Agent Design Patterns for SAP Processes — Tools, Planning, Human in the Loop

What you will learn

• Write an agent design file with expertise, instructions, a tool list with access and identity per tool, a thinking-step budget, a memory rule and five checkpoints
• Score each tool by worst-case failure and accountable owner, and confirm that an MCP tool is exposed through the MCP Server artifact type with principal propagation
• Set a thinking-step budget (for example 5) and define the exhaustion path: partial result, escalation or clean failure
• Choose between a subagent, an MCP tool, an A2A delegation, an n8n workflow and a skill chain for a given callee
• Build and test a three-way-match agent in Joule Studio with a 25-task golden set including injection and missing-record tasks
• Map the five EU AI Act Article 14 capabilities to named, testable mechanisms and register the agent in the SAP AI Agent Hub

Module overview

Who this is for. “AI & LLM Fundamentals for SAP Consultants — Transformers, Tokens, Embeddings, Context” taught you to call a model and let it call a tool. This module is the next step: designing something that keeps acting (read state, choose a tool, observe the result, decide the next step) until the goal is met or a human has to step in. That is an agent. SAP gives you a documented way to build one in Joule Studio and two protocols for wiring it to tools and to other agents, MCP and A2A. You leave with a repeatable design method, a design file you can put in Git, and a test set. It closes the loop of the batch opened by Signavio recommends an agent, the AI Agent Hub governs it and CAP can supply its tools. Statuses are as of 28 October 2026; the platform is SAP Business AI Platform, which brings together SAP BTP, SAP Business Data Cloud and AI Foundation (SAP News, May 2026).

Source: SAP Community — Give Business Agents Memory with SAP HANA Cloud Agent Memory Service (memory, isolation, revocation)

Prerequisites

• Completion of AI & LLM Fundamentals — tool calling, agents, MCP, A2A and strongly recommended CAP and the SAP Cloud SDK for AI
• Recommended: AI in SAP Signavio and AI in SAP LeanIX, since this module closes the agent lifecycle both begin
• Comfort reasoning about risk and accountability for an automated system's actions, not just its correctness

Outcomes

• Design an agent's toolset scoped to the smallest set that does the job, with a named accountability for each tool's worst-case failure.
• Set and justify a thinking-step budget with an explicit failure path (partial result, escalation, or clean failure).
• Choose correctly between subagent, MCP tool and A2A delegation when composing multiple agents for one SAP process.
• Produce a five-line EU AI Act Article 14 mapping for any agent design, naming the exact mechanism behind each oversight capability.