The Golden Hours of Criminal Investigation: AI-Enhanced Decision Support for Data-Driven, Evidence-Based Policing

Background/Objectives: The golden hours, the critical initial phase of each criminal investigation, are crucial for securing evidence, engaging witnesses, and directing investigative efforts, yet this phase often relies heavily on individual expertise and intuition. This paper explores how artificial intelligence (AI) can enhance evidence-based and data-driven policing during these early hours by improving decision-making and investigative quality. Methods: Drawing on a source pool of 1589 open-source case records, covering solved and unsolved homicides, suspected homicides, suspicious disappearances, and deaths initially of unclear cause in the Netherlands, compiled from public repositories that aggregate material originating from police files and published court decisions, the study develops a conceptual framework that integrates AI techniques such as automated narrative generation, 3D reconstruction, and digital evidence analysis into early-stage investigations, and illustrates it retrospectively with a purposively selected, coded twenty-case sample; scenario generation in the proposed system follows the twelve elementary scenario components (ESC12). Results: By linking empirical data from previous cases with new operational information, that is, information that emerges or is newly discovered during the ongoing investigation, AI is designed to support the narrative approach, accelerate scenario reconstruction, and generate candidate hypotheses and leads for investigator assessment. An author-assigned preliminary evidence-and-readiness assessment rates four AI application domains against five golden-hour criteria, and the retrospective case analysis maps each framework component onto documented investigative problems. Conclusions: AI may improve investigative efficiency and consistency and is hypothesised to mitigate the impact of cognitive bias and time pressure; these effects remain to be demonstrated in prospective evaluations. Meanwhile, challenges remain regarding explainability, reliability, and ethical compliance.