Skip to main content

SAP AI Golden Path

Last updated on by SAP

The SAP's AI Golden Path is the starting point for developing AI applications across the SAP ecosystem. It contains recommendations, best practices, and tutorials to help you understand the AI technology stack, identify suitable tools and services, and design, deliver, and extend enterprise-grade AI solutions on SAP technology.

It serves as a central entry point into our approach for building AI-native applications, covering everything from data and AI foundation services to agentic architectures and Joule.

Is This Guide for You?​

  • If you're an architect or AI project lead, this guide helps you to plan the architecture of your AI application and choose the right AI capabilities, data layers, and runtime services within the SAP ecosystem.

  • If you're a developer, this guide helps you to select the appropriate tools, SDKs, and frameworks for building and integrating AI features.

  • If you're a product manager, this guide helps you to understand the AI development lifecycle, available technologies, and best practices for delivering AI use cases.

References​

Before starting your AI development journey, you may want to familiarize yourself with the following documents:

  • SAP BTP Developer Guide Starting point for developing business applications on SAP Business Technology Platform (BTP). It contains recommendations and best practices for development projects on SAP BTP.

How to Use This Guide​

Use this guide to navigate the AI development lifecycle, from understanding the available technology to implementing and scaling AI use cases.

Understand Available Technology​

Architecture Overview

The first step toward building an AI application is understanding the technology stack available within SAP. This includes:

  • Agent Layer — including content-based AI Agents built with Joule Studio, code-based AI Agents built with open-source frameworks and SAP's generative AI hub, Joule, and agent tools (MCP).
  • AI Layer — powered by AI Core (including generative AI hub), and SAP HANA Cloud (including Vector Engine, Knowledge Graph Engine, PAL, SparkML), enabling model training, orchestration, and interaction through generative and agentic interfaces.
  • Data Layer — powered by SAP HANA Cloud, Business Data Cloud, and partner technologies like Databricks and Snowflake for advanced analytics and data science.

See Technology Decision Tree for details.

Select the Right AI Approach​

Choosing the correct AI approach depends on the use case and business problem:

  • Agentic Workflows & AI Agents – Context-aware automation, orchestration, and dynamic decision-making.
  • Generative AI & LLMs – Natural language, summarization, and conversational use cases.
  • Classic Machine Learning (ML) on AI Core – Predictive analytics, optimization, classification.
  • Relational Foundation Models – Structured data, tabular predictions.

See Decide on an approach for detailed recommendations.

Develop AI Use Cases​

AI development follows a design-led, iterative process that ensures business value and technical feasibility. This guide structures recommendations across the key development phases:

AI Development Process

This guide mainly focuses on the Design, Deliver, and Run & Scale phases.

  • Explore, Discover — Identify business challenges suitable for AI, evaluate feasibility, and define success metrics or evaluations (evals).
  • Design — Create an AI architecture and decide between ML, LLM, or agentic approaches.
  • Deliver — Set up the environment (SAP BTP, Joule, AI Core), develop and deploy AI applications, and integrate them into existing processes.
  • Run & Scale — Run the application to provide business value.
  • Evaluate & Improve — Measure accuracy, performance, and business outcomes; continuously enhance through structured evaluations.
  • Extend — Build on top of existing applications or agents, integrating with partner ecosystems and third-party solutions.

Get started with building in the Build and Deliver section.