PoC with SAP Business Network Purchase Order PDF with Document Information Extraction (AI Services)

By /

Given my background in Ariba and AI, I decided to combine both skills to make a very simple PoC combining SAP Ariba, SAP Business Network and SAP BTP’s SAP AI Service.

In this article, you will find a use case where a buyer has generated a purchase order in SAP Ariba and sent it to the supplier via SAP Business Network. After receiving it in SAP Business Network, the supplier needs to extract the PDF information from this purchase order in order to generate a JSON message to integrate into their sales system (e.g. Salesforce).

Here are the steps:

Step-by-step

Giving Ariba context

1. Through SAP Ariba Buying platform, buyer creates a PR that is converted in Purchase Order after all approvals:

VictorSevilla_0-1716974553797.png

2. The supplier has chosen SAP Business Network as the method of receiving purchase orders. The supplier receives this PO via SAP Business Network including PDF smartform:

VictorSevilla_1-1716974607945.png

VictorSevilla_2-1716974642775.png

VictorSevilla_3-1716974653881.png

Using Document Information Extraction (SAP BTP AI Service)

The following steps will then continue with the recognition of the fields of this PDF through the SAP AI Service of Document Information Extraction.

3. Through Document Information Extraction (SAP BTP), we generate the schema of the document to be recognized. In this case, we will use a pre-built schema provided by SAP.

VictorSevilla_4-1716974742146.png

4. Once the schema has been defined, the PDF is uploaded to the application.

VictorSevilla_5-1716974769329.png

5. After pressing ‘Confirm’ tile, the system starts with the recognition of the fields and the generation of the information through GenAI.

VictorSevilla_6-1716974787052.png

5. Finally, the system processes PDF and we extract the results:

VictorSevilla_7-1716974866607.png

Human-in-the-loop is very important to verify that everything is correct. To facilitate this action, the system returns the percentage of possible error through the use of traffic lights (red, orange or green):

VictorSevilla_8-1716974912441.png

Additionally, DIE is able to generate a document in different formats (text, json or csv) that we could use to integrate this order in a SAP or non-SAP application (e.g. Salesforce):

VictorSevilla_9-1716975005421.png

Conclusions

After this short PoC, we can see the potential of artificial intelligence in SAP through SAP BTP. Here are my thoughts:

  • The pre-build schema represents a very important back-bone to start a more robust module for our clients.
  • The use of the artificial intelligence tool is simple and intuitive.
  • The application has been very easy to configure thanks to the SAP BTP boosters.
  • The system allows us to customise any type of document we receive, so its flexibility and adaptability is very high.
  • The extraction of the information in different formats allows the automation of future steps, such as, for example, the creation of an output of a warehouse, sales order, invoice, etc.

We hope you find it interesting how we can combine the different solutions of the SAP offering so that through AI we can automate our processes and humans can perform tasks that bring more value to the business.

Scroll to Top