Construction sector AI safety gains remain difficult to verify

Traders Union
All News
Business & Wealth
Companies
AI safety in construction

Highlights

Amey began rolling out FYLD's AI safety platform to 2,500 field workers in March, with FYLD reporting incident and injury reductions of up to 48%.
FYLD claims its system reduces site administrative workload by automating draft reports, decreasing completion time from thirty minutes to just two or three minutes.
Academic researchers and sector leaders warn that while AI adoption grows, conclusive evidence of reduced accidents remains lacking due to study limitations and implementation challenges.

Adoption grows across field operations

As reported by CNBC, software providers and infrastructure groups are increasing the use of AI tools to support front-line workers, with applications ranging from permit planning to hazard detection and automated reporting.

London-based FYLD says its system helps predict which permits a project will need, reducing delays when crews arrive on site without complete information. Chief Executive Shelley Copsey says the platform also helps teams adjust work around weather conditions, such as moving heavy lifting away from periods of strong winds, while workers can record short videos about daily tasks and safety concerns for AI analysis.

Copsey says that process also reduces administrative work, which can account for about 20% to 25% of time spent on site. She says photos, video and job information can be used to generate draft reports, cutting a task that previously took around half an hour to two or three minutes.

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

Other companies are applying AI to inspections and maintenance. Bentley Systems says its Blyncsy road maintenance platform analyzes driver-captured images to identify issues such as damaged guardrails, overgrown vegetation and debris, while Dutch software company Ultimo says its tools scan work orders and shift logs for signs of safety incidents that might otherwise go unreported.

Evidence gap clouds safety impact

Academic researchers say the main challenge is moving from detection and efficiency gains to proving that AI actually reduces harm. Waleed Umer, an assistant professor at Northumbria University, says many studies focus on what AI models can spot rather than whether they prevent injuries, and he warns that changing site conditions, false alarms and overreliance on automated warnings can all limit effectiveness.

Billy Hare, a professor of construction management at Glasgow Caledonian University, says pilot projects have largely centered on productivity improvements, often assuming that lower exposure to hazards means lower risk. He says that assumption is reasonable, but proving a direct effect on accident prevention needs more investigation.

Bentley Systems says one rail project it supported reduced physical site visits by 50% and manual surveys by 70% using a digital twin. According to the UK Health and Safety Executive, construction worker deaths in work-related incidents in Great Britain fell to 25 in 2025/26 from 35 a year earlier and below the five-year average of 37, but Hare says the decline may reflect economic slowdown as well as any technology effect.

Researchers and executives also say implementation matters as much as the software itself. Umer says human oversight may still be insufficient if AI generates more information than supervisors can realistically review, while Ultimo Chief Executive Steven Elsham argues companies should use time savings to improve safety, first-time fixes and reliability rather than simply push for more output per worker.

In our earlier article on a KPMG U.S. CEO survey, we noted that many large companies are already seeing efficiency and growth gains from AI agents, even if only a small share considers them truly transformational so far. The piece also highlighted that progress depends on harder operational changes—process redesign, data readiness, and hiring and retraining—alongside continued investment in AI infrastructure.

This material may contain third-party opinions, none of the data and information on this webpage constitutes investment advice according to our Disclaimer. While we adhere to strict Editorial Integrity, this post may contain references to products from our partners.