Deploying artificial intelligence (AI) is no longer the main challenge for businesses. The real issue now is transforming AI's potential into measurable business value. This challenge has been dubbed the "AI Value Paradox." While companies, especially in retail and consumer goods, are investing heavily in AI, the tangible results from these investments are not keeping pace with the rapid development of AI capabilities. Companies are becoming more adept at deploying AI, but they still struggle to demonstrate, attribute, and track its business impact. An analysis of public information from ten leading retail and consumer goods companies shows that mentions of AI in annual reports have increased by about eight times between 2020 and 2025. Announcements about AI platforms and pilot projects have progressed quickly, but quantifiable results remain limited. Among the measurable gains, cost savings are mainly seen in inventory management for retailers and in selling, general, and administrative expenses for consumer goods. However, the missing piece is the mechanisms that connect AI resources—such as technologies, platforms, and use cases—to actual business outcomes like revenue, profit margins, and working capital needs. Publicly available information suggests that only about 4% of AI-related communications focus on these mechanisms, compared to 73% on resources and 23% on results. A few companies have successfully linked AI to measurable results. For example, Walmart attributes a 35% increase in order value to its AI assistant, Sparky, which has an adoption rate of about 50% among app users. Mondelēz links an 80% efficiency in sales representative work to a 2 to 4% increase in revenue. Unilever achieved a 98% product availability in a pilot while reducing manual planning efforts by about 30%. These successes highlight that while AI can create potential, its value depends on how it is integrated into business processes and how users interact with it. The realization of value depends heavily on user behavior and trust in AI systems. Users may ignore or bypass AI recommendations if they do not trust the system, which is built through factors like explainability, consistency, and reliability. Companies that neglect these aspects often see adoption rates plateau before reaching expected returns. Many AI initiatives remain stuck in the "purgatory" of pilot projects due to insufficient investment, unclear decision-making roles, or resistance from middle management. Poor data quality can further erode trust, leading to disengagement even when the data improves later. In stores, for instance, AI might be used to plan staffing or evaluate store layouts, but without involving store managers in the design or evaluation of the system, these recommendations may be ignored. The cost and measurement of AI follow different logic. AI costs increase with usage, especially for tasks like inference and governance, while value is built in stages. Early results may not be enough to judge the success of an AI initiative. With agentive AI, which can make autonomous decisions, measurement must also focus on the quality of those decisions and the governance structures in place. Humans still play a role through validation and control, but the value increasingly depends on the system's ability to act autonomously. As AI evolves, the ability to transform its capabilities into demonstrable results will become a critical management discipline and a key source of competitive advantage.