EXPRESS: Agentic AI and the New Architecture of Marketing Relationships

AI agents now search, negotiate, and transact for firms and customers, reshaping how marketing relationships form and evolve. We introduce Business to Agent to Customer (B2A2C) marketing , a triadic exchange structure in which firms and their AI agents interact directly with customers and their AI agents. B2A2C rests on delegation, multiplicity, and infrastructural mediation, and creates coordination and governance problems that existing marketing theories do not fully address. To explain relationship management in these markets, we propose Customer-Agent Relationship Marketing (CARMA) , an organic marketing theory built around five interdependent mechanisms: calibration, accountability, reciprocity, mediation, and adaptivity. CARMA departs from relationship marketing and principal-agent theory by treating exchange not as direct firm-customer interaction or single-agent alignment, but as a system in which both sides delegate to learning agents governed together. We develop eight propositions on the antecedents of delegation, the relational failures arising without CARMA’s mechanisms, and how these mechanisms vary by context and agent capability. We argue that trust becomes an ex-ante condition, infrastructure and machine-readable reciprocity shape loyalty through protocol design, and that alignment across human, infrastructural, and agentic layers becomes a source of competitive advantage. We close with six CARMA KPIs and a CARMA Readiness Assessor for evaluating B2A2C readiness.