Designing a KPI Matrix for AI Performance Management in Retail Operations

The paper addresses a common issue in retail companies: the application of artificial intelligence (AI) is often interpreted and explained differently depending on the operational indicators of a specific business function or the technical indicators of a particular system. While this approach can be useful for operational monitoring, it does not adequately serve management because it fails to demonstrate how much a specific application has actually impacted business results. A more accurate forecast, faster recommendation, or automated decision cannot be considered clear evidence of improved management unless the effects on customers, costs, employees, reliability, and control are evaluated simultaneously. The aim of this paper is to formulate a matrix of Key Performance Indicators that links the application of AI in operations with indicators of business performance, management efficiency, and risk. The matrix is based on a synthesis of 33 papers published between 2022 and 2025. In the analysis, the papers are classified by field of application, the system’s role in decision-making, and the indicators used to measure results. The synthesis identifies situations where the system independently implements decisions and those where it supports the decision-maker. This distinction is important because, in the first case, the focus is on stability, traceability, and control of the decision, while in the second, the emphasis is on how employees accept, modify, or reject the AI system’s recommendations. The contribution is the design of a matrix that enables the monitoring of various initiatives in retail and helps ensure their evaluation is not limited to individual metrics, but instead provides a comprehensive insight into their effects.