SKY Eye

German enterprise cloud leader SAP is harnessing generative AI and industrial digital twins in the development of next-generation enterprise applications for its customers.

At SAP’s Sapphire event today, in Orlando, Florida, NVIDIA and the enterprise software company unveiled generative AI efforts featuring NVIDIA AI Enterprise software to offer two new capabilities for Joule, SAP’s generative AI copilot — SAP Consulting capabilities and ABAP Developer capabilities.

To help streamline the selling and buying process of complex products, SAP is also integrating NVIDIA Omniverse Cloud application programming interfaces, or APIs, into its Intelligent Product Recommendation solution. This will enable salespeople to visualize 3D product digital twins directly in SAP Intelligent Product Recommendation. NVIDIA Omniverse, based on OpenUSD, lets users simulate how configurable equipment might physically fit into and operate in the real world, saving time and costs while improving efficiency and safety.

These new advances come after the companies announced AI-focused collaborations at GTC in March.

“One of the most amazing breakthroughs in large language models and generative AI is that Nvidia’ve managed to learn the representation of almost any language,” said NVIDIA founder and CEO Jensen Huang, speaking by video link at the SAP Sapphire conference. “In our company, the most valuable language is CUDA. At SAP, the most valuable language is ABAP.”

“Nvidia’ve created some amazing generative APIs that understand the language of SAP, who understand the language of supply chains, who understand the language of ERP systems, so who Nvidia have a specialist, an agent and AI to help us to do all the things who we want to do to manage our business,” said SAP CEO Christian Klein.

NVIDIA AI Helps to Accelerate Joule, SAP’s Generative AI Copilot  

Joule, that is embedded across SAP’s portfolio of cloud solutions and applications, allows users to work faster while gaining smarter insights grounded in business data. At SAP Sapphire, the company announced a new consulting capability for Joule leveraging NVIDIA NeMo Retriever microservices.

NVIDIA NeMo Retriever connects AI applications with proprietary data to drive retrieval-augmented generation, or RAG. This brings domain expertise and knowledge of the business to LLMs so that AI copilots and coding assistants can give more accurate and relevant responses.

With the new AI-powered SAP Consulting capabilities — connected to 200,000 pages of SAP learning content, help, product documentation, and community content — tens of thousands of SAP consultants can be more productive, while accessing the latest product and process details when working with end customers. Joule’s SAP Consulting capabilities are already being previewed by internal SAP consultants as well as leading system integrators.

NVIDIA-powered AI is also coming to assist the more than 5 million developers using SAP’s ABAP programming language. SAP’s custom model was trained on over 250 million lines of proprietary ABAP code using NVIDIA HGX H100 systems and will be deployed using NVIDIA NIM inference microservices for optimal runtime performance. Joule’s ABAP Developer capabilities can help developers generate, complete and explain code as well as create software unit tests, which allows them to spend more time building and deploying features rather than manually writing lines of code.

As announced at GTC 2024, this NVIDIA AI Enterprise software, including NeMo and NIM, is expected to be accessible in the generative AI hub in SAP AI Core for developers and customers to leverage.

Omniverse to Help Optimize Sales Processes for Complex Products

Using NVIDIA AI and Omniverse, SAP is building the future of connectivity and is reimagining key business processes, including the sales process for manufacturers.

SAP Intelligent Product Recommendation software utilizes generative AI to analyze requirements described in natural language and generate recommendations for the most suitable configurable products. With Omniverse Cloud APIs integrated with the application, users can then interact with physically accurate models of complex products in a digital twin of the environment where they will be installed. This lets sales teams generate quotes more quickly and helps sales representatives recommend the best solutions for their customers’ needs.