Companies Deploy AI to Improve Infrastructure Worker Safety, CNBC Reports

How Software Anticipates Site Hazards

When Shelley Copsey was appointed by the Ontario Teachers’ Pension Plan to find technology solutions across its global portfolio, she identified persistent safety issues facing front-line crews. Four years later, as CEO of the London-headquartered software firm FYLD, she oversees artificial intelligence applications designed to protect workers maintaining roads, railways, and building sites.

Field workers record short videos detailing their daily tasks and safety concerns, which the AI analyzes to help managers anticipate potential problems before work begins. The system also assists crews in planning around adverse weather, such as rescheduling heavy lifting from days when strong winds are forecast. FYLD predicts which permits a project will require to help workers avoid arrival delays.

Administrative burdens also decline under the system. Field workers typically spend 20% to 25% of their working hours on site handling paperwork. The software uses photos, video, and job logs to generate draft reports, cutting a task that previously took approximately half an hour down to two or three minutes.

Infrastructure services provider Amey announced in March that it would roll out FYLD to 2,500 field workers to improve safety and give supervisors real-time visibility on the ground. Copsey stated that FYLD case studies have reported reductions in workplace incidents and injuries of as much as 48%.

AI Analyzes Imagery to Spot Roadway Damage

Technology companies are also using remote inspections to minimize the time workers spend in hazardous environments. Nathan Marsh, senior vice president for Europe, the Middle East and Africa at Bentley Systems, explained that the firm’s road maintenance platform, Blyncsy, analyzes images captured by drivers to spot damaged guardrails, overgrown vegetation, and roadway debris.

“Applying AI to existing imagery provides complete asset coverage in a fraction of the time a conventional survey would take,” Marsh told cnbc.com.

Bentley Systems reported that one rail project supported by its digital twin technology—a virtual model of a physical asset—cut physical site visits by 50% and manual surveys by 70%. Meanwhile, Steven Elsham, CEO of Dutch maintenance software company Ultimo, noted that his firm’s AI tools scan work orders and shift logs for signs of unreported safety incidents, drafting summaries for human review. For instance, the system might flag warehouse door damage indicative of a forklift collision.

“People should still report things themselves,” Elsham said. “This is there for what gets missed.”

The Limits of Safety Evidence

Academic researchers caution that while digital tools increase productivity and reduce physical exposure, proving a direct causal link to accident prevention remains complex. Waleed Umer, an assistant professor at Northumbria University, stated that many studies focus primarily on what detection models can spot rather than whether they actively prevent harm.

Construction sites change constantly, which makes it difficult for cameras to capture every hazard, Umer noted. He warned that false alarms can cause workers to ignore warnings, while an overreliance on AI can generate a false sense of security.

“Treat AI as an extra pair of eyes, not a replacement for supervision,” Umer said. “And be clear about what it can’t see.”

Billy Hare, a professor of construction management at Glasgow Caledonian University, added that research projects and pilot studies have largely focused on productivity gains, often assuming that reduced exposure to hazards automatically lowers risk. While spending less time in hazardous environments should theoretically reduce accidents, establishing whether the technology actively prevents injuries requires further investigation.

Official safety data shows how difficult it is to measure technological impact. According to the U.K.’s Health and Safety Executive (HSE), the number of construction workers killed in work-related incidents in Great Britain fell to 25 in 2025/26, down from 35 the previous year and below the five-year average of 37.

“The U.K. has seen the lowest construction fatality figure ever recorded this year, but this may be due to economic slowdown,” Hare noted, explaining why it remains difficult to determine whether technology has made a measurable difference.

Human oversight must remain rigorous, according to Umer, particularly if automated systems generate more information than supervisors can properly review, which risks leading to rubber-stamped recommendations without adequate scrutiny. Ultimo’s Elsham emphasized that companies must decide whether AI truly makes difficult jobs more manageable or merely squeezes greater output from the existing workforce.

“Safety teams are already covering more hazards with the same headcount,” Elsham said. “If the only plan for the freed-up hours is more tickets per head, workers will see straight through it, and they’d be right.”