How SAP Customers are using ChatGPT: Blog 2 of Series “ChatGPT and SAP”

Welcome to the second part of our fascinating journey into the vast expanse of ChatGPT and its integration with SAP systems. If you’re just tuning in, in our first blog post, we ventured into the labyrinth of ChatGPT, exploring its inner workings and probabilistic AI model, and even delved into its inherent limitations. We shone light on the complexities of this innovative technology, attempting to simplify it for our readers.

Today, we take it a step further. In this installment, we dive deeper into the practical applications of ChatGPT, specifically within SAP’s vast and diverse client base. We spotlight the victories and troubles, the eureka moments and the stumbling blocks, and the insights gathered from pushing the boundaries of AI technology in a real-world business context. Hold tight as we navigate through this exciting intersection of artificial intelligence and enterprise software.

How SAP Customers are using ChatGPT

According to a recent report by TechTarget from the SAP Sapphire event (Paraphrased):

The expanding field of Generative AI technology has been creating quite a buzz in the corporate sector. However, it is still in its nascent stages and struggling with challenges related to accuracy and ethical considerations. Per TechTarget, at this point, generative AI is seen more as a novelty rather than a necessity by SAP customers. Although some customers are exploring the technology, its use is not widespread due to its novelty​. In fact, the demand for AI functionality among SAP’s customers does not appear to be high.


Vishwas Madhuvarshi using the Generative AI Capabilities of Microsoft Designer

However, the SAP landscape is far from stagnant. A plethora of innovative ideas are being proposed, experimented with, and evaluated. Notably, a majority of these exploratory cases revolve around the technical applications of SAP, while business-oriented applications are still relatively few. Here are 11 intriguing use cases that SAP customers have ventured into. Broadly, we can divide these use cases into two areas: Code generation and Applications. Let us take a look.

Code Generation

ABAP Code Development

ChatGPT, despite its self-proclaimed inability to write software, has demonstrated remarkable proficiency in generating code in response to natural language prompts. For instance, when asked to write an ABAP report based on specific parameters, ChatGPT was able to generate code faster than most developers. It could also modify the code into an object-oriented format and provide an ABAP Unit test class for it. To test this function out, I took a report description from the comment section at the top and asked ChatGPT to generate an ABAP report based on these comments. As the report is 79 lines long, I have not included it all, but a snapshot of what it was able to generate has been shown in the image below.

Database objects generation

ChatGPT can generate SQL statements on standard tables and create scaffolding for ABAP and ABAP Unit classes. While it needs detailed instructions, ChatGPT can generate accurate SQL queries. Similarly, when tasked with building a Data Access Object (DAO) class, it could generate code that could be directly copied and pasted into an Integrated Development Environment (IDE) for further development.

Accelerating everyday development tasks

Many developers have found ChatGPT to be highly effective for code documentation, typically left until the last minute. Another widespread use case is the development of functional specifications. Additionally, ChatGPT is being increasingly utilized to assist with data wrangling and process mining, which are emerging areas of application. Here is an example of generating a Functional Specification Document using ChatGPT for the ABAP report we generated in the previous example.


Vishwas Madhuvarshi: Functional Specification Document

AI-Powered ETL

Some users have explored the integration of ChatGPT, SAP HANA, and Jupyter notebooks to enhance data ETL processes in the SAP Datawarehouse Cloud. By leveraging the AI capabilities of ChatGPT, users could interactively perform ETL tasks and even generate code snippets for tasks like connecting Jupyter with HANA, creating entries in HANA, and adapting example code for new table creation.

Simplified SAP HANA Data Queries

A developer used ChatGPT to access and analyze data stored in a HANA Cloud Database. By leveraging OpenAI’s APIs, the user interacted with the database by inputting questions in natural language. OpenAI converted these questions into SQL queries, which were then executed on the HANA Cloud instance. The resulting data were further summarized by OpenAI, providing the user with live insights.

Automated Fiori App Development:

ChatGPT has shown promise in accelerating Fiori app development by generating boilerplate UI5 code and OData service bindings from natural language prompts.


Drafting Technical Specifications

Some SAP architects and developers have harnessed ChatGPT to draft technical specs and documentation from high-level requirements. Here is an example of generating a Technical Specification Document for the ABAP report we generated in our first example.


Vishwas Madhuvarshi: Technical Specification Document

Automating customer support

Customer support can be automated by leveraging ChatGPT to generate prompt and cost-effective responses to customer queries. A possible solution involves utilizing Azure OpenAI service in conjunction with the SAP S/4HANA system, specifically configuring the Standard Sales Order API and integrating it with SAP Cloud Connector and SAP API Management service.

Improving supplier data collection for ESG reporting

Supplier data necessary for ESG reporting is commonly found in unstructured electronic documents such as PDFs. Unlike standard Machine Learning (ML) models, ChatGPT offers a promising solution that doesn’t require retraining for each new document format. In addition, trials have demonstrated ChatGPT’s exceptional ability to understand the context of questions and accurately extract relevant answers from extensive text documents.

Automated Mapping Program Generation

This use case showcases the application of generative test-driven development (generative TDD) to automate program mapping in integration platforms such as SAP Integration Suite and SAP Process Orchestration. The objective is to eliminate the manual creation of mappings, enhance the efficiency of migration and development procedures, and streamline the overall process. Leveraging generative AI solutions like ChatGPT, mappings can be generated in various programming languages. However, manual validation remains crucial to ensure accuracy and suitability for the specific integration platform. The generative TDD approach necessitates a clear understanding of the desired outcomes and the intended functionality of the mapping.

Using integrated OpenAI API with SAP Commerce

One user has integrated OpenAI API with SAP Commerce Cloud to enable the automatic generation of product descriptions, product summaries, and images without the need for additional PIM (Product Information Management), CMS (Content Management System), or DAM (Digital Asset Management) systems.

We used the following image and ChatGPT plugin Pixellow to test the functionality to generate the product description and product summary.


Created by @we.are.niceday, upscaled by Vishwas Madhuvarshi using Midjourney

Generated Product Description

Experience the harmonious blend of tradition and innovation with this meticulously crafted trumpet. The instrument showcases a sleek, polished brass finish that exudes an aura of sophistication and elegance. Its design is a testament to precision engineering, featuring a seamless bell, three piston valves, and a comfortably curved leadpipe for optimal playability. The trumpet’s compact size and lightweight nature make it a perfect choice for both beginners and professional musicians. Its high-quality construction ensures a rich, warm sound that can fill concert halls and living rooms alike. This trumpet is not just an instrument; it’s a statement of style, quality, and musical passion.

Product Summary

Unleash your musical prowess with this elegantly designed, high-quality brass trumpet. Perfect for all skill levels, it promises a rich sound and comfortable playability.

To conclude, ChatGPT, in its initial stages of adoption within the SAP ecosystem, already demonstrates immense potential. This nascent relationship is not just a synergy of two advanced technologies but is shaping up to be a formidable partnership that could revolutionize how businesses operate. The examples we’ve examined, though just the tip of the iceberg, showcase a promise of even greater developments.

With every challenge overcome and every success story shared, we inch closer to fully harnessing the capabilities of this powerful AI model. The dawn of AI-enabled solutions within the SAP environment is upon us. What we’re witnessing is just a precursor to a deluge of advancements, applications, and innovations waiting in the wings. So, as we stand at the threshold of this technological revolution, one has to wonder: Are we ready for the surge of innovation that ChatGPT might usher into the world of SAP?

As we continue to uncover the myriad applications and opportunities that the synergy between ChatGPT and SAP systems presents, our journey leads us to explore a fundamental facet of any enterprise system – data. Our next blog post in this series will guide you through the riveting domain of data cleansing and visualization. We’ll dive into hands-on examples demonstrating how to sift through, clean, and visualize your data to extract valuable insights. Furthermore, we’ll delve into the captivating world of process mining and unearth areas ripe for improvement using process logs. This exploration is designed to empower you with the knowledge and techniques to elevate your SAP environment to its peak performance. So, stay tuned for an enlightening tour into the lifeblood of SAP – its data and processes.


SAP notes that posts about potential uses of generative AI and large language models are merely the individual poster’s ideas and opinions, and do not represent SAP’s official position or future development roadmap. SAP has no legal obligation or other commitment to pursue any course of business, or develop or release any functionality, mentioned in any post or related content on this website.



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


Thanks a lot for sharing your insights Vishwas Madhuvarshi! Looking forward to reading your upcoming blog posts in the series.

For those of you interested in Generative AI use cases planned by SAP, check this page.

Author's profile photo Vishwas Madhuvarshi

Vishwas Madhuvarshi

Blog Post Author

Thank you, Paul PINARD!

Author's profile photo Abraham MD
Abraham MD

An excellent blog discussing the subject. I attempted to create a Functional Specification Document for an Invoice Display ALV Report, but it doesn’t align with yours. Can you provide details about the prompts you utilized and how ChatGPT gains insights into custom SAP reports to describe their functionalities within the FS?

Author's profile photo Vishwas Madhuvarshi

Vishwas Madhuvarshi

Blog Post Author

Dear Abraham MD,

Thank you for reading the blog and taking the time to share your thoughts. I’m delighted you found it insightful.

It’s important to understand a key characteristic of generative AI models like ChatGPT. Despite being deterministic in nature, it’s highly unlikely that ChatGPT would generate the exact same response to the same question asked twice. The model’s design takes into account a myriad of factors and variables that lead to a wide range of potential responses. This is part of what makes GPT models so dynamic and effective.

Regarding your experience with creating an FSD for an Invoice Display ALV Report, the details and specificity of the prompt given to ChatGPT can heavily influence the output. If the prompts were different than the ones used in my example or less detailed, the resultant FS might differ significantly.

It’s also important to note that while ChatGPT is a highly advanced AI model, its comprehension of custom SAP reports, like the Invoice Display ALV Report you’re referring to, is derived from the input data it was trained on and not from any inherent understanding of SAP systems. Therefore, the quality of the response ChatGPT produces relies heavily on the clarity, specificity, and context provided in the prompts.

When it comes to my process, I unintentionally took an inverse approach. As seen in the blog, I first created the report. Following that, I prompted ChatGPT to create a Functional Specification Document (FSD) based on that report. As a result, the FSD is likely more detailed due to the availability of the report as a reference.

To generate the FSD, the prompt I used was simple: ‘Now generate an FSD for this report’. This instruction was provided to ChatGPT after the report was generated. It’s crucial to note that the effectiveness of the output depends on the quality and detail of the input data (in this case, the report).

I hope this clears up your queries. Let me know if there’s anything else I can help you with.