PROBES snapshot: An agricultural knowledge assistant using large language models

This one-page brief, developed by the Digital Transformation Accelerator (DTA) team at the Alliance of Bioversity International and CIAT, is based on the Agricultural Knowledge Assistant probe led by Patil Mukund, Senior Scientist - Soil Physics at ICRISAT. The probe explores how large language models (LLMs) can be used to help researchers and practitioners access, synthesize, and navigate large volumes of agricultural knowledge more efficiently. By enabling users to query documents, datasets, and technical resources in natural language, the tool aims to reduce time spent searching for information and improve evidence use in research and decision-making. Implemented as an early-stage, safe-to-fail experiment, the probe focuses on understanding the accuracy, usefulness, and limitations of LLM-based knowledge assistants in a scientific context. The brief highlights key insights to inform future development and responsible integration of generative AI tools within CGIAR research workflows.