This article explores how retailers and retail-technology vendors envision the use of AI-enabled technologies in loss prevention for brick-and-mortar retail. Using retailer and vendor marketing materials, interviews, and observations at industry events, we identify three key approaches to retail theft prevention: eyes (observation systems), locks (physical deterrents), and patterns (anomaly detection, often powered by AI). While eyes and locks have long been central to loss prevention, patterns are increasingly layered onto these methods to address their limitations. Despite the promise of AI, these systems still depend on frontline workers to interpret and intervene in theft prevention. We argue that patterns-driven solutions could instead shift the nature of tasks and discretion involved in frontline retail work, and future research should consider the conditions that occasion these changes.