VelocityEHS research finds most safety professionals now trust AI-driven insights, even as measurable return on investment trails behind
Most safety professionals now trust artificial intelligence to guide their decisions, yet far fewer can point to measurable returns, according to new research from workplace safety software company VelocityEHS.
The 2026 EHS 360 Benchmark Report, released by the environmental, health and safety (EHS) software firm, found that 70 per cent of EHS professionals trust AI-generated insights to inform their safety decisions, while 76 per cent believe AI can meaningfully reduce the administrative burden that weighs on safety teams. The report draws on a survey of 1,008 EHS professionals conducted in March 2026, combined with anonymized usage data from the company's software. A further 44 per cent identified AI and automation as the single biggest force shaping the future of the profession.
Trust is outpacing measurable returns
Matt Airhart, chief executive officer of VelocityEHS, said the findings confirm a shift that many in the industry expected would take far longer.
"EHS leaders have definitely signalled that they trust AI to help them run their businesses better," Airhart said in an interview with Canadian Occupational Safety. "They trust the results that they're getting from the AI tools that they're using. That's a really good shift, because you obviously need the trust before you can get the adoption."
Airhart said VelocityEHS had been "a little bit ahead of the curve," building AI-based research and functionality for roughly four years before the technology gained broad momentum. The question now facing Canadian teams mirrors the ongoing debate over AI adoption across Canadian workplaces.
Why the return on investment gap is a matter of timing
Trust, however, is not yet translating into hard numbers. The benchmark report found that only 34 per cent of EHS professionals report measurable return on investment (ROI) from AI, a wide gap from the share who say they trust the technology.
Airhart attributes that distance to timing rather than disappointment.
"I think that gap is probably a timing gap," he said. "They've probably adopted AI tools effectively at most within the last 12 months, and in many cases probably in the last three to six months. Any ROI is hard to measure, calculate and report on in less than a 12-month period."
He said hard cost savings, such as lower workers' compensation claims, take a full reporting cycle to surface, while efficiency gains inside safety departments are harder still to quantify, a challenge tied to how safety leaders are measuring the return on their safety investments. He expects the picture to change quickly.
"My prediction is that if you ask this question again in 12 months, and we do plan to do this benchmarking report annually, it will probably tick up pretty meaningfully," Airhart said.
Where safety teams are putting AI to work
The report also drew on usage data from the company's base of more than 15,000 customers. Airhart said more than 80 per cent of its users worldwide had adopted its potentially serious injuries and fatalities (PSIF) tool, despite it being on the market for only about six months.
He tied that rapid uptake to a problem the profession has struggled with for decades.
"Over the last 50 years, industrial safety has seen a pretty drastic reduction in minor injuries," Airhart said. "But if you look at the serious injuries and fatalities rate over those 20, 30, 50 years, it has not really reduced at all. People jumped in with both feet because they see an opportunity to solve a problem that hasn't been solved in a couple of generations of professional safety. And it works."
Airhart said an AI-assisted job safety analysis tool was also seeing heavy use, allowing workers who are not safety specialists to complete assessments while being coached through hazard identification and controls, part of a broader shift that includes the expanding role of wearable and sensor technology on Canadian worksites.
Keeping people at the centre of AI adoption
Airhart returned to the theme of trust, arguing that credible AI in safety must be purpose-built and kept firmly under human control.
"We're doing models that were trained on actual decades of safety data," he said. "It gives people coaching and suggestions and tries to flag potential risks that a person might not see. But we don't make decisions, and I think that is key. We're not here to displace EHS professionals. We're here to augment them and make them better."
