Smart Construction Technology: How Machine Learning Predicts and Prevents Construction Safety Incidents

November 9, 2018 Josh Kanner

Technology is changing everything about the way we build. From 3D modeling software to UAVs to VR technology, the benefit of new construction technology on the worksite is increasingly undeniable. Time saved, a reduction in waste, and an increase in quality--the rising tide of innovative technology really is lifting all boats. With construction safety being a major concern across all facets of the industry, it only makes sense that new technology can create opportunities for improvement.

Predictive analytics are the leading edge of Machine Learning and Artificial Intelligence (AI) applications and their capabilities are just now being explored in the construction industry. is leading the charge with the application of this technology in AEC systems and construction workflows, enabling users to mine existing data and images from worksites to gather even more data and increase safety onsite. built their technology to seamlessly integrate with Autodesk BIM 360 software, which enables them to gather data direct from the source.


Construction Safety: Finding New Ways to Work Together partnered with Suffolk Construction to test the ability of AI to predict construction safety incidents on specific sites. The driving force behind this initiative was Suffolk’s “Safer Together” program, a partnership between Suffolk and contractors with the aim of promoting jobsite safety in an industry that is notorious for accidents and injuries. They focus on collaborating with subcontractors and emphasize positive safety discussions rather than negative call-outs. Their ethos is simple: “Everyone goes home safely at the end of the day.”

To test the capabilities of predictive analytics, used historical data from Suffolk going back ten years, which was fed to “Vinnie,” their artificial intelligence engine. The idea is that Vinnie would be able to compare this data with current jobsite images, incident reports, and environmental data. By tracking things like personal protection equipment (PPE) usage, weather, project type and phase, and comparing that data against incident reports, Vinnie was able to generate a predictive model for safety issues.

Vinnie’s predictive model would support “Safer Together” by tracking overall compliance with observed use of PPE and drive positive reinforcement of safety protocols and compliance, then issuing warning alerts to site managers.


Construction Technology: Predicting the Future by Studying the Past


Some companies are moving AI beyond simple observation and are instead training AI to predict safety issues


Before launching into a real-time test, Vinnie was tasked with predicting incidents over a 3 year period where the predictions could be compared to what was known to have happened. A joint-release from and Suffolk lays out the numbers and shows some real possibilities in risk-reduction via AI predictions. 

  • Over the 3 year period, Vinnie predicted 20% of all incidents at an 80% accuracy rate, when the manager elected to receive 4 alerts/year, with one of the 4 being a “false alarm.”
  • If the manager was open to more alerts, they could be warned of a full 40% of incidents, with 66% accuracy (2/3 of predicted incidents occurred, or roughly 12 alerts/year, with 4 being “false alarms”).


In trying to predict the potential impact of the early warning system, a conservative assumption is useful.

  • If only 1 in 4 predicted incidents are avoided, a company with 50 projects/ year can avoid 40-100 incidents/year.
  • At an estimated cost of roughly $36,000/incident (in 2018 dollars,) that’s somewhere between $1.4M and $3.6M in safety related savings per year.
  • If we assume that 50% of the alerts from the “early warning system” result in preventing an incident (a more likely assumption), the financial benefits are double this amount.



If just 1 of 4 predicted incidents is avoided, that translates to $1.4-$3.6 million in savings


These numbers show that it is possible to use data and AI to create a predictive safety management system, based on what we already know, and incorporate statistical probabilities to help reduce safety risk. 

It’s also worth noting that while an elevation in safety clearly presents a cost savings, it also stands to reason that creating a safer work environment promotes talent retention in an industry that is already feeling the stress of filling the “skilled labor” gap.


AI Driving the Business Model of Tomorrow

Where do we go from here? The possibilities of artificial intelligence go well beyond taking the guesswork out of safety management. As we continue to study these areas and provide more high-quality data, the predictions are likely to be even more precise. While these predictions could well be used to drive better rates for insurance, that’s just the tip of the iceberg. There are opportunities to explore how predictive analytics can be used to manage all areas of project risk.

Safety, quality, cost and schedule are the four critical drivers of project success. Could Vinnie start predicting early warnings of cost overruns before they happen? That’s one prediction that may be too early to make. To learn more about predictive analytics, and how the “Vinnie” and the early warning system from can be implemented on your next project

- eBook, "How suffolk and learned to predict and prevent construction incidents"

Previous Article
4 Practical Steps to Evolve Your Construction Quality Control Plan
4 Practical Steps to Evolve Your Construction Quality Control Plan

When it comes to winners and losers in the digital age, one of the biggest potential winners is your constr...

Next Article
Construction Daily Reports: The Value of Software for Connectivity on the Jobsite
Construction Daily Reports: The Value of Software for Connectivity on the Jobsite

Cliché terms are often cliché for a reason. They can hold timeless truth. So, I'm going to throw a cliché p...