Transforming Safety Management with AI One Step at a Time

Safety leaders, like other professionals, are being inundated with promises and pitches of various technologies that can simplify their work lives while keeping their employees out of harm’s way, and artificial intelligence (AI) is the current tech du jour. Daily incantations of the capabilities of the technology can often instill fear in employees who feel they will be replaced and create a great deal of pressure on safety managers who are required to quickly move AI into daily operations.

One time-tested strategy to determine if a technology is actually useful is to match that technology to a real-world situation. That has proven to be the most judicious way to discover early benefits that can be scaled as needed. This focused AI implementation strategy is paying off, and in this article we’ll look at two companies that offered insights on their use cases as well as their philosophies on the relationship between humans and technology.

Skanska Taps AI as a Sidekick

Skanska USA Building, which employs 6,500 and is part of a global construction company, created an internal AI tool that perfectly positions it as one that benefits users via collaboration. Introduced a year ago, Safety Sidekick operates just as its name indicates. It’s an automated tool that analyzes safety data, provides real-time safety guidance, and helps manage risks on construction sites. Who wouldn’t want a sidekick like that?

“Safety Sidekick provides a hand when someone is making a decision,” explains Brian Karas, national program director for EHS at Skanska USA Building. “It’s not making the decision.”

Karas notes that when creating this tool, the goal was to provide employees with a source that could assist with tasks they were already undertaking. Just as a sidekick can offer guidance, this tool allows for easier and more time efficient access to safety documents and processes. “When you are writing an incident report, you can use Sidekick, or when you are making a safety observation, the tool can be helpful,” Karas notes.

The tool was created specifically to increase access to information. “The primary object is to use AI to bring safety standards and institutional knowledge to real-time,” says Karas. “It brings decision-making to the field where the risk actually exists, rather than back in the office. The focus is to put the right information in the right hands at the right time.”

The tool serves the company’s overall safety strategy. “We are using AI to reduce risk exposure during high-hazard work by improving the quality of information and the timelines of safety decisions,” says Boris Kiprovski, an EHS director for Skanska USA Building.

The company uses the energy-based safety approach, which is a scientific method to preventing serious injuries and fatalities. For example, in the field when searching for a potential hazard, AI can be trained to research a scenario so an employee can document observations, and AI will provide a live answer.

By increasing consistency and quality in safety observations and classifications, Kiprovski notes that it can increase the quality of data both in his region and then be extrapolated to a national region to ensure a match across locations.

Change Management

While bringing new technology inside company walls has always been a challenge, bringing AI is turning into one of the greatest challenges ever faced by companies. That might be seen as an overstatement given the sea changes companies have experienced over the past decades, but fear of job loss as a result of AI is real. And in many cases, it is not unfounded, as several companies are explicitly pointing to AI improvements as a reason for layoffs.

However, most companies tend to view AI in the same light as Skanska does, as being yet another tool, albeit powerful, to help improve safety processes.

“AI decision support is the goal,” says Kiprovski. “Not decision-making or replacement.” Karas echoes this sentiment. “We are augmenting decision-making, and we keep our employees in the loop with every AI step we make. By allowing AI to access safety material that is based on the company’s own internal manuals, it allows employees to have quicker access to a proven knowledge base. Our AI is hallucination-free,” notes Karas. AI is not solving issues but assisting the employee who has the authority to make safety decisions.”

Kiprovski explains that setting clear boundaries for information used in AI applications is essential. “We explain that this is not ChatGPT. The information is based on our internal documents and OSHA regulations.”

And employees are trained on how to use AI through a multi-pronged approach, including small groups led by Kiprovski, as well as lunch-and-learn sessions and online training that employees can use when convenient for them.

Adoption has been positive and word is spreading from one employee to another. “People who use this tool love it and talk about it with colleagues and supervisors,” Kiprovski says. This word-of-mouth method allows for a more organic adoption.

Measuring Success

Technology and return on investment are usually closely intertwined, but given the global push for incorporating AI into organizations, there is less pressure than with other technological advances. “In the safety department, AI is able to access low-hanging fruit when it comes to expectations,” says Karas. “We are tracking its use and monitoring adoption.”

The key to adoption is that the company doesn’t position AI as a separate initiative. “This is not one more thing that employees have to do,” explains Karas. “They are already doing an incident investigation and now they are getting help from AI and therefore have a more consistent outcome.”

AI is also valuable as a method of addressing an area that all companies are struggling with—a demographic shift. “A lot of knowledge is walking out the door,” says Karas. “Experts who are able to answer questions based on experience are retiring. Trying to capture that knowledge is difficult. For example, an experienced electrical supervisor can understand an issue and plan out the solution, all in their head. With AI we can democratize that information and try to help fill that gap.”  

Cargill’s AI Use

Making information accessible is also the key driver at Cargill in their use of AI. Christina Brundage, an EHS specialist at a Cargill ground beef manufacturing plant with 350 employees, implemented an AI-based program two years ago. “My primary objective is to help enhance efficiency and consistency in our safety processes,” she says.

The first step in doing that is to reduce the administrative burden. “This will increase access, improving the understanding of information, with the ultimate goal of trend identification,” she explains.

And like Skanska, she says AI has a specific place in her safety organization. “I view AI as a partner and decision support tool, not a replacement for our professional judgment. So I highly believe that human oversight, ethical use and accountability still remain essential. Our goal is to enhance how we work and help improve safety outcomes, not to remove our human decision-making.”

For Cargill, a specific advantage of using AI is the ability to address the particular needs of the plant’s employees. The technology assists Brundage in refining safety tools for ease of use, as a number of employees at her plant have not graduated from high school. There is also a language issue in that for some employees, English is their second language, or they might not speak English at all. So Brundage uses AI to help ensure that her documents are written in a way that reflects the varied needs of the employees.

Given that her safety department consists of only two people, and cost is a concern, Brundage has become adept at finding inexpensive ways to gain access to this technology. She uses Microsoft’s Copilot app, and to further enhance capabilities, she is using Microsoft’s Power Automate and Power BI to create training reports.

“One issue that I have found, and I think that it applies across many organizations, is that there’s a lack of an effective, accessible way to track training. The solution I’m building pulls training reports automatically from our learning management system using Power Automate. And then Power BI translates it into an intuitive, supervisor-friendly dashboard, where supervisors can now see which employees are trained in specific job areas. And they’ll be able to filter by department or by supervisor. So, if someone calls off work in one department, the supervisor can access these documents to find another employee who is cross-trained and send them over.” 

She notes that this is a work in progress and part of an overall objective she has to help supervisors in the area of training. Her system enables supervisors access to training information such as the current month’s completion percentages, a year-to-date completion, overdue training percentages, and a breakdown of completion by supervisor.

Once completed, she wants to see how supervisors react to the tool. “Some people don’t realize that they are already using AI when they use Siri or Alexa. These are all machine learning tools. So, I hope to get a positive reaction from supervisors.”

Of course, the fear employees might have in terms of losing their jobs is one that must be addressed. “Many people put a lot of stock into AI and are afraid of it. I don’t think that it’s going to be able to do everything that people have dreamed of it, at least not anytime soon. I don’t think it will take away jobs. I don’t foresee it taking my job in my lifetime.”

Moving Forward

As with any technology journey, a few lessons are learned along the way. “Starting with AI requires a few basic things,” says Karas. “First, you need to digitize your data. Stick with the knowledge base that you can maintain and answer questions really well within that framework. Start by answering questions from your manual and adding job hazard analysis. This establishes credibility for those who use the system. Users must trust the source, since if you lose trust, it’s hard to get it back.”

Both Skanska and Cargill are building trust in their approach to integrating AI by providing credible information based on known safety standards, as well as talking with employees about how this is a tool to help them make better safety decisions and isn’t replacing employees.

These are examples of ways companies are exploring how AI can be brought into their organizations. The ultimate goal, however, is to significantly improve safety outcomes. “Our true return on investment will come with time and the ability to reduce SIFs,” says Kiprovski.