After the BP Oil Spill, the U.S. federal government and Gulf states pursued compensation for economic losses resulting from diminished recreational beach use. The industry standard for measuring change in beach use requires human analysts to count beach-users manually from aerial imagery. Recent advancements in machine-learning (ML), artificial intelligence (AI) and drone technology have created opportunities to improve the efficiency and cost-effectiveness of this manual method. In this presentation, we introduce an ML-enabled method to automatically count beach-users from aerial imagery. Although developed in the context of oil spill damage assessments, this new approach may provide reliable information at a reasonable cost and therefore enable use of aerial imagery in other contexts (e.g., resource inventories, crowd forecasting, etc). We discuss successes and challenges operationalizing this method and outline a realistic path forward for further intergration of ML into real-world applications of remote sensing.