GeoAgriSense is an AI-assisted crop-scouting application for smallholder horticultural farmers and farm managers. Farmers can record crop concerns, attach images and context, and use Gemini Flash 3.8 to interpret the available evidence, surface missing information, and suggest what to check next. Built with Python, Django, Streamlit and SQLite, the prototype organises observations into an explainable scouting workflow that helps users decide where to investigate first. The goal is to make limited scouting time more useful without replacing farmer judgement or professional agronomic review.