When AI shops for me: How agentic retail representation turns assistance into purchase delegation

Abstract

Agentic artificial intelligence is moving digital retail from recommendation toward autonomous consumer representation. This research develops agentic retail representation capability as the perceived capacity to acquire retail intelligence, adapt choice strategies, preserve consumer goals, and execute bounded actions. The study includes a qualitative pretest, two scale validation samples, three experiments, and an independent category validation study, comprising 4680 quantitative participants and 210 qualitative respondents. Studies 1A and 1B assess four reflective dimensions combined as a formative composite. Study 2 finds higher delegation intention under a bundled capability intervention and statistical indirect associations through perceived diagnosticity and algorithm trust. Study 3 identifies expertise dependent responses across assortments jointly varying in size, priority information coverage, and displayed prices, preventing attribution to complexity alone. Study 4 finds stronger capability related trust under a bundled transparency intervention and higher authorization rates within an experimental shopping task. Study 5 examines measurement transportability and correspondence with independently coded agent activity. These findings concern perceived capability, measured reliance, and task authorization rather than verified commercial purchases. The contribution distinguishes functional capacity from reliance judgments and permission decisions while specifying the limits of measured mediation and bundled interventions. Retail applications require further validation with real payments, delivery, returns, and repeated consumer use.