AI Across the Implementation of H&S Management Systems on Construction Projects: What Does the Evidence Actually Tell Us?

Artificial Intelligence is increasingly being applied across different processes involved in the implementation of Construction Health & Safety (H&S) management systems, yet much of the available evidence evaluates individual technologies or use cases rather than their interaction within the wider system. This article examines evidence gathered through an ongoing independent structured research programme on AI, Construction H&S and Risk Management, focusing on AI solutions applied or evaluated during construction execution and on relevant studies involving H&S practitioners in the evaluation of AI-supported solutions. The evidence indicates growing but uneven AI involvement across risk assessment, safety information and documentation, monitoring and inspection, decision support, corrective action and feedback. However, technical performance, field performance, preventive effectiveness and organisational value represent distinct levels of evidence and should not be treated as equivalent. As AI-generated information contributes to multiple H&S processes, further questions arise concerning the completeness and currency of context, the configuration and adaptation of AI-supported processes, professional judgement, verification and challenge, and the coherence of information and controls as project conditions change. The available evidence does not yet establish that multiple AI capabilities automatically produce a coherent or more effective H&S management system. The analysis therefore shifts attention from isolated AI performance towards how AI-generated information, context and human judgement interact across the implementation of H&S management systems on construction projects.