Artificial intelligence is increasingly being used in construction, utilities and infrastructure to identify hazards, plan fieldwork and reduce administrative tasks. However, researchers caution that the technology’s ability to prevent accidents has not yet been clearly demonstrated.
London-based software company FYLD uses AI to predict required permits, help crews adjust plans around weather conditions and analyze short videos recorded by workers about their tasks and safety concerns. The aim is to give workers and managers earlier warnings about potential problems while reducing paperwork.
According to FYLD CEO Shelley Copsey, field workers can spend between 20% and 25% of their time on administrative work. The company’s system uses photos, videos and job information to prepare draft reports, reducing some tasks from about 30 minutes to between two and three minutes.
Infrastructure services provider Amey said in March that it would deploy FYLD among 2,500 field workers to improve safety and provide supervisors with real-time visibility of work. Copsey said case studies involving the platform had reported reductions in incidents and injuries of up to 48%.
Other companies are applying AI to inspections and incident reporting. Bentley Systems’ Blyncsy platform analyzes images captured by drivers to identify issues such as damaged guardrails, overgrown vegetation and debris. The company said a rail project using a digital twin reduced physical site visits by 50% and manual surveys by 70%.
Ultimo’s software scans work orders and shift logs for indications of safety incidents that may not have been reported. It can draft a report for human review, such as in a case involving a warehouse door apparently damaged by a forklift.
Researchers say, however, that much of the existing research examines what AI systems can detect rather than whether they reduce harm. Waleed Umer, an assistant professor at Northumbria University, said constantly changing construction sites make it difficult for cameras to capture every hazard. He also warned that frequent false alarms could cause workers to disregard alerts, while excessive reliance on AI might create a false sense of security.
Billy Hare, a construction management professor at Glasgow Caledonian University, said pilot projects have often emphasized productivity and assumed that reducing workers’ exposure to hazards automatically lowers risk. Although spending less time in dangerous environments could be beneficial, he said more research is needed to establish whether the technology actually prevents accidents.
Great Britain recorded 25 construction worker deaths in work-related incidents during 2025/26, according to the Health and Safety Executive, down from 35 the previous year and below the five-year average of 37. Hare noted that the decline may also reflect an economic slowdown, making it difficult to isolate the impact of technology.
Human supervision remains essential. Umer said supervisors could struggle to review the volume of information generated by AI, potentially leading to recommendations being accepted without sufficient scrutiny. He advised treating AI as an additional set of eyes rather than a replacement for supervision, while clearly defining what the systems cannot detect.
The way companies use time saved by AI is also central to its impact on workers. Ultimo CEO Steven Elsham argued that businesses should measure safety, successful first-time repairs and unexpected breakdowns, rather than focusing only on output. If efficiency gains simply lead to more tasks per worker, he said, employees may see the technology as an intensification of work rather than a safety improvement.
Sources: CNBC Markets.