The Ethical Question: AI In Worker Safety

Julia Penfield, Ph.D., Chief AI Officer at VelocityEHS, develops AI systems that identify workplace risks before serious incidents occur.

For over two decades, safety professionals have worked to reduce the most serious workplace incidents. We’ve made progress on lower-severity injuries, but the number that matters most has barely moved: In the United States, a worker dies from workplace conditions roughly every two hours. That rate has been essentially flat for 15 years. Despite collective effort, this is still unacceptable.

The plateau isn’t a mystery. Preventing serious injuries is fundamentally limited by time and human resources. Before a job is performed, someone must understand it, identify its hazards and plan controls to reduce or eliminate risk. If a task is repetitive, they must judge whether the workstation will wear the worker down through recurrent strain. When something goes wrong, someone must find the root causes and corrective actions to prevent recurrence. In a plant with thousands of chemicals or hundreds of contractors, someone must manage it all.

In a perfect world, we’d have unlimited experts, time and budget to do these things well. But we don’t.

New EHS 360 Benchmark Report data found that most EHS professionals say limited budgets and staffing constraints undermine their ability to do their jobs effectively. So, the work goes undone, and workers get hurt.

I’ve seen what this costs up close. Before I worked in software, I spent two years at a hydroelectric power plant in British Columbia, 600 feet underground, surrounded by heavy machinery and stored energy that can change a life in an instant. It’s where I learned what safety actually means to the people in steel-toed boots, and it became part of who I am.

It taught me something that still shapes how I build. You cannot close this gap by asking people to work harder. You need to provide them with better tools.