The role of artificial intelligence in reducing dispensing errors for patient safety and quality: a systems approach

Abstract

Dispensing errors, often driven by look-alike/sound-alike medicine names, similar packaging, and complex workflows, pose a persistent threat to patient safety and care quality. Artificial intelligence (AI) offers new opportunities to detect discrepancies and support decision-making in near real time, yet its impact depends on how it is embedded within the wider healthcare system. In this perspective, we use a systems approach to synthesize current AI-enabled strategies for reducing dispensing errors and to outline a roadmap for their safe and effective implementation. We focus in particular on an AI-based natural language processing (NLP) decision-support application as an exemplar, examining how it can be integrated into dispensing workflows to flag high-risk prescriptions and labelling discrepancies before medications reach patients. Using systems thinking, we organise our analysis around four interrelated perspectives: people (training, human–AI teaming, trust), system (interoperability, data pipelines, monitoring), design (human-centred interfaces, uncertainty displays, workflow fit), and risk (ethical oversight, bias assessment, safety assurance, and governance). Across these perspectives, we identify priorities such as multimodal data use, external validation across sites and populations, prospective evaluation with safety and equity metrics, and continuous model monitoring with clear rollback mechanisms. AI can enhance safety, timeliness, and efficiency in dispensing; however, its value depends on disciplined sociotechnical integration and feedback within learning healthcare systems, rather than on standalone algorithmic performance.