“Can AI predict—and thus help prevent—safety incidents?”
Our team was posed this simple question in 2017. We’d seen that our AI engine could identify safety risk indicators in job site images. Now we wanted to know: could these visual indicators be combined with other data to help avoid incidents altogether?
As I’ve written about previously, this simple question kicked off 12 months of collaboration between major national contractor Suffolk and Newmetrix (formerly Smartvid.io). Suffolk contributed 10 years of photo and project data from the company’s internal systems, including Autodesk, Procore, Oracle and Oxblue. Newmetrix analyzed the images, then fed additional project data (such as project type, weather and phase) into a multi-layered machine learning predictive model to see if it was, indeed, possible to predict incidents on job sites.
The results were compelling. As published in this case study, we predicted roughly one in five safety incidents a full week ahead of time with 81 percent confidence. In other words, if our AI engine, “Vinnie,” thought an incident was going to happen, it happened—more than four out of five times. We shared these results at conferences like ENR FutureTech, Autodesk University, and Procore’s Groundbreak, and they were featured in MIT Tech Review.
This industry attention validated our first phase, and showed that predictive analytics is possible through collaboration. However, this work is not about one company’s risk; it’s about reducing risk across our industry. And getting there will require partnership, resources, data and analytical horsepower. We’re lucky to have each of these elements on our side as we enter phase two of our work to reduce safety risk through predictive analytics.
“This work is not about one company’s risk; it’s about reducing risk across our industry.” - Jit Kee Chin, Chief Data Officer, Suffolk
The partners
Suffolk (as council chair), Barton Malow, Clayco, DPR Construction, JE Dunn, Messer Construction Co., Mortenson, Shawmut Design and Construction, and the Bouygues Group of France comprise the inaugural Predictive Analytics Strategic Council. In addition to these participants, the council also welcomes Aon as an inaugural member from the insurance ecosystem. Newmetrix will serve as the technical chair of the council. The council’s mission is to learn and explore by doing—building new models for areas of risk and actively discussing operational, social and business impacts along the way. All members have pledged to provide data to help advance the cause of predictive risk in the architecture, engineering and construction (AEC) industry.
And that’s not all. Technologies fail if they are pursued without careful consideration of the people and processes they affect, no matter how valuable they may be. In addition to gathering data, the council’s goal is to chart the course through these non-technical obstacles in order to help bring the benefits of predictive analytics to the AEC industry as a whole.
The resources
As of 2/15, we’ve closed the first tranche of a strategic funding round led by Suffolk, Leawood Venture Capital and Building Ventures. This support, in addition to the council’s guidance and data will accelerate development for the predictive analytics module and more. Leawood Venture Capital’s investors include many notable owners and executives from some of the top national engineering, design, and construction firms based in the midwest.
The next step
With its founding members in place, the predictive analytics council will continue to grow. Interested in contributing data and strategic input? Reach out to our team. We’ll also continue to work with Autodesk, Procore, OxBlue and our new partner, Oracle, to advance our mission to apply AI to construction risk.
Our simple question has turned into a collaborative initiative between ten companies (and counting). The work to find answers will keep us busy for years to come, with implications for the industry as a whole to reduce risk for safety and beyond.
Read how Suffolk is using AI and Newmetrix to predict and prevent incidents >
