Open-field garlic cultivation is highly sensitive to external factors such as climate change, pest outbreaks, and soil moisture imbalance, often resulting in unstable crop growth and reduced quality. To address these challenges and improve agricultural productivity, this study proposes an AI-based precision agriculture system that integrates automated decision-making and control. The proposed system consists of physical components, including drones equipped with NDVI and RGB cameras, zone-specific soil moisture sensors, and automated irrigation and pesticide sprayers. It also includes software modules for analyzing imagery and moisture data using YOLOv8 and CNN models, a GPT-based language model for natural language decision-making, and a mobile application for user interaction. Field validation was conducted in a 1582 m2 open-field garlic testbed. Compared to conventional methods, the proposed system reduced water usage by 17.4% and pesticide consumption by 22.0%, while increasing NDVI-based crop stability by 22.3% points. In addition, the system achieved a pest detection accuracy of 93.5%, and enabled real-time decision-making and task execution without user intervention, demonstrating its applicability in aging rural environments. The system shows potential for scalability across various crops and climate conditions and is expected to serve as a practical model for AI-driven agricultural automation. It contributes to accelerating digital transformation in agriculture and fostering a sustainable agricultural ecosystem.