Extending SAP with AI in just a few hours: GenAI Accelerated in action

Every summer, the operating temperature in the customer service department at Aspen Pumps rises, even when the sun is not shining at the company’s headquarters in Hailsham, UK. This is because the manufacturer’s products are particularly popular in summer. Millions of its pumps are used in air conditioning systems around the world to remove the condensation that forms when indoor air is cooled.

Be it a hotel room, warehouse or event venue – when air conditioning technicians call Aspen Pumps to resolve technical problems, they need quick answers: Do you have a replacement product for this part? Do you have anything to fit this description? The support team at Aspen Pumps must immediately clarify which model is suitable, which version is needed, and which retailer has it in stock. “In the past, first-level support staff referred lots of problems to experts,” explains Jon Curtis, Group Business System Consultant at Aspen Pumps. “Customers were transferred and had to ask their questions several times. That’s quite annoying for a technician when hotel managers and event organizers are standing next to you and putting pressure on you.”

Quick answers to many questions

Meanwhile, Aspen Pumps’ experts were frequently interrupted from more valuable work to answer the same questions over and over again. Optimizing these processes was important to Curtis, ERP and systems expert, which is why he was immediately enthusiastic when long-standing partner NTT DATA Business Solutions developed the GenAI Accelerated tool for SAP Business Suite. “This partnership has never been exclusively about SAP technology, but always also about innovation in business processes,” emphasizes Mark Wheeler, Global Head of Product Engineering & AI Customer Success at NTT DATA Business Solutions.

“Processing customer inquiries faster, reducing waiting times for customers, and using internal resources more efficiently – these arguments gave me a solid basis for drawing up a proof of concept for an AI assistant in the service team,” recalls Curtis of Aspen Pumps. The manager recognized the potential of GenAI in conjunction with SAP data early on and pulled out all the stops to exploit it fully. It later became apparent that the AI assistant also simplifies and speeds up the training and induction process for new customer advisors. The assistant also delivers a call to action at the end of each chat that matches the respective request. “This lets customers know immediately where they can buy the recommended products. That saves the technicians time, is good for our turnover, and boosts customer relationships,” says Curtis.

Rapid implementation

The proof of concept (PoC) itself saved time: “It just happened very, very quickly,” says the Aspen manager. “Mark presented the technology to me in 20 minutes and the subsequent discussion of the setup for the PoC took a similar amount of time.” Aspen’s in-house consultant then created a data extraction from the SAP system in just a few minutes. By the third meeting, Mark’s team had already delivered a functional demo version. “The effort required for the PoC at Aspen Pumps therefore took less than two hours of my time,” says Curtis.

Automatic connection of SAP and LLM

The plug-and-play connector, which connects the existing SAP solution with any large language model (LLM), made this rapid process possible. “Manual integration is completely unnecessary,” emphasizes Wheeler. Ready-made templates for the retrieval augmented generation (RAG) process speed up the grounding, testing, and deployment of the AI assistant. This meant the project team was able to focus on the content design of the assistant right from the start. “At Aspen Pumps, we actually configured the AI model based on the SAP product catalog in a few minutes so that it provided useful answers,” reports Wheeler. Jon Curtis adds: “Our CEO was impressed after the rigorous testing – he immediately saw the possibilities and approved the necessary funding.”

For all his enthusiasm for the rapid results, Curtis had one crucial issue on the agenda from the outset: “It was clear to me that with AI, as with any new technology, our management wants to know exactly how secure it is.” The SAP AI Core service, part of the SAP Business Technology Platform ensures end-to-end encryption, content filtering, and the guarantee that no third-party models will be trained on your data, ensuring the security and compliance of the intelligent assistant at Aspen. Mark Wheeler explains: “Trust in AI depends on absolute privacy. That’s why the architecture of GenAI Accelerated gives customers control over every byte of their data.”

Flexible use with different LLMs

Another advantage of GenAI Accelerated is its flexibility. Wheeler initially used OpenAI’s GPT-4 model for Aspen’s AI chatbot. In the tests with Curtis, however, the Claude model developed by Anthropic proved to be better suited for more complex questions. “Being able to easily switch from one model to another without changing anything in our system is a key factor in getting the most out of our investment in SAP and AI,” explains Curtis. After all, GenAI is a young technology that is developing rapidly. “Thanks to GenAI Accelerated, we will be able to continue to use the best model for the job with all our SAP processes immediately.”

Improving efficiency in customer service and order management

Aspen harnessed the versatility of GenAI Accelerated to reprioritize their automation goals. The background to this is the company’s promise to its retail partners: Every order received by 3 p.m. will be delivered on the next working day. This means that hundreds of orders often have to be processed in a short time. An incorrect product name, transposed digits in the item number, or another incorrect entry quickly result in errors that can be costly if they lead to returns and a drop in customer satisfaction. “That’s why we are now using the AI model first to compare customer orders with the product data in SAP,” says Curtis.

Which SAP process will we improve next with AI?

The next step is to extend the AI Assistant for the customer service department. And that’s just the beginning, agree Curtis and Wheeler. They have actually already talked about rolling out the solution in other countries and company divisions, says Wheeler. “Our experience with numerous companies has shown that bringing SAP applications and AI technology together opens up innovation potential that no one has seen before.” The question is then no longer, “What should we do with AI?” but “What core SAP process shall we improve first with AI?”

GenAI Accelerated enables the seamless integration of SAP systems with common large language models (LLMs) such as OpenAI, Anthropic, and others. The most important functions and features include:

• Secure connection to third-party LLMs
• Applying AI to trusted business data
• Agent capability with LangChain
• Suitable for pro-code and low-code teams
• Development of AI chatbots in minutes
• Integrates more than 40 major third-party language models
• More than ten SAP S/4HANA Cloud AI solutions available as validated partner use cases