SAP’s Responsible-AI policy and the operating guides for SAP AI Business Services state that any productive AI agent must (1) be embedded in a formal AI-governance structure that can approve use cases and mitigate ethical or regulatory risks, (2) be continuously measured, logged and re-evaluated in operation so that concept drift, bias, or performance degradation are detected early, and (3) remain under explicit human (“expert”) oversight so that critical decisions can be reviewed, overridden or explained to stakeholders.

Providing customer data for additional training or instituting generic organisational change-management initiatives are not listed by SAP as baseline requirements for bringing an agent AI into production.

A. SAP recommends retrieval or fine-tuning only if legally permissible; it is not a general prerequisite for adopting agent AI.

E. Change-management activities can be useful, but SAP documentation does not list them as mandatory integration steps for AI agents.

1. SAP SE – “Guiding Principles for Artificial Intelligence”

Version 3.0

2022

pp. 4-6 (sections “AI Ethics & Governance” and “Human Agency & Oversight”).

2. SAP Help Portal – “AI Business Services: Operations & MLOps Guide”

rev. 04-2023

Section 5.3 “Continuous Evaluation and Monitoring”.

3. SAP SE – “Responsible AI Policy Framework”

Doc-ID 2023-RAI-01

Section 2.1 “AI Governance Bodies”.

4. SAP Note 3259811 – “Best Practices for Deploying AI Agents in SAP Solutions”

paragraphs 7-9 (“Human-in-the-Loop” and “Operational Monitoring”).

5. University of St. Gallen / SAP

“Governance of Enterprise AI” (Working Paper

2021)

pp. 12-13 (governance structures and expert oversight in SAP landscapes).