More Than a Tool? How AI Design Shapes Customer Responses

Abstract

The first essay is a meta-analysis that provides the first systematic examination of how the AI–human gap in customer responses varies across customer journey stages, and how anthropomorphic design factors moderate this gap at each stage. Analyzing 1,425 effect sizes drawn from 401 studies reported in 129 publications, I document a small but robust baseline disadvantage for AI agents relative to human agents (r̄ = −.140). The gap is largest in pre-purchase stage, where inflated performance expectations amplify disappointment when shortfalls occur, and smallest in post-purchase stage, where lower expectations for AI emotional capability mean that adequate performance can exceed consumer baselines. To explain these patterns, I introduce experience-dimension primacy, a mechanism specifying that cognitive, affective, and social customer experience dimensions shift in relative salience across journey stages, systematically altering how AI agents are evaluated at each point in the journey. Identity cues, form cues, and interaction cues produce pronounced stage-contingent reversals: the same design choice that reduces the AI–human gap at one stage can widen it at another. Taken together, these findings challenge uniform AI deployment strategies and provide actionable guidance for aligning agent design with the dominant experiential dimension at each stage of the customer journey. The second essay examines AI social role design through the lens of brand biography. Firms increasingly deploy AI agents as the primary consumer interface, yet whether their appropriate social role depends on the brand's narrative identity remains unknown. Drawing on the congruity schema, this research explores whether partner-framed AI is congruent with underdog brands and servant-framed AI with top dog brands, advancing perceived congruence as the mechanism and examining perceived power imbalance as an initially proposed mediator, with self-brand connection as a moderator. In Study 1, partner-framed AI produced more favorable responses under both brand biographies, and perceived power imbalance did not mediate. In Study 2, the predicted congruency pattern emerged only among consumers with low self-brand connection, on social experience, satisfaction, perceived congruence, and customer experience, operating through perceived congruence. Therefore, Brand-AI congruency is conditional, holding for consumers without an established brand relationship.