<![CDATA[SAP S/4HANA condition determination traditionally relies on static and deterministic access sequences that evaluate condition records in a predefined order and apply the first eligible match. This mechanism provides predictable execution but does not dynamically account for demand context, customer behavior signals, or competitive pricing conditions. This paper proposes an AI-driven condition determination architecture for SAP S/4HANA that introduces a machine-learning prediction layer to select eligible condition types and determine their values while preserving standard pricing governance. The proposed architecture uses the PRCG_DOC_CONDITION_AMOUNT Business Add-In as the injection point, a Business Technology Platform-hosted API proxy as the integration layer, and a supervised learning model trained exclusively on SAP order-context data. Model features are derived from transaction and pricing objects, including XKOMV, KOMK, KOMP, and PRCD_ELEMENTS, enabling predictions from the live order context available during condition determination without external data enrichment. A confidence-gated governance mechanism controls the use of model predictions by applying validation and fallback rules when prediction confidence or eligibility criteria are not satisfied. Implementation evidence from IBM's Blue Harmony global enterprise program was evaluated across multi-country pricing scenarios to assess the accuracy of condition-type prediction and the reliability of condition-type determination. The results demonstrate measurable improvements in condition-type accuracy and reductions in pricing-determination errors compared with the conventional deterministic process. The architecture also maintains traceability of AI-influenced decisions and provides a controlled fallback to standard SAP determination when required. The findings indicate that machine-learning augmentation can extend SAP S/4HANA condition determination from a purely rule-based mechanism toward context-aware pricing decisions while maintaining auditability, operational safeguards, and clean-core compliance]]>
