AI tools are being deployed on construction and infrastructure sites to flag hazards and cut paperwork, but researchers say rigorous evidence that the technology reduces workplace injuries remains limited.
Artificial intelligence is being deployed across construction, utilities, and infrastructure worksites to flag hazards, reduce paperwork, and remotely inspect assets — but researchers say the evidence that the technology actually prevents injuries remains thin.
Software company FYLD, headquartered in London, uses AI to help field workers anticipate problems before they arise. Workers record short videos describing their daily tasks and safety concerns, and the system analyzes that information to help crews and managers spot potential issues in advance.
FYLD CEO Shelley Copsey, who previously worked with the Ontario Teachers' Pension Plan — one of the world's largest infrastructure investors — said the company's case studies have shown reductions in serious worksite injuries and incidents of up to 48%.
Copsey said a persistent gap in the industry had been the lack of attention paid to frontline workers' day-to-day experience. "What nobody ever did was just sit and say: well, what goes on in those field workers' lives and how do we make the life of the front line easier?" she told CNBC.
The platform also cuts administrative overhead. Field workers currently spend roughly 20% to 25% of their time on site handling paperwork, Copsey said. AI-generated draft reports have reduced a task that previously took around half an hour to between two and three minutes. Infrastructure services provider Amey announced in March that it would deploy FYLD across 2,500 field workers.
Beyond hazard detection, AI is also reducing the need for physical infrastructure inspections. Nathan Marsh, London-based senior vice president for Europe, the Middle East, and Africa at U.S. software company Bentley Systems, said the company's road maintenance platform, Blyncsy, analyzes images captured by drivers to identify damaged guardrails, overgrown vegetation, and debris. One rail project supported by Bentley Systems cut physical site visits by 50% and manual surveys by 70% through the use of a digital twin, a virtual model of a physical asset.
Steven Elsham, U.K.-based CEO of Dutch maintenance software company Ultimo, said his company's AI tools scan work orders and shift logs to surface safety incidents that might otherwise go unreported — such as a warehouse door showing damage consistent with a forklift strike. "People should still report things themselves," Elsham told CNBC. "This is there for what gets missed."
Despite these applications, academic researchers caution that the safety case for AI on construction sites has not yet been established. Waleed Umer, an assistant professor at Northumbria University in England, said most studies to date have focused on what AI models can detect, rather than whether they prevent harm. Construction sites change constantly, he noted, making it difficult for cameras to capture every hazard. He also warned that frequent false alarms can condition workers to ignore warnings, while overreliance on the technology can create 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 in Scotland, said research projects and pilot studies have focused primarily on productivity gains, often assuming that reduced exposure to hazards translates directly to lower risk — an assumption he said needs further scrutiny.
According to the U.K.'s Health and Safety Executive, 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. Hare cautioned, however, that the decline may reflect an economic slowdown rather than technological progress, making it difficult to attribute the improvement to AI.
For Elsham, the more fundamental question is whether AI makes demanding jobs more manageable rather than simply increasing output expectations. As the industry continues to expand deployments, the pressure on companies will be to define clearly what any time saved is actually for — and what, precisely, AI is and is not equipped to see.
Sara Montes de Oca is the Editor in Chief of TechEchelon. Previously a correspondent and producer in Washington, D.C., covering business, finance, and politics.
More from Sara →Google's Suncatcher prototype satellite — carrying a single Tensor Processing Unit and running DeepMind's Gemma 4 — has completed its commissioning phase and begun active operations in orbit, marking the first time an AI model has run inference workloads on a space-based data center.
U.S. employee engagement dropped to 29% in Q3 2026, its lowest level since 2011, according to Gallup, with remote and hybrid workers at AI-integrated organizations driving the steepest declines.
Microsoft cut its full-year capex forecast to $170 billion from $175 billion, while Alphabet held guidance at $195 billion to $205 billion — both below analyst estimates — as GPU delivery delays from third-party providers push some AI infrastructure capacity into 2027.
Written for readers who already know the basics — markets, AI, and the policy decisions that shape both.