Generative AI with SAP
Discover the many possibilities of generative AI in the SAP environment! In this book, the team of authors shows you how to develop a customized AI assistant step by step using SAP AI Core, Generative AI Hub, and other SAP tools. You will be guided through the entire process, from the basics to personalization and deployment. In addition, you will learn everything you need to know about the legal framework and security and quality aspects of using AI.
- Program your own AI application
- Use of Generative AI Hub, SAP AI Core, SAP AI Launchpad, and much more
- Including legal framework, AI security, and quality assurance
You'll learn about:
- Understanding the basics:
What is generative AI and how does it work? The team of authors clearly explains the concepts behind machine learning, neural networks, and large language models. Get to know SAP's AI tools and discover typical use cases.
- Develop your own AI assistant:
Follow the instructions to program a custom AI model with SAP AI Core and SAP AI Launchpad. All steps are explained, from selecting the language model to training with your own data to deploying the assistant.
- Secure and compliant use:
AI also carries risks and is subject to legal requirements. Learn how to eliminate vulnerabilities, monitor the quality of your AI system, and meet the requirements of the EU AI Act. Then nothing will stand in the way of productive use.
Key Highlights:
- Basics of AI and machine learning
- Large language models
- SAP AI Core
- SAP AI Launchpad
- Selection and fine-tuning of language models
- Personalization with Retrieval Augmented Generation
- Training with the SAP HANA Cloud Vector Engine
- Agents
- System integration for real-time data queries
- Quality assurance
View Full Table of Contents
- Introduction
- PART I Fundamentals
- 1 Fundamentals of AI and machine learning
- 1.1 What is AI/ML?
- 1.2 Types of ML: Supervised, Unsupervised, Reinforcement
- 1.3 From Statistics to Generative AI
- 1.4 GenAI, a new AI paradigm for generalized systems
- 1.5 Summary
- 2 What happens in AI?
- 2.1 Python: The right programming language for AI
- 2.1.1 Python
- 2.1.2 The Python ecosystem for AI
- 2.1.3 Python setup
- 2.1.4 Hello, World!
- 2.1.5 Pip
- 2.2 How do machines learn?
- 2.3 Neurons as the basic building blocks of AI
- 2.4 Neural networks as architecture for AI models
- 2.5 Transformers as architecture for large language models
- 2.6 Structure and Functionality of Large Language Models
- 2.7 Summary
- 3 The SAP AI portfolio
- 3.1 SAP Business AI
- 3.2 Joule and SAP Build Code
- 3.3 The SAP Data Architecture: SAP Business Data Cloud, SAP Datasphere, and SAP Analytics Cloud
- 3.4 AI Foundation on SAP BTP
- 3.4.1 SAP AI Core
- 3.4.2 SAP AI Launchpad
- 3.4.3 Generative AI Hub
- 3.4.4 SAP HANA Cloud Vector Engine
- 3.4.5 SAP Knowledge Graph
- 3.5 Summary
- 4 Examples of AI Applications in a Business Context
- 4.1 Practical AI use cases from SAP
- 4.1.1 Purchasing
- 4.1.2 Expenses
- 4.1.3 Sales
- 4.1.4 Service
- 4.1.5 HR
- 4.1.6 Process analysis
- 4.2 Best practices for introducing an AI assistant
- 4.3 Customer-specific AI use cases
- 4.3.1 Processing offers in purchasing
- 4.3.2 AI assistant with company knowledge
- 4.3.3 Personnel resource planning
- 4.4 Summary
- 5 Your tools for AI development: SAP AI Core and SAP AI Launchpad
- 5.1 Architecture: The SAP Generative AI Reference Model
- 5.1.1 Security and Governance Mechanisms in the SAP Generative AI Reference Model
- 5.1.2 Technical Process of an Interaction with the AI Model
- 5.2 Functions of SAP AI Core and SAP AI Launchpad
- 5.2.1 Functions of SAP AI Core
- 5.2.2 Functions of the SAP AI Launchpad
- 5.2.3 Interaction between SAP AI Core and SAP AI Launchpad
- 5.3 Service plans for SAP AI Core and SAP AI Launchpad
- 5.3.1 Free plan for SAP AI Core and SAP AI Launchpad
- 5.3.2 Standard Plan for SAP AI Core and SAP AI Launchpad
- 5.3.3 Extended Plan for SAP AI Core
- 5.4 Runtime Environments
- 5.4.1 Cloud Foundry
- 5.4.2 Kyma
- 5.4.3 Cloud Foundry and Kyma in Comparison
- 5.5 Sizing and Licenses
- 5.5.1 Licensing models
- 5.5.2 Sizing
- 5.6 Summary
- 6 Setting up SAP AI Core and SAP AI Launchpad
- 6.1 Configuration
- 6.1.1 Prerequisites
- 6.1.2 Setting Up SAP AI Core and SAP AI Launchpad
- 6.1.3 Best Practices for Operational Use
- 6.2 Security and Authorizations
- 6.2.1 Architecture of Identity and Authorization Control
- 6.2.2 Business Roles in SAP AI Core
- 6.2.3 Roles in SAP AI Launchpad
- 6.2.4 Basic Principles for Role Assignment
- 6.2.5 Integration with Identity Providers
- 6.2.6 Examples of Typical Role Scenarios
- 6.2.7 Future developments and challenges
- 6.3 Administration
- 6.3.1 Git repository
- 6.3.2 Applications
- 6.3.3 Resource Groups
- 6.3.4 Object Store
- 6.3.5 Secret Docker Registry Keys
- 6.3.6 Generic Secret Keys
- 6.4 Monitoring
- 6.4.1 Monitoring Functions of SAP AI Core
- 6.4.2 Monitoring Functions of SAP AI Launchpad
- 6.4.3 Monitoring Functions in the Generative AI Hub
- 6.4.4 Extended Integration into the SAP BTP Monitoring World
- 6.4.5 Best Practices for Effective Monitoring
- 6.5 Summary
- PART II Developing an AI Assistant
- 7 The LLM as the Heart of Your AI Assistant
- 7.1 Selecting LLMs in SAP AI Core
- 7.1.1 Functions of the Generative AI Hub
- 7.1.2 Decision-making for a model
- 7.2 SAP AI Core as a runtime for LLMs
- 7.2.1 Overview
- 7.2.2 Scenarios
- 7.2.3 Configurations
- 7.2.4 Executions
- 7.2.5 Implementations
- 7.2.6 Schedules
- 7.2.7 Datasets
- 7.2.8 Models
- 7.2.9 Result sets and other artifacts
- 7.3 AI service providers as an alternative to your own server
- 7.3.1 Advantages of external AI service providers
- 7.3.2 Disadvantages of external AI service providers
- 7.3.3 Selection criteria for an external AI service provider
- 7.4 Fine-tuning as a way to personalize an LLM
- 7.4.1 Implementation of fine-tuning
- 7.4.2 Alternatives to fine-tuning
- 7.4.3 The right operating model for fine-tuning
- 7.5 Summary
- 8 Personalizing your AI system with retrieval-augmented generation
- 8.1 How hallucinations arise
- 8.2 Extending LLM with retrieval-augmented generation using your own data
- 8.2.1 How does retrieval-augmented generation work and what are its advantages?
- 8.2.2 Chunking as a method for reducing and dividing context
- 8.2.3 Keeping data up to date and adding metadata
- 8.2.4 Authorization in RAG systems
- 8.3 Implementing a RAG system
- 8.4 Semantic text search with embeddings
- 8.4.1 What are embeddings?
- 8.4.2 Embedding model training: Sentence pair scoring and contrastive learning
- 8.4.3 Use cases for semantic similarity
- 8.4.4 Pre-built embedding models
- 8.4.5 Implementing semantic search with Python and strengths and weaknesses of embedding models
- 8.4.6 Designing and implementing a RAG system
- 8.5 Embeddings in SAP AI Core
- 8.6 Vector databases as storage locations for embedded data
- 8.7 SAP HANA Cloud Vector Engine as a knowledge pool
- 8.8 Testing, applying, and integrating the RAG system
- 8.8.1 Testing the application
- 8.8.2 Ensuring data quality
- 8.8.3 Developing a Python FastAPI interface for separating indexing and query logic
- 8.8.4 Developing the User Interface and Evaluating the System
- 8.9 Summary
- 9 Interaction of an AI model with external systems via agents
- 9.1 Using agents
- 9.1.1 Necessity of agents
- 9.1.2 Practical example 1: Interaction with external systems
- 9.1.3 Combining information gathering and interaction
- 9.1.4 Practical example 2: Interaction based on external information
- 9.1.5 Further possibilities for using agents
- 9.2 Developing an agent with LangChain
- 9.2.1 Setting up LLM
- 9.2.2 Developing the necessary tools
- 9.2.3 Providing tools to the LLM
- 9.3 Connecting SAP systems to query data in real time
- 9.4 Software development kits from SAP
- 9.4.1 Architecture
- 9.4.2 SAP Cloud SDK for AI: JavaScript
- 9.4.3 SAP Cloud SDK for AI: Java
- 9.4.4 SAP Cloud SDK for AI: Python
- 9.4.5 SAP HANA ML Libraries
- 9.5 Summary
- 10 Training LLMs
- 10.1 Approaches to Customizing an LLM
- 10.2 Fine-Tuning an LLM
- 10.2.1 Selecting a Base Model
- 10.2.2 Collecting training data
- 10.2.3 Setting up the training environment
- 10.2.4 Training the LLM
- 10.2.5 Evaluation
- 10.2.6 Different fine-tuning techniques
- 10.3 Fine-tuning an LLM on SAP BTP
- 10.4 Summary
- PART III Security, quality, and legal principles
- 11 Security of LLMs and AI
- 11.1 Special security requirements for AI systems
- 11.2 Injections
- 11.2.1 The different types of injections
- 11.2.2 Security strategies
- 11.2.3 Implementation of security strategies
- 11.3 Adversarial attacks
- 11.4 Summary
- 12 AI quality: Enabling transparency
- 12.1 Tracing AI systems
- 12.2 Monitoring LLMs using tracing
- 12.3 Summary
- 13 Legal framework
- 13.1 EU AI Act
- 13.2 National implementation
- 13.2.1 Implementation in Germany
- 13.2.2 Implementation in Austria
- 13.2.3 Special status of Switzerland
- 13.3 Summary
- The team of authors
- Index
Generative AI mit SAP
Entdecken Sie die vielfältigen Möglichkeiten von generativer KI im SAP-Umfeld! In diesem Buch zeigt Ihnen das Autorenteam, wie Sie mit SAP AI Core, Generative AI Hub und weiteren SAP-Tools Schritt für Schritt einen maßgeschneiderten KI-Assistenten entwickeln. Von den Grundlagen über die Personalisierung bis hin zur Bereitstellung werden Sie durch den gesamten Prozess geleitet. Zusätzlich erfahren Sie alles über die rechtlichen Rahmenbedingungen sowie Sicherheits- und Qualitätsaspekte beim Einsatz von KI.
- Programmieren Sie Ihre eigene KI-Anwendung
- Einsatz von Generative AI Hub, SAP AI Core, SAP AI Launchpad u.v.m.
- Inkl. rechtlicher Rahmenbedingungen, KI-Sicherheit und -Qualitätssicherung
Du lernst etwas über:
- Grundlagen verstehen:
Was ist generative KI und wie funktioniert sie? Das Autorenteam erklärt anschaulich die Konzepte hinter Machine Learning, neuronalen Netzen und Large Language Models. Lernen Sie die KI-Tools von SAP kennen und entdecken Sie typische Anwendungsfälle.
- Eigenen KI-Assistenten entwickeln:
Folgen Sie der Anleitung, um mit SAP AI Core und SAP AI Launchpad ein individuelles KI-Modell zu programmieren. Von der Auswahl des Sprachmodells über das Training mit eigenen Daten bis zur Bereitstellung des Assistenten werden alle Schritte erklärt.
- Sicherer und regelkonformer Einsatz:
KI birgt auch Risiken und unterliegt gesetzlichen Vorgaben. Erfahren Sie, wie Sie Schwachstellen eliminieren, die Qualität Ihres KI-Systems überwachen und den Anforderungen des EU AI Act gerecht werden. So steht dem produktiven Einsatz nichts mehr im Weg.
Aus dem Inhalt:
- Grundlagen von KI und Machine Learning
- Large Language Models
- SAP AI Core
- SAP AI Launchpad
- Auswahl und Finetuning von Sprachmodellen
- Personalisierung mit Retrieval Augmented Generation
- Training mit der SAP HANA Cloud Vector Engine
- Agents
- Systemintegration für Echtzeitdatenabfragen
- Qualitätssicherung
Komplettes Inhaltsverzeichnis
