European Journal of Sustainable Agroecosystems

Background: Modern agriculture faces the dual challenge of increasing global food production to meet the demands of a growing population while reducing environmental degradation under changing climate conditions. Conventional agricultural Decision Support Systems (DSS) often depend on rigid, deterministic models that require extensive manual data input and lack the flexibility to process real-time, heterogeneous, and unstructured environmental information.
Objective: This paper aims to explore the integration of Generative Artificial Intelligence (GenAI), particularly Large Language Models (LLMs) and multimodal vision-language architectures, into agronomic decision support systems. It seeks to develop a comprehensive enterprise-level GenAI-based DSS architecture, evaluate its effectiveness across diverse ecological zones, assess its socio-economic implications for both smallholder and industrial farming systems, and identify strategies to address AI hallucinations and systemic biases.
Method: A systemic GenAI-driven DSS architecture was designed by integrating multi-modal data sources, including satellite imagery, historical weather records, localized IoT sensor streams, and unstructured agronomic literature. The framework leverages LLMs, multimodal vision-language models, Retrieval-Augmented Generation (RAG), and domain-specific fine-tuning to generate contextualized, real-time recommendations for irrigation scheduling, fertilizer management, and integrated pest management. The architecture was conceptually evaluated across different ecological conditions and farming contexts.
Result: The integrated GenAI framework demonstrated the potential to provide accurate, context-aware, and real-time agronomic recommendations by effectively synthesizing diverse data sources. The architecture significantly reduced the technical barriers to technology adoption, particularly for farmers with limited technical expertise, while supporting more informed decision-making across varying agricultural environments. However, challenges related to AI hallucinations, bias, and reliability were identified, emphasizing the need for robust validation mechanisms.
Conclusion: Generative AI has substantial potential to transform agricultural decision support systems by enhancing adaptability, accessibility, and data-driven decision-making. Nevertheless, ensuring ecological safety and operational reliability requires the implementation of rigorous validation frameworks, Retrieval-Augmented Generation (RAG), and domain-specific model fine-tuning. These measures are essential for the responsible deployment of GenAI in sustainable and precision agriculture.