SAP AI Core
SAP AI Core is a service on the SAP Business Technology Platform (BTP) that enables organizations to run AI models in a scalable and standardized way.
Particular emphasis is placed on the operational management of machine learning models (MLOps), including deployment, versioning, and monitoring. SAP AI Core provides the (runtime) environment to efficiently orchestrate machine learning workflows and integrate them into existing SAP landscapes.
What are the key functionalities of the service?
SAP AI Core is a cloud service on SAP BTP for implementing, executing, and managing AI workflows (training phase). It is built on various frameworks (e.g., LangChain) and provides an environment for integrating additional modern AI/ML frameworks and libraries. The service enables flexible use of both custom and pre-trained models.
Key features at a glance
| Area | Description |
| Orchestration | Control of AI workflows and pipelines |
| Scalability | Use of containers (Kubernetes) for scalable resources |
| Openness | Support for common ML frameworks and libraries |
| Integration | Connectivity to existing SAP and non-SAP systems and data sources via standardized APIs and BTP services |
| Governance | Control over models, versions, and deployments |
How does SAP AI Core work?
SAP AI Core is based on a containerized architecture and uses containers (Kubernetes) for the flexible orchestration of workloads. This allows training and inference processes to be efficiently distributed and scaled.
A typical sequence looks as follows:
- Development of an AI solution (e.g., using Python)
- Orchestration of trainings and workflows
- Deployment via serving templates
- Integration of results (inference) into SAP processes
Overview of SAP AI Launchpad and SAP AI Core
SAP AI Core is embedded in SAP BTP and, in close interaction with services such as SAP AI Launchpad and SAP Business Data Cloud, forms the foundation for end-to-end AI and data processes.
An AI use case is implemented in SAP AI Core as a scenario. Scenarios are realized using reusable templates (e.g., workflows), which are defined through parameters, datasets, and configurations. These components are bundled for each use case in a dedicated workspace, known as a resource group. Based on these components, a deployment is instantiated. As input, a deployment uses one or more models and parameters from a configuration. Deployments are implemented in a runtime that creates endpoints, enabling access to the deployed models for inference/live predictions.
The following diagram provides an overview of SAP AI Core and the interaction of its individual components:
SAP AI Core and Launchpad: Central Control and Monitoring
While SAP AI Core handles the execution of AI workloads, SAP AI Launchpad provides a graphical interface for managing and monitoring these processes. The Launchpad complements the API with a visual control layer and simplifies access, for example, for business analysts.
Functions of the Launchpad:
- Overview of all models and deployments
- Monitoring of training runs
- Management of scenarios and pipelines
- Analysis of performance and logs
What role does the SAP AI Core AI API play?
The AI API is based on the general AI API specification and serves as a central interface for the lifecycle management of AI assets. These include, among others, training scripts, datasets, and models, which can be managed across different runtimes. This enables standardized management and execution of AI artifacts.
SAP AI Core provides a concrete runtime implementation of the AI API specification and extends it with additional capabilities.
Key functions of the AI API:
- Management of AI assets such as models, training scripts, and datasets
- Control of workflows and deployments via so-called executables (e.g., Argo Workflows and serving templates)
- Execution and monitoring of training and inference processes
- Access to logs, metrics, and status information
- Organization of resources into groups
A key advantage of the AI API is that it serves as a standardized interface for different runtime environments. This allows clients, such as SAP AI Launchpad or the SAP AI Core SDK, to interact with any AI API–compatible runtime.
The following diagram provides an overview of the interaction between SAP AI Core, SAP AI Launchpad, and the AI API:
Orchestration
Orchestration is a service within SAP AI Core that provides unified access to various generative AI models.
A key advantage is the ability to connect different AI models and versions without requiring changes to the client code of custom AI solutions.
Key orchestration features:
- Templating: Creation of dynamic prompts with placeholders that are filled at runtime
- Content Filtering: Control and restriction of inputs and outputs of generative AI models
- Data Masking: Anonymization or pseudonymization of sensitive data before processing
- Grounding: Enrichment of models with external or domain-specific data for context-aware results
- Translation: Integration of translation logic for inputs and outputs within the process
Development with the SAP AI Core SDK
For development, SAP provides the SAP Cloud SDK for AI, which serves as the official SDK for SAP AI Core, the Generative AI Hub, and Orchestration. It supports multiple programming languages, including JavaScript, Java, and Python, and enables standardized integration with the AI services of SAP BTP.
The SDK abstracts the underlying APIs and simplifies the integration of AI capabilities into custom applications.
Key SDK features:
- Access to SAP AI Core and AI APIs via language-specific libraries
- Integration of generative AI capabilities (e.g., chat completion) into applications
- Use of orchestration features such as templating, grounding, data masking, and content filtering
- Support for lifecycle management of AI assets (e.g., models and deployments)
- Simplified setup and configuration of SAP AI Core
Using custom and pre-trained models
With SAP AI Core, organizations can manage different types of models:
1. Custom models
- Development and training of custom models using proprietary training scripts
- Deployment and execution via SAP AI Core (e.g., using executable templates such as Argo Workflows)
- Adaptation to specific business requirements and data
2. Pre-built models and services
- Use of foundation models via the Generative AI Hub
- Integration of preconfigured AI services for standardized use cases
These models are managed through SAP AI Core and the underlying AI API, which enables centralized lifecycle management, including versioning, deployment, and monitoring of AI artifacts.
Overview of SAP AI Core benefits
Why do organizations choose SAP AI Core?
Technical benefits
- Support for modern ML frameworks
- API-first approach
Business benefits
- Faster time-to-value for AI projects
- Integration into existing SAP processes
- Improved decision-making
Governance & Security
- Model versioning
- Access control
- Traceability of decisions
Challenges and Best Practices
In addition to its benefits, there are also typical challenges to consider when implementing and using SAP AI Core:
Challenges
- Complexity of implementation
- Need for ML expertise
- Integration of existing data sources
Best Practices
- Define a clear MLOps strategy
- Select pilot projects strategically
- Strengthen collaboration between business and IT
- Leverage standard services on SAP BTP
FAQ on SAP AI Core
What is the difference between SAP AI Core and SAP AI Launchpad?
SAP AI Core is the technical execution environment for AI models, while the Launchpad provides a user interface for management and monitoring.
Which programming languages are supported?
Primarily Python, as many ML frameworks are based on it.
Is SAP AI Core only relevant for S/4HANA?
No, but integration with S/4HANA is a major advantage. SAP AI Core can also be used independently within SAP BTP.
How does SAP AI Core differ from traditional AI tools?
SAP AI Core is specifically designed for enterprise requirements:
- Integration into business processes
- High scalability
- Governance and compliance
Conclusion: Why SAP AI Core is a key building block of modern SAP architectures
SAP AI Core enables organizations to leverage artificial intelligence not just experimentally, but in a structured and scalable way. Through its tight integration with SAP BTP and SAP S/4HANA, AI becomes an integral part of operational business processes.
Organizations looking to benefit from AI in the long term will find it hard to avoid a service like SAP AI Core. Especially in the context of S/4HANA transformations, the service offers significant potential for automation, optimization, and innovation.
![[Translate to English:] Handschlag Icon](/fileadmin/_icons/_mit-hintergrund-bogen/Kollaboration_d.svg){kind=icon}
