Artificial Intelligence (AI) is widely adopted across the e-commerce industry, yet firms still struggle to show that customer-facing AI produces measurable financial returns. Three literatures address the parts of this problem separately: AI-in-retail research treats improved customer experience as an endpoint rather than as the start of a financial claim; marketing and accounting research finds that measures such as satisfaction predict financial performance only inconsistently; and IT-value research shows the challenge of isolating a single investment’s contribution. How a firm joins these steps into one argument, and where that argument might fail, has not been examined. The thesis therefore asks how e-commerce firms link AI-driven customer experience (CX) tools to return on investment (ROI) in their public disclosures and where these linkages break down. 

The thesis is a qualitative embedded multiple-case analysis of thirty AI-CX tool-cases across six e-commerce firms (Klarna, Zalando, eBay, ASOS, Wayfair and Stitch Fix), drawing on 113 public documents. Each tool-case is coded against an integrated three-gate framework, encompassing measurement capability (Gate 1), attribution and causal identification (Gate 2), and financial translation (Gate 3), built on the dynamic-capabilities, multi-touch-attribution, balanced scorecard and hard/soft ROI literatures.

Attribution and causal identification is the most common breakpoint: nine of the thirty tool-cases break at Gate 1, seventeen at Gate 2, and the four reaching Gate 3 do not clear it. No tool-case discloses a benefit netted against its cost, and disclosed rigour and disclosed monetary value never occur in the same tool-case. The AI ROI gap is therefore not a single deficiency but a sequential breakdown in disclosed reasoning. The three-gate framework gives firms and researchers a structured way to locate where a given AI-CX value claim loses its support.