent processes using  SAP Cloud Platform. In this scenario, we developed an integrated process involving different services to automate order entry. Please note that this is not a standard Software-as-a-Service (SaaS) solution that is provided out-of-the-box. It is an integrated scenario of standardized business services that can be used in a modular approach, like building with LEGO bricks. The key advantage of this setup is that as a customer, you now have the ability to implement technologies that are highly contextual to your specific business needs. Once you identify the process flow from a business perspective, you can easily pick and choose the necessary components and connect them seamlessly via SAP Cloud Platform.

If you want to learn more about how artificial intelligent is influencing SAP’s solution portfolio – please check out this blog from SAP Product Management.


 Credits to svenhuberti, tomasz.janasz and tim.nusch who participated at this project!