This paper proposes a methodology for implementing Artificial Intelligence (AI) in a dairy barn system at CAETEC (Experimental Agricultural Field from Tecnologico de Monterrey) to optimize dairy production, enhance milk nutritional value, and promote environmental sustainability. The project focuses on integrating data science, predictive models, and AI to select the best dairy cows. The core challenge lies in synthesizing vast amounts of data—including cow genetics, feeding behavior, rumination, milking results from a robotic system, and animal welfare—to inform decision-making. The methodology outlines four steps for implementing AI, including sensor identification, system interconnection, continuous training, and data analysis.

Various monitoring systems and data sources are utilized to feed the databases for the AI algorithms, including the Sense Hub system for rumination and health, the DeLaval VMS CLASSIC robotic milking system, the DeLaval activity meter system for behavior and heat detection, and the GREENFEED system for measuring methane emissions. The crucial, non-digital knowledge of the expert (veterinarian) is also digitized and integrated.

The selection of breeding sires, a crucial step in genetic improvement, is presented as a complex decision-making problem that is ripe for AI intervention. The analysis highlights the combinatorial complexity in selecting a stallion, even when considering only four specific genetic types (A2A2, High HHP$, Mastitis resistant PRO, and Polled Genetics) with multiple variables and sire options. The total number of possible ways to select sires, considering badges and characteristics, exceeds 172,000, which necessitates the use of AI to identify the optimal matches.

The paper proposes that a 'cow-match' tool can be developed using data science and algorithms to identify the optimal cow-sire match. This tool is designed to support decision-making in an academic context, offering a practical application for students. The significant number of variables and possibilities demonstrates the immediate need for an AI-powered prompt and decision theory to assist students from different backgrounds, like agronomy (with prior knowledge) and industrial engineering (without previous knowledge), in selecting the best sires based on multiple, often conflicting, objectives (e.g., milk production, cost, reproductive issues, and pollution). The goal is to minimize errors and measure the probability of success in the selection process.