Skip to Content
Product Information
Author's profile photo Tomasz Janasz

Intelligent finance invoice processing with SAP AI Business Services

As part of the Hyperautomation Webinar Series, I presented a session how users can combine different AI Business Services to provide instant business value to enterprises. This blog post presents the key content for you to take away.

Reducing the burden of business document processing

Many processes in enterprises still involve handling business documents — for example, supplier invoices in accounts payable or customer orders in sales, and so on. Even though there are various ways to electronically exchange information, large numbers of documents are still processed manually.

The challenge is to automatically transform unstructured documents into structured information and embed it directly into business processes to generate instant business value. This is exactly the definition of the value proposition of AI Business Services related to Business Document Processing.

Business Document Processing typically consists of three steps (see also figure 1):

  1. Classifying: When document arrives, you need to understand what type of document it is in order to know what to do with it.
  2. Extracting business-relevant information.
  3. Enriching extracted information with additional master data and posting it to the system of record for further processing.

Figure 1: Business Document Processing with AI Business Services.

The business value can be seen in:

  • increased automation rate, meaning less human involvement in the process,
  • higher accuracy by leveraging the latest machine learning algorithms,
  • increased processing speed and operational efficiently.

Implementing such processes, companies can achieve significant cost reductions by reducing the need for human labor (for more information about the related AI Business Services you can read my previous blog post: Simplify Business Document Processing with SAP AI Business Services).

Business Scenario: Supplier Invoice Processing

The business scenario at hand refers to Supplier Invoice Processing in Accounts Payable (AP). AP is the most time and labor-intensive finance and administration function. In fact, processing costs per invoice range from 5 to 12 US dollars, and roughly 62% of those costs come from manual processing (Driscoll, 2015). Often what we observe with our customers is that invoices arrive without a reference document (such as a purchase order number), causing AP departments to invest lot of time manually inserting the required information for correcting or completing the invoice or trying to determine the correct cost object (such as General Ledger account). It is a highly repetitive and cost-intensive task.

AI-enabled components on SAP Business Technology Platform can provide significant support here:

The optimized process is as follows (see also figure 2):

  1. An invoice arrives in the system (via email or web interface) and data is automatically extracted.
  2. The invoice is automatically enriched with supplier data from the corresponding master data records.
  3. A customized ML-model predicts a G/L account and the cost center.
  4. The final version of the invoice is created in the system of record for further reconciliation processes.

Figure 2: Intelligent finance invoice processing with SAP AI Business Services.

The benefits are:

  • Accelerated information extraction from unstructured documents,
  • Correct G/L assignment choices,
  • Reduced redundancy and effort spent on post-processing account corrections,
  • Enhancing Financial reporting thanks to minimized account discrepancies,
  • More time for teams to focus on strategic tasks such as optimizing payment strategies.

If you are interested in gaining a greater understanding of the business process and the integration architecture, I recommend reviewing the recording from my session (30 minutes), which can be found on SAP Technology’s official YT-channel.

For further information please refer to following assets below:


Blog posts

Also check out SAP Community Page for further information and the latest content: SAP AI Business Services.



Driscoll, Mary (2015): Metric of the Month: Accounts Payable Process Cost, URL:, retrieved: September 02, 2021.

Assigned Tags

      You must be Logged on to comment or reply to a post.
      Author's profile photo Khyati Patel
      Khyati Patel

      Dear Tomasz,

      Per the below blog and SAP User Guide, a new feature was released to create templates based on sample document files.

      The "Template" icon appears only for few seconds in the left panel when loading the DOX application. It then disappears. I am using google chrome.

      Also, per the user guide, I have the Document_Information_Extraction_UI_Templates_Admin Role assigned.


      How to get the "template" icon on the left panel of the DOX application?


      Author's profile photo Tomasz Janasz
      Tomasz Janasz
      Blog Post Author

      Hi Khyati,

      please make sure you have activated the correct service plan: blocks_of_100. The „default“ is not enabled. The service plan can be chosen when selecting the service on the BTP.



      Author's profile photo Khyati Patel
      Khyati Patel

      Dear Tomasz, thank you for your prompt response. Appreciate it. I think I am using "default". I will try with the blocks_of_100.

      Author's profile photo Khyati Patel
      Khyati Patel


      I had given it a try with Edge Browser too.

      Author's profile photo Khyati Patel
      Khyati Patel

      Dear Tomasz,

      The templates worked with the "blocks_of_100" Service Plan.

      I had activated the DOX trial service through the "booster" option. It was taking the "default" plan by default. It was not offering me the option to choose a service plan. I am not sure if I was doing it incorrectly or that is how the "booster" option is designed.

      I activated the DOX trial manually; this way I was able to choose the "blooks_of_100" plan.

      Thank you so much for your prompt assistance, I would not have figured that out myself.

      Thank you again for your time. I really appreciate it.



      Author's profile photo Paul van Rouwendaal
      Paul van Rouwendaal

      Hi Tomasz Janasz, me and my colleague found this blog very interesting! We still have a few questions. Would you be able to help us on getting some answers to these?

      Question 1: In the video, you are showing an integration process flow for the entire process (from receiving e-mail to calling the various APIs). Could you tell us which service in the SAP BTP you are using for this integration? Any blog recommendations about this?

      Question 2: regarding the Document Extraction Services. We would like to use it for custom documents, e.g. sales orders. We tried to set it up in our trial account, but it looks like this is not using any ML or AI techniques for recognition. We could only learn the OCR by point and click, which does not work that well. Are there any blogs or other documents available on how to build custom OCR using AI and/or ML techniques?

      Many thanks and kind regards, Paul

      Author's profile photo Tomasz Janasz
      Tomasz Janasz
      Blog Post Author

      Dear Paul,

      thank you for getting back to me. I am happy to answer your questions.

      Q1: this particular demo was built based on Cloud Integration, selected SAP AI Business Services, and Connectivity Services. This architecture allows to build custom iFlows to connect to email servers, to integrate SAP AI Business Services, or to communicate with an SAP S/4HANA or SAP ECC in a secure way. Please start with these blogs, which also provide some relevant best practices partially related to the use case:


      Q2: Document Information Extraction service supports Purchase Order extraction (with some limitations regarding language and country support). It is based on a global ML model. If you want to extend the service to a custom document type you can use the Templating feature. This feature is rule-based (no ML), unless you want to combine it with the existing standard model (e.g. for Purchase Orders. In the latter case the defined template is combined with the global ML model as a fallback scenario. Please note: Templating can only be applied to one template, it is not generalizing on unseen layouts.

      • Supported document types:
      • Extension Capabilities (please note: for template creation it is always required to have a schema):
      • Tutorial on Schema Creation:
      • Tutorial on Template Creation:

      I hope this clarifies your concerns.

      Best regards,

      Tomasz (from Product Management)