Construction safety has made significant strides over the past decade, but the industry’s fatality rate remains stubbornly high. New technologies—including artificial intelligence—are giving contractors a way to move beyond reacting to incidents after they occur and instead identify risk before accidents happen.
Construction Executive spoke with Josh Kanner, vice president of AI and analytics strategy for Oracle Construction and Engineering, about why predictive safety programs are gaining momentum, how AI is changing field operations and what contractors should know before implementing the technology.
Construction continues to rank among the most hazardous industries. Why has improving jobsite safety remained such a difficult challenge?
It’s a stubborn problem. For years, safety leaders have talked about “shifting the curve”—reducing both the frequency and severity of incidents while also bringing down the industry’s fatality rate. Despite significant effort, we’ve been relatively flat for nearly a decade.
Part of the challenge is that construction processes don’t change quickly. Introducing new technologies and new ways of working into an active jobsite isn’t easy. On top of that, every project brings together multiple contractors and trade partners, each with their own systems and safety practices. Aligning everyone around a consistent approach can be difficult.
One encouraging trend is seeing owners become more engaged in safety. Whether through owner-controlled insurance programs or broader project oversight, we’re seeing stronger alignment across project teams. That shared responsibility creates a better environment for improving outcomes.
How are predictive AI tools changing the way contractors think about safety compared with traditional approaches?
Traditional safety programs have often been reactive. They tell you what happened after an incident or near miss. Predictive AI changes that by helping project teams understand where risk is likely to emerge before something happens.
We’ve seen organizations reduce incident rates by as much as 50% within the first year of implementing predictive safety programs because they’re able to focus attention where it matters most.
The predictive models evaluate dozens of variables simultaneously—typically between 50 and 75 factors—including weather conditions, manpower levels, supervisory ratios, project activity and safety observations. Every one of those variables is something safety professionals already recognize as meaningful. AI simply analyzes all of them together and identifies patterns humans can’t easily see.
Equally important is the cultural shift. Predictive safety encourages more engagement from everyone on the jobsite. Instead of safety being something owned exclusively by inspectors or safety managers, it becomes everyone’s responsibility. That increased participation strengthens safety culture while also producing better data for future predictions.
What does implementation actually look like on a project?
There are really two components.
The first is an observation-based safety program. Workers use a simple mobile application to document both positive observations and potential risks while they’re in the field. At Oracle, that capability is integrated directly into Aconex and Unifier, so it becomes part of normal project workflows rather than another standalone tool.
The second component is the predictive engine itself. We collect data such as manpower, project location, weather forecasts and observation trends using a standardized data template. That information feeds an industry model we’ve already trained using more than 10,000 project-years of historical safety data, including incidents and near misses.
The technology is important, but executive commitment is equally critical. Organizations need leadership that prioritizes operational improvement and safety. Without that top-down support, any technology implementation becomes much more difficult.
Observation-based safety is becoming increasingly common. How does that differ from the old checklist approach?
It’s almost the exact opposite of where the industry was 10 or 15 years ago.
Back then, safety often revolved around lengthy inspection checklists completed by dedicated safety personnel. Those inspections frequently pulled people away from the field because they had to return to an office to finish paperwork. Safety became something done by specialists instead of something owned by everyone.
Observation-based safety is designed to be lightweight. A field employee can complete an observation in just a few moments using a mobile device, even offline. They identify whether they’re recognizing positive behavior or documenting a potential hazard. If it’s a hazard, they categorize the risk and rate both its potential severity and how frequently workers are exposed to it.
That information becomes valuable both for immediate follow-up and for the predictive models that identify emerging risk across the project.
How receptive have field teams been to this technology?
Like any technology, there’s always some resistance to change. Construction teams already have a tremendous amount to manage, especially once a project is underway.
What we’ve found, though, is that results speak for themselves.
One feature that resonates particularly well with field personnel is that the system can recommend how many observations should be completed each week based on actual project conditions. Because the model understands labor hours, project phase and overall risk, observation targets adjust dynamically.
That means we’re not simply telling every project manager to complete five observations every week regardless of what’s happening onsite. The recommendations reflect actual project conditions, making the process feel much more relevant. Perhaps most importantly, workers can immediately see how the observations they’re submitting influence weekly risk predictions. They’re not collecting data simply because management asked them to—they’re contributing directly to safer project decisions.
Can predictive AI also improve communication between the office and the field?
Absolutely. At the enterprise level, safety and operations leaders can compare projects across an entire region or business unit and quickly identify where they should focus their attention.
At the individual project level, every team receives its own weekly risk assessment along with recommendations for reducing exposure. They can also compare predictions week over week and determine whether recommended actions were implemented and whether those changes improved the project’s overall risk profile.
The technology isn’t asking teams to reinvent their safety programs. It’s helping them make better decisions using information they already have.
Many contractors worry about implementation and adoption. How do you balance safety improvements with productivity?
Improving outcomes almost always requires changing behavior to some extent. You can’t expect better results without doing some things differently.
That said, we’ve worked very hard to minimize the implementation burden. Our goal is to leverage systems companies are already using rather than asking them to recreate processes from scratch.
For example, manpower information, observation data and other operational metrics already exist within most organizations. We simply work with customers to map those data sources into a common structure that supports predictive modeling.
Lowering that activation threshold makes adoption much easier because companies aren’t duplicating effort. When leadership decides it’s time to improve safety, we want implementation to be as straightforward as possible.
After more than two decades working in construction technology, what continues to motivate you—and what excites you most about the future?
What keeps me excited is the people. Construction is filled with builders who genuinely want to create something meaningful. I build software, but it’s still part of helping people build the world around us.
The industry is full of smart, practical, collaborative people, and that’s something I’ve always appreciated.
Looking ahead, I’m incredibly optimistic about AI’s potential. This is a technology shift on the scale of the internet, mobile computing or cloud technology. At the same time, construction faces enormous challenges: growing workloads, an aging workforce, the loss of institutional knowledge and increasingly complex projects.
AI can help address those challenges by improving how information flows through organizations and helping people make better decisions faster.
At Oracle, we’re combining decades of construction expertise—from scheduling and payments to document management, project controls and analytics—with major investments in AI. Bringing those together creates an opportunity to solve problems the industry has struggled with for years.
It’s an exciting time because AI isn’t replacing the people who build our world. It’s helping them build more safely, more efficiently and with better information than ever before.
SEE ALSO: AI IS REACHING THE JOBSITE: IS YOUR WORKFORCE READY?
