Enterprise investment in supply chain artificial intelligence has grown substantially, yet autonomous, multi-step agentic AI systems remain at pilot stage in the vast majority of organizations. This paper investigates the gap between investment and adoption through a structured qualitative study: protocol-guided conversations with 32 purposively sampled supply chain executives, directors, and senior managers across retail, manufacturing, logistics, and consumer goods sectors, conducted throughout late 2025 and early 2026 using a consistent four-domain conversation protocol and analyzed using reflexive thematic analysis following Braun and Clarke (2022). Three foundational barriers emerged with high cross-industry consistency. First, the absence of shared domain ontologies prevents agents from reasoning coherently about supply chain entities defined differently across source systems. Second, the lack of persistent, relational knowledge structures deprives agents of the contextual intelligence that experienced human planners carry as tacit knowledge. Third, the absence of live decision context across agent cycles produces stateless systems incapable of multi-step coordination. These barriers are sequential dependencies: each layer of deficiency compounds the one before it, and organizations that skip foundational infrastructure investments consistently encounter the highest remediation costs. In response, this paper introduces SCALE (Supply Chain Agentic Layer Enablement), a three-layer infrastructure framework addressing each barrier through an Ontology Layer, a Knowledge Graph Layer, and a Context Graph Layer. SCALE reorients the supply chain AI conversation from model capability to system readiness — the precondition that existing literature has systematically underexamined.

Keywords: Agentic AI adoption, Supply Chain Management, Knowledge Graphs, Ontology, Context graphs, Qualitative research, AI infrastructure, Thematic analysis