AI & Analytics Legends The knowledge platform for SAP Analytics Sign inSign up
Academy module

SAP Domain Models and SAP Knowledge Graph for Agents — Design, Build and Run Grounded Joule Agents

SAP Domain Models and SAP Knowledge Graph for Agents — Design, Build and Run Grounded Joule Agents — Analytics Legends section illustration for the SAP Analytics knowledge base (concepts, studies, Academy)

As of 2026-10-16

Edited by , sole editor and publication director — 27 years in the SAP ecosystem. How these pages are produced, and where automation is used.

A design-build-run module on grounding custom and extended Joule agents with SAP Domain Models and the SAP Knowledge Graph. It splits grounding into four layers with four owners, states the dated status of SAP Domain Models (Early Adopter Care since 13 May 2026, GA planned Q3 2026, not confirmed as of 15 October 2026) and of the Knowledge Graph, and shows what to build in the layers a customer controls: data-product semantics, a minimal own ontology, a scoped tool set and layer-specific evaluation. The SAP TechEd 2026 sessions RA108, ST103, RA161, ST118, AS209, RA110, RA163 and BT162 are used as an agenda and a list of questions to verify after the event.

What you will learn

  • Name the four layers that ground a Joule agent (model knowledge, structure, data, context and memory), their owners and the dated status of each SAP component
  • Write a grounding map that routes each question type to a path, an owner, a source of truth and a test
  • State what can and cannot be assumed about SAP Domain Models as of 15 October 2026 and keep unconfirmed capabilities on a depends-on-SAP list
  • Decide when SAP's Knowledge Graph suffices and when customer-specific semantics need an own ontology or data-product semantics
  • Build four layer-specific evaluation sets (path, fact, semantics, refusal) and run them as scheduled regression
  • Turn the SAP TechEd 2026 session descriptions into a written question list for SAP to verify after 30 October 2026

Who this is for. You can already build a Joule agent (M326), you know what the SAP Knowledge Graph is and how the HANA Cloud engine works (M037) and you have read the concept cards on SAP Domain Models (C390) and the Knowledge Graph (C028). This module answers the next question a delivery lead gets: "What exactly makes our agent's answers right, and what do we build, configure and monitor to keep them right?" It does not re-teach RDF, SPARQL or the engine labs of M037, nor agent construction (M326), MCP mechanics (M327), identity (M368) or production testing (M374). It connects them into one grounding design for a custom or extended Joule agent. Platform vocabulary: SAP Business AI Platform is the name of the platform that was SAP BTP.

1. Four layers of grounding, and who owns each

An agent that acts on business data can be wrong for four different reasons, and each has a different owner.

Prerequisites

  • Completion of M037 (SAP Knowledge Graph and the HANA Cloud Knowledge Graph Engine) or equivalent hands-on experience
  • Completion of M326 (Building Joule Agents with Joule Studio) or equivalent
  • Optional: read access to a Business Data Cloud tenant to run the coverage test

Outcomes

  • Diagnose a wrong agent answer by grounding layer instead of blaming the model.
  • Produce a grounding map a steering committee can approve.
  • Tell a sponsor, with source and date, what is announced and what is confirmed for SAP Domain Models.
  • Specify the minimal custom ontology that closes a customer-specific gap.
  • Run an evaluation that would catch a wrong traversal, a wrong number and an invented definition.

Full module available to members. The full module adds: the decision framework · the end-to-end scenario walkthrough · the KPI scorecard · the anti-patterns · the code blocks · the knowledge check · the diagrams.

Open in the app →

Next in this track