Artificial intelligence (AI) is rapidly becoming operational in workplace safety programs, but most safety professionals say clear guidelines should be set, according to a report released today by the National Safety Council and Wolters Kluwer Enablon.

The Safety Shift: EHS Readiness in 2026 reveals the status and perceptions of AI across core environmental, health and safety (EHS) activities, including incident prediction, hazard identification, compliance, analytics and reporting. However, this excitement is tempered by caution regarding overreliance and the importance of human judgment in decision-making. 

While there is strong interest in adopting AI in EHS activities, more than half of the surveyed organisations have not yet established the fully digital foundation needed.

"Digital tools and AI are becoming more commonly embedded in workplace safety programs, but findings show organisations are still working to connect systems, improve data quality and balance new risks alongside operational responsibilities," said Jay Vietas, senior director of research at NSC. "Effective adoption of these principles will be instrumental for EHS professionals to make workplaces safer in the years ahead."

The report is based on survey responses from 1,053 professionals across EHS, operations and risk management roles in U.S. industries.

Key findings include:

Concerns are widespread: 90 per cent report concern about the implications of AI use, including 65 per cent citing overreliance as a key risk

AI adoption is now mainstream across EHS programs: 20 per cent report extensive AI integration within EHS programs, while 62 per cent report moderate or limited use

Readiness is high but uneven: More than 80 per cent say their organisations are mostly or fully ready to deploy AI, including 42 per cent who consider themselves fully ready

Digitisation remains incomplete: Only 11 per cent report being in fully digital EHS systems; 71 per cent operate in hybrid environments that combine digital and manual workflows

Despite progress, most organisations remain in transition, limiting the effectiveness of AI due to fragmented data and inconsistent systems.

"Organisations have made meaningful progress digitising foundational safety processes, and many are now using AI to analyse and apply that data," said Raj Jayaraman, Vice President, Product Management, Wolters Kluwer Enablon. "What the results make clear is that technology alone is not enough. Effective EHS programs depend on strong data foundations, clear governance and maintaining human accountability as analytics and AI become more widely used."

The report also highlights how organisations define risk. While 87 per cent agree mental health belongs within the EHS mandate, near-term priorities remain focused on operational risks such as fatigue, infectious disease and an ageing workforce.