While artificial intelligence promises to transform many occupations, the technology can also be used to better understand the workplace and even lower safety risks.
One electrical contractor using A.I. and machine learning to improve awareness of workplace hazards and inspire a more engaged workforce—which has been shown to decrease injuries and fatalities—is Cupertino Electric Inc., a national electrical engineering and construction company based in San Jose, Calif.
For the last six years, Cupertino Electric has been using A.I. and machine learning, along with human observation, to assess workplace hazards on its job sites. In 2019, Cupertino Electric began to work with FactorLab, Pleasanton, Calif., a developer of people-centered applications and data science platforms designed for continuous improvement and injury-free operations, according to its website.
Cupertino Electric trains its safety professionals and field technicians on the use of FactorLab’s SmartTagIt, an application for sharing and mitigating potential hazards whenever they notice, find or fix a workplace hazard. Since 2019, Cupertino Electric’s field workers and safety professionals have input nearly 74,000 field observations.
“We had to be comfortable with people telling us bad news, and we want to be sure the field trusts us when they give us information we might not want to hear,” said Brent Bowers, corporate safety manager at Cupertino Electric. “If we don’t have trust, we’ll just get a facade instead of the truth.”
Thousands of construction workers and hundreds of companies use the SmartTagIt application every day on more than 5,000 work sites, according to Barry Nelson, FactorLab’s president and CEO. The application can collect observations through text, video and audio message. It automatically transcribes and translates Spanish into English, and if a recording is captured in a noisy atmosphere, the application is able to filter that excess noise out—both new advancements within the past year or so.
Now, those workplace hazard observations make up a large database, which safety professionals can comb through with A.I. to gain a better understanding of the hazards, assess the level of engagement among workers in the field and find out the baseline level of observations. Hazards will always exist on job sites, Bowers said, but the A.I. tool can help assess how often hazards should be occurring (what a typical rate is) and when there is a higher-than-normal number of incidents. The goal is to use this information to help prevent injuries.
Better than a stagnant assessment
Before working with FactorLab, Bowers said the safety department at Cupertino Electric was a third of its current size and the company was using iAuditor, an inspection software and mobile app.
“They were very stagnant assessments, form-based without observations,” he said. And the observations being input “weren’t that good, frankly.”
Six years later, “we’ve trained folks in hazard identification. We’ve taught a model to make things more reliable than they ever were before,” Bowers said. “That’s helped people see the value, because the tool is giving accurate data. That’s the reason we’re seeing people adopt it more.”
“We can listen now,” Nelson said. “We’ve given [ECs] a phone and told them, ‘Just talk about what’s happening and what you saw and what you learned.’ You start collecting symptoms and problems.”
A.I. has helped the EC find and create topics out of the massive database of field observations. Viewing hazard observations through A.I. in this way enables them to better understand types of hazards, how often STCKY (stuff that can kill you) hazards occur, where and how often they are happening, and more.
This helps Bowers and his colleagues see patterns and group observations into topics, such as limited (not enough detail describing the hazard), better (sufficient detail describing the hazard), STCKY and non-STCKY.
“It’s a new lens to look through,” Bowers explained, “as opposed to picking apart every single event.”
“We entered a whole other world with observations,” he said. “We turned the corner from a reactionary department to more forward thinking, being able to anticipate some of these things.”
Previously, the data was always old by the time it was put into charts and presentations, Bowers said.
“Now, we have data that isn’t a PDF on my desk,” he said. “We now have the ability to look at this information in real-time to identify trends and act on them.”
The SmartTagIt system has enabled Cupertino Electric to create risk profiles for each area of the company to see which teams are more engaged in inputting hazard observations.
“When it comes to A.I. taking over, we can’t do it ourselves. We don’t have enough people” to run an advanced model like the A.I. can, Bowers said.

FactorLab’s SmartTagIt system is an additional safety resource for companies, just like the energy wheel and the STCKY wheel, but with the added benefit of A.I.
“The novel thing Cupertino Electric is doing is using A.I. to better understand the quality of and ultimately improve the health of safety systems, so there are better accident outcomes,” Nelson said. “Almost everyone has an observation system where they collect details from the field, but they struggle with what that information means. We built this model in a way that A.I. can pick out serious versus non-serious hazards.
“The industry is learning, but Cupertino Electric is pushing ECs forward,” he continued. “It’s nudging us to find better ways, to do what has been beyond our grasp.”
Increased engagement, lower incidents
While it’s not a foolproof formula, Bowers found there is a correlation between the number of observations that workers and safety professionals are putting into the system and lower workplace incidents.
Cupertino Electric assessed the number of employees inputting workplace hazards and incident rates over a four-week period. What the contractor found was that as engagement numbers rise, incident rates decrease.
“Engagement can be a good proxy indicator of cultural state,” he said. “And we want employees more engaged than just during their weekly safety meeting.”
Cupertino Electric has seen success with this safety system because it involved the field. The best performing division, in terms of highest number of observations, actually has the fewest number of safety individuals in it, Bowers said.
“The safety department can’t do it alone,” he said. “We need the field’s engagement because they outnumber us probably 75 to 1. We want to see the field engaged more than the safety department to look out for one another.”
Be aware of the source when collecting observations. Trained safety professionals’ reports will be high quality. Observations from the field may have less detail, but will provide a fuller picture of the safety reality, Nelson said.
While there are already safety professionals who do this job well, Nelson explained how critical it is to make it easier for people in the field to observe and call attention to workplace hazards.
“What we at FactorLab want is to encourage more people like Brent to communicate with each other and talk collectively about how to remove constraints, what does good look like, what are they doing to drive trust, what are they doing to make sure people understand they—and their hazard observations—are valued,” he said.
Safety professionals have to give themselves permission to look at problems differently and know they may not get it right the first time, Nelson continued.
“I think there is a movement afoot, a fundamental shift,” he said. “Everyone is on their journey.”
Nelson wants to meet electrical contractors wherever they are, to introduce them to A.I. and help them see how it can improve safety culture or be an added tool in their already well-built safety tool belt.
Nelson and Bowers assured that when using the application, the user is in control—the A.I. does not control you.
“We’re not saying A.I. is the be-all, end-all,” Nelson said. “We’re saying this is a logical way to look at the information you’re collecting and have different conversations.”
It took FactorLab and Cupertino Electric many times over several years to get the technology right. But they found that it helps them look at information differently, enables them to do something that wasn’t possible years ago.
“What we’re saying is, borrow this. This is a model everybody can use,” Nelson said. “Look at it and go build your own. The value of A.I. is the cost and time to build models has been dramatically reduced; the math hasn’t changed. We don’t have to be constrained by the current shackles that just said ‘hazard, hazard, hazard; control, control, control.’ I think technology will help transfer knowledge and really show people what [hazards and safety] look like.”
The promise of technology
The next phase: using A.I. to gain a better understanding of why hazards are happening, what led up to them occurring, where and when they take place most and similar questions, Nelson explained.
“This would be understanding precursors,” he said.
For instance, Bowers asked whether the “better” category of observations would be required to prevent problems. “Is this truly the information I need to prevent problems, or do I need to evolve my system?” he said. “Information like why did this happen, what were you thinking at the time, etc.—the new technology allows me to collect more information than ever before.”
“This is the promise of technology,” Nelson said. “It’s not the charts it generates, it’s what Cupertino Electric is doing [with the technology, with what they learn] that is going to move things forward. They’re brave enough to use these emerging capabilities to think about and solve problems that previously were simply too challenging to tackle.”
stock.adobe.com / cofficevit/ J.V.G. Ransika / kiimoshi
About The Author
Chertock is a poet and renewable energy and science journalist in the Washington, D.C., area. Contact her at [email protected].