Content
In today’s enterprise landscape, artificial intelligence has moved from experimentation to expectation. Businesses want smarter automation, predictive insights, and intelligent processes and they want all of it tightly integrated with their core systems like SAP S/4HANA.
But there’s a challenge: how do you innovate rapidly with AI without compromising the stability and upgradeability of your SAP environment?
That’s exactly where SAP’s Clean Core strategy comes in. Clean Core encourages organisations to keep their ERP system free of custom modifications ensuring performance, maintainability, and smooth upgrades, while shifting innovation and extensibility to the cloud. In this model, the SAP Business Technology Platform (BTP) becomes the home for innovation. From integrations and extensions to automation and data orchestration, BTP acts as the secure, scalable foundation where enterprises can build the intelligent solutions they need.
One of the most exciting parts of this evolution is the rise of AI-powered extensions. With SAP AI Core and AI Launchpad, SAP now provides a complete framework for developing, deploying, and managing custom machine-learning models while keeping the ERP core clean. Even better: this allows data scientists and developers to continue working in Python, the world’s most popular language for AI, and run their models at enterprise scale within SAP BTP.
In this article, we’ll explore how the Clean Core principle naturally leads to AI innovation on BTP and how you can integrate Python with SAP AI Core and AI Launchpad to bring custom intelligence into your SAP landscape without compromising stability.
What is SAP Clean Core?
Let´s give you a brief overview: SAP’s Clean Core approach ensures that your SAP ERP system, whether it´s S/4HANA Cloud or on-premise evolving toward cloud standards, remains stable, secure, and upgrade-friendly (you can read more about it here).
The idea is simple:
- Keep the ERP core clean (no Z-coding or modifications inside the system)
- Build extensions and innovation side-by-side using SAP BTP
- Adopt standard processes and upgrade continuously without disruption
In practice, this means you avoid old-school Z-customisations embedded in the SAP backend and instead use modern cloud-native services, APIs, and event-driven extensions.
This foundation is critical for AI adoption. To innovate with AI rapidly, without risk to your core business systems, you need a flexible, scalable platform. That platform is SAP BTP.
What is SAP BTP and how does it relate to AI
The SAP Business Technology Platform (SAP BTP) is SAP’s cloud innovation and extensibility platform. It enables organisations to integrate systems, manage data, build applications, and deliver intelligent automation and AI all while keeping the digital core of SAP S/4HANA clean, stable, and upgradeable.
In simple terms, SAP BTP acts as the innovation layer for SAP landscapes. It gives companies a secure space to build and run extensions without modifying the core ERP, aligning perfectly with the Clean Core philosophy.
SAP BTP brings together several capabilities: modern application development tools, integration services to connect systems inside and outside the SAP world, data and database services like SAP HANA Cloud and SAP Datasphere, and analytics through SAP Analytics Cloud. Together, these services provide a complete environment for building scalable enterprise solutions in the cloud.
AI is now a major pillar of SAP BTP. The platform provides everything required to embed intelligence into business processes from out-of-the-box SAP AI services to advanced capabilities for running your own AI and machine learning workloads.
SAP AI Core, AI Launchpad, and built-in AI services
Once SAP BTP is established as the innovation platform, the next step is understanding how AI becomes operational inside the SAP ecosystem. This is where SAP AI Core, SAP AI Launchpad, and SAP’s out-of-the-box AI services come together.
SAP AI Core is an execution environment for custom AI workloads. It allows you to run AI models using modern container-based infrastructure whether you’re training models, running batch jobs, or serving real-time predictions. The power of AI Core lies in flexibility: you can build models in Python using familiar libraries like TensorFlow, PyTorch, or Scikit-learn, package them into containers, and deploy them into a scalable SAP-managed environment. Once deployed, these models can be exposed as secure endpoints and consumed from SAP applications or external systems, all without modifying your SAP S/4HANA or business core.
SAP AI Launchpad complements this by acting as the operational and governance layer. It provides a unified cockpit to deploy and monitor models, track performance and logs, compare model versions, and enforce controls and access. It turns AI into a managed enterprise service, rather than an experiment running on someone’s notebook or private virtual machine.
Together, AI Core and AI Launchpad give organizations full lifecycle AI management: develop, deploy, observe, govern, and continuously improve.
But SAP does not expect you to start from scratch. Alongside custom AI, the platform also provides ready-to-use AI services like SAP Document AI. For example, document extraction, invoice recognition, sentiment analysis, and business-specific machine-learning services across SAP applications. These pre-trained capabilities allow companies to adopt AI quickly without building models themselves, while still leaving room for custom innovation where needed. And importantly, SAP supports large-language-model (LLM) workloads as well. With SAP’s generative AI capabilities and AI Core as a runtime, organizations can deploy fine-tuned language models, host custom NLP pipelines, or integrate enterprise-secured foundation models. This means you can build AI copilots, natural-language interfaces for SAP systems, intelligent document processing, or domain-specific text generation, all governed under enterprise compliance and connected to SAP business data.
SAP offers a vast variety of different LLM providers. This includes both LLMs from providers like OpenAI, Google, or Azure as well as SAP-hosted open-source models from Mistral, Llama and DeepSeek. This allows you to select an appropriate solution for your business case without risks for your data. Moreover, the SAP AI Launchpad allows you to experiment and configure usage of your LLM. For instance, you can easily create a configuration of your favorite LLM and define different prompting setups that solve your tasks. In addition, you have access to administration center, where you can define accessibility and scalability options and track your usage statistics.
So, whether you want to use SAP’s pre-built AI, deploy your own machine-learning models in Python, or run and adapt LLMs for enterprise scenarios, SAP BTP provides the environment to do it cleanly, securely, and without touching your core ERP logic.
Example of an AI model deployment using SAP AI Core
Let’s look at some of the easiest ways to serve your AI models on SAP BTP. Deploying a custom machine-learning model with SAP AI Core follows a clear and structured workflow. First, you develop and package your model along with a serving application, for example, by exposing a prediction endpoint using a Python framework such as FastAPI or Flask. Once the serving code and model artifacts are ready, you containerize them in a Docker image.
The next step is to define a serving executable using a YAML configuration file. This file describes key deployment properties such as the image location, resource configuration, scaling parameters, and version details. In most scenarios, this YAML file is stored in a Git repository that is connected to your SAP AI Core instance, ensuring automated and traceable deployment.
With the configuration in place, you can trigger the deployment either through the SAP AI Launchpad user interface or programmatically using the SAP AI Core SDK and APIs. After a successful deployment, SAP AI Core exposes a secure inference endpoint running on SAP BTP. This endpoint can then be consumed by your applications whether they are SAP extensions, automation workflows, UI applications, or backend services, allowing you to integrate your custom AI seamlessly into business processes while keeping your core systems clean and upgrade-safe.
Why choose PIKON as AI implementation partner?
We at PIKON can support you in your way of implementing modern AI solutions within the SAP ecosystem. That includes:
- Configuring out-of-box SAP AI services on SAP BTP
- Development of custom AI solutions and deployment using SAP BTP
- Connecting deployed AI services to your systems
Contact
