IBM Granite LLM Now Available Through the Generative AI Hub in SAP AI Core

⇨ IBM has introduced the Granite 13b.chat large language model to SAP users, enabling efficient Q&A and conversational AI applications by leveraging SAP's rich data ecosystem on the SAP Business Technology Platform.

⇨ Despite progress in AI adoption, many organizations remain cautious about integrating AI into core business functions, with only 27% currently using AI for customer service and a majority still evaluating its use in critical areas such as finance and supply chain.

⇨ Granite models are designed to be transparent, performant, and efficient, making them suitable for enhancing insights and automating workflows within SAP systems, ultimately driving productivity and operational efficiency.

IBM recently announced that SAP users can now harness the power of IBM watsonx and IBM Granite, beginning with Granite.13b.chat large language model, now available through the generative AI hub on SAP AI core on the SAP Business Technology Platform (SAP BTP). IBM and SAP have a shared approach to generative AI, built on an open ecosystem, prioritizing trust, transparency and the use of purpose-built models to help optimize business efficiency. The Granite announcement appears to be well timed, as 56% of surveyed business executives say they plan to divert spending from other budget areas to SAP-specific generative AI, according to new research from the IBM Institute for Business Value.

Granite 13b.chat is optimized for Retrieval Augmented Generation for Q&A and is commonly used to create assistants and chatbots. Tapping the rich data in SAP systems is the ideal starting point for getting value from generative AI. The combination of Granite conversational AI capabilities with SAP’s domain-specific finance, human capital management, supply chain and CRM data sets, will allow enterprises to scale AI in an ethical and responsible way.

Recent SAPinsider research suggests that while AI adoption is progressing, organizations are cautious toward its application in core operational areas, emphasizing the need for continued development and validation of AI technologies to meet these specialized demands. The recent research report, AI: State of Adoption 2024, showed 27% of organizations currently using or have used AI for customer service, but outside of this function, AI adoption is extremely low for other business functions. Surprisingly, a significant number of organizations are still in the evaluation phase for integrating AI into core business functions such as Finance & Accounting, Procurement & Sourcing, Product Service Design and Development, Risk Management, and Supply Chain and Logistics. This hesitancy could be attributed to the complexity and critical nature of these functions, where the integration of AI requires robust data management, high accuracy, and compliance with stringent regulatory standards.

Granite models are lightweight, cost-effective, open and customizable models underpinned by stringent data transparency and training principles. Here’s how these models stand out:

• Transparent: Granite models are among the most transparent models available today, according to Stanford University’s Foundation Model Transparency Index 2024. IBM offers indemnification for usage of the Granite models and transparency into the datasets used to train them.
• Performant: Trained on curated datasets relevant to enterprise use cases, Granite allows users to tune and guide models to meet business needs.
• Efficient: Smaller models require less computing and inferencing capacity. Granite models can help organizations experiment and scale generative AI applications.

Bringing transparent, performant and efficient Granite models to SAP AI Core on SAP BTP infrastructure helps turn ideas into real-world applications. The initial use cases Granite 13b.chat can enable include bringing Q&A, generative tasks, insight extraction, summarization and classification, especially to ERP processes. And because a considerable portion of Granite model training data is from the finance domain, they are optimized to excel in finance-specific tasks.

For example, assistants and decision support solutions can be built to quickly answer questions related to documents, cases and solutions, products, KPIs and sales metrics, or provide classification of customer problems or user feedback. For supply chain management, summarization of inventory or supplies can be immediately accessed through a chat interface to increase transparency and visibility. Classification can help spot challenges and cross-selling potentials.

This news builds on IBM’s collaboration with SAP around embedding IBM Watson AI technology into SAP solutions. As part of IBM’s Value Generation Partnership initiative announced with SAP in May 2024, IBM Consulting has been building out an extensive portfolio that will include 100 AI solutions across industry, line-of-business and product delivery, with 25 delivered as of October 2024. Many of these are expected to be offered on Granite, such as IBM Intelligent Invoice Match and Verification for SAP Solutions, IBM Intelligent Asset Maintenance for SAP Industrial Manufacturing, IBM Intelligent Warranty Processing for SAP Automotive and IBM Intelligent Direct Distribution to Stores for SAP CPG & Retail. In addition, IBM Consulting plans to build extensions to Granite 13b.chat for select customers.

What this means for SAPinsiders

Participate in SAPinsider research. SAPinsider will be publishing a research report in December on AI in the Supply Chain that will explore the key drivers causing companies to adopt AI in supply chain, the primary strategies companies are deploying to operationalize AI in supply chain, the critical underlying capabilities that must be in place in order for AI to deliver durable outcomes, and the leading technology tools and providers companies are using to embed AI into supply chain processes and systems. Take the survey today and let your voice be heard!

Tap Granite models for enhanced insights. IBM Granite models can complement SAP Datasphere and analyze and process data from various SAP modules (e.g., ERP, CRM, SCM) to generate advanced insights from structured and unstructured data. By aligning Granite with SAP Datasphere, organizations can centralize data to enable more sophisticated predictive analytics and reporting. Further, Granite models, combined with SAP’s analytics and reporting tools, can bridge data silos, allowing SAP customers to draw deeper insights across departments. This is particularly useful in supply chain forecasting, demand planning, and customer insights, where large language models (LLMs) can spot patterns or anomalies that traditional analytics might miss.

Plan for productivity bump via workflow automation. Granite models can help users query SAP data in natural language, reducing the need for specialized SQL or other data-querying knowledge. This enhances productivity by making it easier for teams to access data insights quickly, empowering non-technical users. LLMs can identify and recommend automation opportunities across SAP-driven processes. For instance, Granite could be applied to streamline customer service by automating responses to common queries, invoice processing, and procurement workflows, thereby reducing manual overhead and freeing resources.