Following the broad public availability of generative artificial intelligence (AI) tools, a managerial narrative has emerged within the small and medium-sized business (SMB) community: that AI is now a commodity input, that any provider delivers equivalent results using the same underlying foundation models, and that the rational decision criterion has accordingly collapsed to price. This paper examines whether the commoditization thesis withstands empirical scrutiny. Drawing on the foundational literature on tacit knowledge (Polanyi, 1966; Nonaka, 1994), the resource-based and knowledge-based theories of the firm (Penrose, 1959; Barney, 1991; Grant, 1996), the cognitive science of expert judgment (Ericsson, Krampe Kahneman Dell'Acqua et al., 2023; RAND Corporation, 2024; MIT NANDA, 2025), the analysis finds that aggregate failure rates between 80% and 95% in generative AI deployments cluster systematically around projects lacking integrated domain expertise, while vendor-led implementations succeed at approximately twice the rate of internal builds. The empirical record reframes generative AI not as a commoditized capability but as a delivery layer whose value is determined by the tacit business judgment encoded into its configuration. The paper proposes the Agentes Para Tu Negocio framework — a bottleneck-first, expertise-led implementation model for owner-operated SMBs — as a theoretically grounded corrective and identifies directions for further empirical validation in Latin American SMB contexts.
Humberto Inciarte (Sun,) studied this question.