How AI Is Changing Safety on the Jobsite And WhatIt Still Can't Replace
- Chris Fredette
- Apr 14
- 4 min read
The tools are real, the data is compelling, and adoption is accelerating fast. But a camera that flags a missing hard hat still can't have a conversation with a struggling crew member at the end of a long shift.
For years, "AI is coming to construction" was the kind of sentence that showed up in conference decks and then got filed away. Not anymore.
In 2026, the shift is real and it's moving fast. A recent industry report found that 38% of contractors now report measurable business impact from AI which is up from just 17% the year before. Platforms like Fyld, which analyzes jobsite video footage for safety risks, reported 82% year-over-year growth in 2025 and counts firms like Kiewit among their users.
Contractors using it report reductions in serious incidents of up to 48%.
Those numbers are worth taking seriously. As a safety professional, my job is to protect workers and if a technology is genuinely doing that, it deserves a clear-eyed look. But so do its limits. Because right now, there's a lot of noise around AI in safety, and the hype tends to outpace the reality on the ground. Let's dive into it a bit!
"AI is moving from pilot programs to essential tools on every job site. 2026 will be the year it proves its value but also the year we find out what it can't do."
What AI is actually doing well on jobsites right now
The strongest use cases aren't futuristic. They're happening on active sites today!
Here's where the technology is genuinely earning its place:
PPE compliance monitoring
Computer vision cameras detect missing hard hats, high-vis vests, and safety glasses in real time and push alerts to supervisors before a violation becomes an incident. I saw this first hand on a project with Shell.
Predictive risk analytics
AI models trained on historical incident data can flag elevated risk conditions: fatigue patterns, high-hazard task overlap, weather triggers before the shift begins.
Proximity & exclusion zones
Wearables combined with AI geofencing alert workers when they enter high-hazard zones near heavy equipment, overhead work, or energized systems.
Documentation & reporting
Voice-to-text field notes, AI-assisted inspection reports, and automated incident logs are cutting administrative time freeing safety pros to spend more time in the field.
The thread connecting all of these is consistency. AI doesn't get tired at the end of a 10-hour shift. It doesn't develop blind spots from familiarity with a crew. It monitors continuously, across an entire site, without the bandwidth constraints a human observer has. That's genuinely valuable.
What it still can't do
Here's where I want to push back on some of the more breathless coverage. AI is a powerful tool. It is not a safety program. And the difference matters.
Reading a crew's state of mind
A camera can flag a missing hard hat. It cannot tell you that a worker just got off a 14-hour shift, is three days behind on sleep, and is working through a personal crisis. But an experienced safety pro can and we all know that situation. That awareness changes everything about how you manage risk in the moment.
Building a safety culture
Culture is built through relationships, trust, consistency, and accountability over time. No algorithm builds that. The foreman who stops a task, has a real conversation, and doesn't move on until the hazard is genuinely resolved, that's irreplaceable.
Adapting to novel hazards
AI systems learn from historical data. They are not well-equipped for genuinely new conditions, a first-time task, an unusual sequence of work, a site layout that doesn't match the training data. Judgment in unfamiliar situations still requires human expertise.
Accountability and authority
When a hazard is identified, someone has to own the stop-work decision. AI can surface the alert. It cannot stand in front of a foreman and back it up. That authority and the willingness to exercise it lives with trained safety professionals.
Worker trust
Multiple studies have flagged a real concern: workers may perceive AI cameras and wearables as surveillance tools, not safety tools. I have seen this first hand with some AI security camera systems we have used on projects. That perception erodes the trust that a safety program depends on. Managing that relationship requires skilled, credible humans in the room. And can erode the trust that management earns from its workforce.
THE LIABILITY QUESTION NOBODY'S TALKING ABOUT
Legal experts are beginning to argue that as AI-powered risk forecasting becomes more available, firms that fail to adopt it could face greater liability exposure after incidents essentially, if the tool existed and you didn't use it, that's a problem. That's a significant shift. It also means that deploying AI poorly, or without proper human oversight, carries its own legal risk. The technology and the accountability are moving together.
How to think about AI as part of your safety program
The right frame isn't "AI vs. safety professionals." It's "what should AI handle, and what should humans handle?" Use computer vision to catch PPE gaps at scale. Use predictive analytics to prioritize where your safety resources go. Use documentation tools to reduce the administrative burden that keeps safety pros stuck behind a desk. Then deploy that freed-up time back into the field, building relationships and making judgment calls that no algorithm can make.
Construction recorded 1,075 work-related deaths in 2023, more than any other industry. Falls, slips, and trips alone accounted for nearly 40% of them. If AI tools can meaningfully move that number, we should use them. But workers aren't dying because there's a gap in the camera coverage. They're dying because of failures in culture, training, communication, and accountability. Technology can support all of those things. It cannot replace them.
AUTHOR
Chris Fredette, CHST
Chris is the founder of TriCore Safety, a construction safety consulting firm. With 8+ years in the industry and a Certified Health and Safety Technician credential, he works with contractors and project owners to build safety programs that actually stick in the field.
Comments