Artificial intelligence (A.I.) is all over the news. New technologies called large language models are front and center. ChatGPT (owned by OpenAI, San Francisco) is one example of an A.I. application that could supercharge construction productivity.  If information is key, easy access to it will be one benefit of this technology. Add in a chatbot to a large language model and you have a conversational tool. Software designers and others are adopting this and other A.I. to create new, better means for safety and other functions in construction. A revolution is near.

While the public is familiar with website customer service chatbots or virtual assistants (e.g., Siri or Alexa), today’s A.I. large language models are the next evolution by a wide mile. They don’t just gather information—they learn. The better the input information, the better an A.I. chatbot can provide accurate responses, and then go one step further by offering advice based on the knowledge it has been given. ChatGPT grows from its gathered intelligence.

Chatbots are one way A.I. can give users easy access to information. In construction, for example, it could augment safety programs, making them comprehensive, mobile and immediately responsive. 

Kris Lengieza is vice president of global partnerships and alliances for Procore, Carpinteria, Calif., which produces a collaborative cloud-based construction management platform. The company employs A.I. to improve products for its customers.

“The conversation around A.I. and safety is not new, but it’s now in the forefront,” Lengieza said. “About 96% of construction data goes completely unused, whether that’s safety data, project data or financial data. A.I. can help better utilize all that data to help reduce risk.”

A productive advance

Lengieza observed that a large share of contractors still commit to pen and paper and spreadsheets. 

“That’s very unstructured data,” he said. “Then you have folks that are putting things in what we’ll call horizontal solutions like Google Drive or SharePoint. Maybe they collect information from their mobile phones, take pictures, fill out a form on a tablet. And then we have more structured data, which is being collected in construction management software and other tools. The more structured the data becomes, the more that we can leverage things like A.I. to put it to work for us. In fact, A.I. is very tightly coupled to data collection. Know that A.I. is only as good as the information you input into it.”

Structured or quantitative data is computer-friendly. It is highly organized and can be read by machine learning algorithms, allowing for easier analysis and search capability.

For Lengieza, A.I. is a better tool than a manual. He explained that crafting an effective large language model for a safety program isn’t just a matter of scanning in National Electrical Code safety manuals.

“There are many steps to think about in building out a safety program where A.I. helps. There’s a lot of different ways data can be input (e.g., OCR [optical character recognition]) but a standardized process and enforcement of people doing things in a repeatable way vastly improves a data set,” he said. 

“For example, the NEC has a tremendous amount of content around good safety practices and probably has checklists and forms and things like that. You must make sure that people are doing things right and filling forms out correctly. Once you get that down, you now have a very large data set to look at trends, and the A.I. can help you with the analysis,” he said.

While a safety director shouldn’t have to be a data expert, they do know what a good inspection process looks like and can then leverage an A.I.-supported product to make sense of the data and provide insight to workers.  

Lengieza added that a safety director shouldn’t have to be a data expert, but they should know what a good inspection process looks like to be able to train and teach people on it, and then leverage an A.I.-supported product to make sense of the data and provide insight.

David Brennan is CEO and a co-founder of Safety Evolution Software Systems, Kelowna, British Columbia, serving the construction and oil and gas industries. He has seen various applications of A.I. in software development, including its integration into his company’s health and safety software. His firm is also working on integrating large language models into its software.

“You couldn’t upload a whole safety manual,” Brennan said. “It’s too much, but you could go section by section. You might add things like, ‘I’m in this state and this is the industry that we work in.’ You’re building context to a question you’ll ask. With A.I. trained on your company-specific and detailed data, when you ask it a question, it will provide an answer but maybe go further and ask some clarifying questions. I believe large language models and user-friendly interfaces help democratize technology so everyone can use it.”

Brennan sees several benefits to A.I., including its use in job planning, resource allocation and decision-making, which enables more proactive safety actions and analyzing data to find solutions for issues. He also likes how it can summarize complex information. 

“I spent a lot of time in my career reading regulations. It was very time-consuming, complex and sometimes confusing. ChatGPT summarizes dense information and breaks it down a bit,” he said.

An A.I. evolution

Lengieza explained that large language models are not new, but how we’re interacting with them and where they can gather information is original.

“There’s a lot of different ways that you can use ChatGPT,” he said. “For example, if you’re a safety director and recently you’ve had a lot of ladder incidents, you could go into ChatGPT and ask it to provide you with a toolbox talk script focusing on ladder safety based on job sites you’re working. Then there’s the use of large language models that are using the same technology as ChatGPT based on your data. 

“Let’s say that you are a Procore user. In the future, imagine asking for the most common safety incidents across all your job sites. You could probably get that information by pulling a bunch of reports, but now your ‘ChatGPT’ is going to go and look through your set of data and give you feedback. It also makes data conversational. It changes how you interact with it,” Lengieza said.

Beyond large language models, Lengieza cited other uses of A.I. suited for construction, including a predictive analysis that takes data sets and comes up with a prediction methodology of what might happen next. These models use a lot of historical data, including photography. 

A.I. can also leverage the internet of things (IoT) to read data captured by smart safety gear such as wearables to measure workers’ exposure to risk in the field. Lengieza finds A.I. analysis will make IoT more actionable in relation to safety, building and business operations. Autonomous machines, such as Boston Dynamics’ four-legged Spot robot, is another application of A.I. that could make its way onto job sites.

“And then, there’s a big opportunity around training,” Lengieza said. “There are some technologies out there that can create training videos in about 5 minutes, completely created with A.I. Supply a topic and script and it’ll make a video.”

The need for digitalization

Brennan found tech adoption, including A.I., increased during the COVID-19 pandemic as companies looked for ways to optimize their operations and be more productive. 

“The need for digitalization during the pandemic, especially with paper-based processes, has driven the adoption of A.I. and other digital tools,” he said. 

Brennan said software solutions geared to customer management, accounting and human resources are making it easier for companies to transition to digital processes. A.I. could help with proactive safety measures by analyzing past data and providing recommendations.

“Incidents often occur due to workers not having the right information or training,” Brennan said. “A.I. can help address these issues by providing workers with instant access to information and guidance.”

Brennan used this example: “Say I’m a journeyman electrician. I go into the field to change out circuit breakers on a panel that is out of date. To prepare, I pick up my A.I.-enabled device and I say, ‘Today I’m going to be at this location, and changing out these breakers (which you identify).’ 

“The A.I. hears you and replies, ‘It looks like you’re probably going to need these tools and you’re going to want to bring these breakers.’ My device becomes my assistant. It might add, ‘Do you want me to notify and get you a lockout permit?’ Now I have an A.I.-enabled tool to help me with proper procedure, compliance and overall safety,” he said.

Brennan and Lengieza feel A.I. will be an integral part of a contractor’s business by simplifying tasks, improving productivity and perhaps revolutionizing safety procedures. It bears repeating that the success lies in the data fed to the A.I., which needs to be accurate, specific and concise. A.I. then augments human decision-making and planning, rather than replacing it. 

“I feel like the question happening in executives’ offices right now is they’re all asking their teams, ‘This A.I. thing, what are we doing about it?,’” Lengieza said. “Those in the construction industry aren’t likely to be software developers. I’m hoping that most contractors out there are not going out and trying to build their own A.I. solutions. Rather, start by asking the vendors you already work with if they are using A.I. to improve their product, to help improve workflows, to make people safer. If not, if A.I. is something you want to explore and learn more, search out vendors that are involved with this technology.”

Shutterstock / ProStockStudio / Procore