Concept card

Joule Agents Library and Agent Catalogue

As of 2026-07-24T14:00:00Z

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

What is Joule Agents Library and Agent Catalogue?

SAP disclosed 30+ pre-built Joule agents by Q1 FY2026, amortising the 3-6 months a custom agent build normally takes — the cited proof point is Daimler Trucks NA's bid-win rate, 10% to 40%+, €70M impact.

Knowing when to use a pre-built agent from this library versus building a custom one is as important as understanding what it is. The Joule Agents Library is SAP's growing catalogue of pre-built, named AI agents — discrete units of autonomous, multi-step execution that go well beyond Joule's original copilot role of answering questions inside an SAP screen. Where the copilot pattern retrieves information and drafts a response for a human to act on, an agent in this library plans and executes a sequence of steps against live SAP data and transactions, pausing for human approval only at defined checkpoints. SAP has disclosed more than thirty specialized agents available or actively in development, spread across five application domains: finance and controlling, supply chain and procurement, human resources, sales and customer service, and IT and platform management. The two most frequently cited examples are the Bid Analysis Agent in the sales domain and the Production Planning Agent in supply chain, both offered as evidence of the shift from copilot question-answering to genuinely autonomous workflow execution.

The Problem Pre-Built Agents Solve

Why it matters

Each library agent ships with a pre-validated EU AI Act risk classification and defined human-in-the-loop checkpoints — exactly the work that otherwise takes months to design from scratch.
The step-change is from copilot Q&A to autonomous multi-step workflow execution — a materially different capability tier than earlier Joule releases.
A DAG of typed steps (datasphere-query, s4-action, llm-invoke, action-invoke, human-checkpoint) is the actual architecture — not a black-box agent.

Key points

30+ specialised Joule agents available or in active development as of Q1 FY2026 across Finance, Supply Chain, HR, Sales, IT.
Two flagship agents: Bid Analysis Agent (SD/Sales — cited in Daimler Trucks NA 10%→40%+ bid-win rate case) · Production Planning Agent (PP/MRP — exception clustering and resolution options).
Agent anatomy: DAG of step types — datasphere-query · s4-action · llm-invoke · action-invoke · human-checkpoint.
Every pre-built agent ships with EU AI Act risk classification, validated data-access pattern, and documented human-checkpoint placement.
Custom agents via Joule Studio: 2–4 week build for experienced developer vs 3–6 months for equivalent Python/LangChain custom stack.
Human-checkpoint at transactional steps is mandatory for EU AI Act Article 14 compliance — agents do not take autonomous S/4 transactional action without operator approval.
Agent catalogue is in BTP AI Foundation — versioned, governed, subject to same change-approval as prompt templates.
Joule Agents Library and Agent Catalogue is mastered only when it changes a named buyer decision.
Start with the semantic contract and control model before demonstrating the tool.
Use current SAP, analyst, study, KG, and news signals as evidence, not decoration.

Terms used on this page

Joule Agents Library: SAP's catalogue of pre-built named AI agents for specific SAP application domains — each with risk classification, validated data-access, and human-checkpoint configuration.
DAG (Directed Acyclic Graph): The workflow structure of a Joule agent — a set of steps with defined dependencies and execution order, where no step depends on a later step (no loops).
human-checkpoint: A Joule agent step type that pauses execution and requires an operator to approve before the agent continues — mandatory at transactional S/4 actions under EU AI Act Article 14.
s4-action: A Joule agent step type that invokes an S/4HANA transactional API — create PO, approve invoice, reschedule work order. Always preceded by a human-checkpoint in production agents.
Joule Studio: SAP's low-code agent builder on BTP — public preview Sapphire 2025, GA target H2 2026. The authoring and deployment surface for both pre-built and custom agents.
Confidence threshold: A configurable parameter on llm-invoke steps — below this probability the agent escalates to a human-checkpoint rather than proceeding autonomously.
Decision owner: The accountable person who accepts the trade-off and funds the next action.
Semantic contract: The shared definition of business terms, metrics, entities, and access rules used by tools and teams.

Sources

Full card available to members. What the full card adds: the full decision framework · the SAP vs Snowflake / Databricks / Fabric comparison · the common pitfalls and their fix · the cheat sheet · the architecture schemas · the code blocks · the facts worth quoting.

Open in the app →

Guides that answer with this page

These guides cite this page as one of the sources their answer rests on.