Analytics Legends The knowledge platform for SAP Analytics
Academy module

SAP Build Process Automation with Agents — From Workflow to Joule Skill to Joule Agent

SAP Build Process Automation with Agents — From Workflow to Joule Skill to Joule Agent — Analytics Legends section illustration for the SAP Analytics knowledge base (concepts, studies, Academy)

As of 2026-09-25

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

Intermediate module that untangles the two meanings of 'agent' in SAP Build Process Automation — the Desktop Agent RPA runner versus the Joule Agent — and builds the documented, end-to-end bridge between them: a BTP Destination (with the commonly-missed sap.processautomation.enabled property), an Action Project (published to library), a Joule Skill authored as a Process Automation project (Call Action and Send Message steps, precise trigger-phrase descriptions), a Joule Agent (Instructions, foundation model, skills as tools, Human-in-the-Loop), and per-environment deployment through a Joule Formation. Covers required entitlements and the hard-gate Joule Booster step, principal propagation across the identity chain (shared ground with M368), and a documented table of six specific failure modes and their causes, sourced from named SAP practitioners' published field guides and SAP's own agent-documentation vocabulary.

What you will learn

  • Distinguish a Desktop Agent (RPA execution component) from a Joule Agent (a reasoning, tool-calling entity built in Joule Studio) and avoid conflating them in scoping
  • Build the documented bridge from a BTP Destination through an Action Project to a Joule Skill authored as an SAP Build Process Automation project
  • Write a skill description that reliably drives agent routing, applying the documented vague-vs-precise rule
  • Configure and deploy a Joule Agent that calls one or more skills as tools, including Human-in-the-Loop for high-impact actions
  • Explain principal propagation across the Joule-to-CAP-action chain and diagnose an identity-trust failure
  • Diagnose each of six documented failure modes (missing destination property, unpublished action, overlapping skill descriptions, destination mismatch, incomplete Joule Booster run, missing IAS trust) to its specific cause

Who this is for. You know SAP Build Process Automation as workflow and RPA tooling, and you keep hearing that it is now part of "agentic SAP" — but the word "agent" means two different things in the same product suite, and mixing them up wastes a design meeting. This module untangles the vocabulary first, then builds the real, documented bridge: a process automation project becomes a Joule Skill, and a Joule Agent calls that skill as a tool. It assumes M333 and benefits from M369's companion module on generative AI hub fundamentals (M325).

1. Two meanings of "agent" in the same product — name them before you design anything

SAP Build Process Automation's own documentation uses "agent" for something specific and much older than generative AI: a Desktop Agent, the software that runs on a Windows or macOS machine and executes recorded or designed RPA automations — screen capture, application interaction, attended and unattended runs. Desktop Agent 3 is the current version, registered and connected to the platform, updated automatically, with its own installation and management documentation. Nothing about a Desktop Agent involves a language model.

Prerequisites

  • M333 (AI & LLM Fundamentals for SAP Consultants)
  • Working knowledge of SAP Build Process Automation (Business Processes, Automations, Decisions) or willingness to explore a trial
  • Basic familiarity with SAP BTP destinations, CAP/OData services, and the BTP cockpit
  • M325 (SAP Generative AI Hub hands-on) helpful for context on the foundation models a Joule Agent can use, not required

Outcomes

  • Scope an 'agent' request correctly in a project conversation, distinguishing Desktop Agent RPA work from Joule Agent design.
  • Build and publish the Destination-to-Action-to-Skill chain for one real SAP API, avoiding the documented configuration gotchas.
  • Write skill descriptions precise enough to produce reliable agent routing, and verify it through skill-level and agent-level testing.
  • Troubleshoot a misbehaving agent by mapping the symptom to one of six documented failure modes rather than guessing.

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