How one construction company is using AI to make workers more safe

• Suffolk is the largest construction contractor based in Massachusetts.
• The company built an artificial intelligence model that predicted safety risks on construction sites.
• Insights from the model reduced incident rates by 25%.
• This article is part of "AI in Action," a series exploring how AI is being used across different industries. 

Suffolk is a construction company headquartered in Boston, Massachusetts. It's the largest construction contractor in Massachusetts with additional locations in California, Florida, New York, and Texas.

Situation analysis: what problem were they trying to solve?

According to Kelsey Gauger, Suffolk's national director of operational excellence, the construction industry could experience a "brain drain" in the next five to 10 years.

The construction sector is facing a labor shortage, with more than half a million additional workers needed on top of the average pace of hiring to meet labor demands. As well as struggling to hire new workers, the industry is set to soon lose many of its current ones, as nearly a quarter of construction workers are older than 55.

"All of our wisdom is going to be leaving the industry," Gauger told Business Insider. "[We have] younger people that are from the next generation that have to get up to speed really quickly with best practices."

Construction is also a high-hazard industry. In 2021, nearly one in five workplace deaths occurred in the construction sector.

Informed decision-making is a key part of ensuring safety standards are upheld on construction sites. However, with its most experienced workers aging out of the industry, Gauger said that Suffolk wanted to shift from "intuition-based decision-making" to "data-driven decision-making."

"Safety's not about numbers, or about technology, it's about people. But there's ways to use data and numbers and technology to make those people safer," Gauger said.

Key staff and partners

Suffolk appointed its first chief data officer, Jit Kee Chin, in 2017, and has a team of 30 data analysts. Having a strong internal team enabled the company to do a lot of the development in-house.

Suffolk also sought out external partners. Suffolk Technologies Group, a venture capital group associated with Suffolk, invests in construction startups and uses Suffolk jobsites to test out potential solutions. One of their first investments was in NewMetrix, formerly known as SmartVid, a startup based in Cambridge, Massachusetts. Suffolk then worked in partnership with NewMetrix to develop a safety solution that leverages artificial intelligence.

AI in action

In order to maintain safety on jobsites, construction workers conduct safety observations, where a team member records any incidents that could pose a safety risk, such as someone not wearing the appropriate protective clothing.

First, Suffolk worked with NewMetrix to digitize its safety observations and incident data.

"The shift from manual and analogue to digital enabled us to collect data, and that's the most important piece of all of this," Gauger said. "It plays into our data strategy."

The next step was putting this data to use. NewMetrix developed a predictive model that leveraged artificial intelligence to give a risk rating to jobsites and to predict the likelihood of an incident occurring.

One of the model's findings was that if one safety observation occurred for every 150 hours worked on a job site — that being, if there are 10 people working a 15 hour day, 150 hours will accumulate — then incident rates will be driven down significantly.

With this new insight, Suffolk changed their KPI for jobsite safety. Instead of having a target of reducing the total number of on-site incidents, they focused on the number of safety observations carried out per workforce hours accumulated.

Gauger said that one challenge has been "convincing people to trust the data" and to "buy into the information that the model is providing."

"I think just holistically, our industry does not always leverage technology and data to the fullest capacity and historically [it] has not done that," she added.

Did it work, and how do they know?

After implementing this data-driven strategy, Suffolk saw a 25% improvement rate in their total recordable incident rate in the past fiscal year.

"Over the course of the last few years since we've provided more focus on observations, we've seen an overall downward trend closer to 60% in terms of incident rates," Gauger said.

Gauger added that the machine-learning model now accounts for 40 different factors correlated with safety outcomes.

"One of the things that we've been able to do is actually flag at-risk projects, in addition to driving safety observations, so we can actually plug into that model all of this information and it can tell us, 'here are the three jobs in your portfolio that are the most at risk.'"

What's next?

Gauger said that in future, Suffolk plans to use AI to create "micro-efficiencies." For instance, Suffolk's operations team planned to roll out Microsoft's AI-powered Copilot capabilities and integrate the AI into tools that the team is already using.

Gauger added the company wanted to implement AI in "big strategic ways" as well. "We're sitting on 40 years of construction data," she said, adding that Suffolk aimed to leverage AI "to fuel better design decision-making up front."