Modern physical security is shifting from reactive, human-dependent models to a proactive stance powered by agentic AI—artificial intelligence capable of independent reasoning and multistep action. This session explores how traditional “jobs to be done” in security, such as detection and deterrence, are being revolutionized by behavioral detection that understands context, such as distinguishing a picnic from a catalytic converter theft. Attendees will learn how these systems act as force multipliers by automating the validation of threats and executing dynamic, unpredictable deterrence—including LLM-generated custom talk-downs and “berserk mode” lighting—to prevent desensitization and reduce alert fatigue for human operators. Finally, we will examine how cloud-native platforms streamline the expensive prosecution phase through natural language forensic search and automated evidence packaging, creating a more efficient, scalable security ecosystem.