AI-Driven Digital Transformation in E-Commerce Platforms: A Systematic Literature Review from Automation to Agentic Systems
Abstract
AI shapes e-commerce platforms across customer interaction, operational decisions, and platform governance, yet executives have no consolidated framework for assessing where their platform sits on the AI maturity curve or what conditions support progression across stages. Global enterprise AI spending reached $307 billion in 2025, with retail among the top-three AI-spending industries. Existing reviews cover segments of this arc, including machine-learning optimization, AI personalization, generative AI in business, and early agentic systems, yet each takes the enabling factors as a static inventory, none tracks the full automation-to-agentic trajectory through an organizational-maturity lens with business-outcome evidence, and none has asked whether what most limits progression itself changes from stage to stage. Following PRISMA 2020 guidelines, we synthesize 371 peer-reviewed studies from Scopus and Web of Science (2021–2026) into the AI-Commerce Maturity Model (ACMM), a four-stage capability instrument covering rule-based automation (S1), machine-learning optimization (S2), generative-AI augmentation (S3), and agentic operations (S4). Enablers and barriers were coded under the Technology–Organization–Environment (TOE) framework and outcomes under an inductively-derived eleven-category scheme. The data indicate that the binding constraint on progression shifts predictably — technology-dominant at S1→S2, organization-dominant at S2→S3, and environment/governance-dominant at S3→S4 — evidenced by step-changes in agentic-architecture and leadership-vision factor counts at the S4 boundary. ACMM gives platform executives a stage-differentiated diagnostic for AI investment and governance planning, and reframes Information Systems (IS) maturity-model theory by recasting TOE from a static enabler checklist into a stage-conditional diagnostic, with implications for any domain where AI maturity must be tracked against organizational and governance readiness rather than computational investment alone.