A Path from Generative AI to Agentic AI in Marketing: A Critical Review, Capability Framework and Future Research Agenda

Abstract

Purpose. Marketing organizations are moving from prompt-driven Generative Artificial Intelligence (Generative AI) toward agentic systems that plan, act, and adapt across multi-step tasks with limited human supervision. The scholarly literature on this transition is dispersed across marketing, information systems, and computer science, and it rarely distinguishes technological capability from the organizational conditions under which autonomous agents create legitimate and durable marketing value. This article critically reviews that relationship and develops a capability-based framework for autonomous marketing. Design/methodology/approach. The article is a critical narrative review rather than a systematic or empirical review. Google Scholar was the principal literature-discovery platform, supplemented by backward and forward citation searching, during July–August 2026. The review audits the 35 references in the source manuscript, retains all as traceable and topically relevant, and adds 8 supplementary scholarly sources identified through citation chaining, producing a cited corpus of 43 scholarly works. A structured narrative process covered identification, relevance screening, design-sensitive appraisal, thematic coding, and narrative synthesis. Dynamic capabilities theory structures the analysis. Findings. Agentic AI does not automatically improve marketing outcomes. Its value is contingent on an organizational capability-conversion process in which AI affordances are translated, through data infrastructure, analytical and creative skill, redesigned workflows, and governance, into accountable marketing routines. Evidence specific to Agentic AI in marketing remains overwhelmingly conceptual, vendor-authored, or transferred from adjacent domains such as personal selling and customer service; controlled, longitudinal, or audited evidence of autonomous-agent performance in marketing is scarce, and widely circulated efficiency and return-on-investment figures originate predominantly from industry and consultancy reports rather than peer-reviewed research. Originality/value. The review replaces a linear technology-adoption narrative with a capability-conversion explanation, separates reviewed evidence from proposed relationships, and contributes a four-stage evolutionary model, a governance-sensitive AI–capability–marketing relationship framework, and testable propositions distinguishing technological affordances from organizational capability, marketing practice redesign, and verified outcomes.