The year is 2033. The owner of a planned high-rise in a big city is looking to build prefabricated modules with recyclable, low-carbon materials. Work is expected to begin as soon as possible, so the contractor cannot afford to spend a great deal of time figuring out how much of each material it will need to complete the project. Throughout the course of the project, the owner expects the entire project team to be able to use artificial intelligence-powered construction software to sift through relevant information across thousands of pages of documents in mere seconds.
“With all those things combined, I think you’re about to see huge changes,” says Patrick Murphy, Chief Investment Officer at Coastal Construction. “And in 10 years, I think it’ll be an unrecognizable industry for the most part, compared to today.”
Coastal Construction can count itself among the earliest adopters of AI for big projects. The company recently won an award from the Associated General Contractors of America for its use of AI for estimation work on luxurious residences like the Plaza Coral Gables in Miami. But according to Murphy, the company’s foray into AI wasn’t just about being innovative; it was in response to a real need.
“I decided to look at our financial statements and see where the biggest sort of spend was, and the biggest piece of overhead in our company, and most construction companies, is the estimating department,” Murphy said. “And as we were analyzing that, we realized that that takeoff piece … that measuring piece, is what’s taking around 50-60% of all the time and money. So that’s when the light bulb went off.”
According to a Coastal Construction press release about its AGC honor, the company cut nearly 14,000 hours of work and saved almost $1 million in its first year using Togal.AI, a spinoff company Murphy founded that promises to reduce the takeoff process from weeks to mere minutes. Murphy notes the tool now comes with ChatGPT functionality that allows builders to use natural language queries to search for information across different types of planning documents, specifications, contracts, invoices, schedules and budgets. Togal.AI’s ChatGPT models understand the context and intent of the queries. Murphy says the technology will eventually be able to convert flat 2D images of construction plans, extract copious data from them based on millions of other plans, and convert them into text format.
Murphy notes that AI could allow humans that normally spend two or three days coloring and tracing plans to focus their efforts on more strategic tasks of a construction project that lead to a better outcome and ultimately less financial burden on general contractors and their subcontractors.
“There are fewer missed things and fewer change orders, and you can really understand how that job is going to go up,” Murphy says. “Right now, most humans don’t have the luxury of that. They’re so busy and they’re spending their time on silly tasks.”
But what about the not-so-silly tasks? Ro Bhatia, president and CEO of PlanHub, says AI has great potential to efficiently detect and address defects in cement production. On the purely operational side of construction, Bhatia believes AI can accelerate data analysis and help project teams make faster, better decisions with a greater degree of certainty. By analyzing data from past projects, AI systems can identify patterns and predict which components of a building are most likely to fail, allowing contractors to offer more comprehensive and accurate warranties to their clients. Experts like Bhatia and Murphy say AI systems can also use insight from past projects to identify potential delays and help contractors to schedule work more efficiently.
“In preconstruction, I expect AI and ML will rapidly become more relevant in both customizing and sending bids, for bid leveling … and risk management,” Bhatia says.
Another aspect of construction operations that could evolve with the AI boom is marketing. Chris Martin, founder and president of Atlas Marketing, says ChatGPT could be useful for general business research and developing keywords and hashtags to strengthen a construction company’s SEO strategy. However, he warns that ChatGPT has a tendency to produce inaccurate content and, even when heavily prompted, isn’t always able to convey a brand’s story authentically. And the information it is able to grab could cause legal problems for contractors if it is plagiarized.
On the physical side of risk, some see potential for AI to enhance construction safety due to the unfortunate amount of data on how accidents occur in the industry. Many of those accidents, Murphy notes, are repeatable and predictable, so being able to use AI to plan ahead and preempt danger on tasks such as digging ditches or placing railings on balconies can help contractors “work backwards.”
In a similar vein, Kris Lengieza, Vice President of Global Partnerships and Alliances at Procore Technologies, notes that AI programs can verify progress on certain tasks. For example, if a human construction worker says he or she is done with one-quarter of the drywall on a particular floor of a multistory building, an AI program can corroborate or refute that. It takes a human to reconcile those two data inputs, but every time someone does that, it trains the AI and makes it more sophisticated.
Lengieza believes contractors with large tranches of data will be best suited to influence the way AI evolves in the industry. ChatGPT is only as potent as the data in its machine-learning model, so what happens when that data isn’t readily available?
“[In] construction, and especially construction technology, a lot of the companies’ approach has been ‘I want to keep my data, it’s my data, and I’m going to hold on to it and keep it real tight,’” Lengieza says. “And it’s certainly not integrated across platforms. It’s certainly not publicly available. And so you can’t train those models if the data’s not there, right?”
But data is just one part of the equation. The other part is trust. And that will take some time to build. Contractors and owners alike won’t truly feel comfortable with AI until they understand how it works.
“Think of it like self-driving cars; how many people are really comfortable getting into a car in San Francisco that’s not an Uber where there’s no driver?” Lengieza asks.
That can change, though, with proper education. The ultimate goal, Lengieza says, is for the technology to reach a point where contractors don’t necessarily realize they’re even working with AI; they just know their jobs are easier.
“We’re gonna need a lot of case studies,” Lengieza says. “We’re gonna need a lot of case studies of where it went right and we’re gonna need case studies of where it went wrong so that people can avoid the common pitfalls.”
If history is any indication, contractors are well-equipped to not only handle, but harness, disruptive innovation. The ongoing AI boom may seem daunting for an industry that consistently struggles to find qualified workers, but not long ago, building information modeling also used to be shiny and new.
Lengieza notes that BIM is now a standard part of construction operations. He believes AI could follow a similar path, with the “largest of the large” contractors trying it out at first and then seeing the technology “trickle its way down” to subcontractors. But as that happens, Lengieza hopes the industry will understand the responsibility it has to use AI without putting anyone at a disadvantage.
“If we can democratize AI to everyone, and the things we use it for are to improve the overall project, improve the deliverable, improve the profitability of the product, then we can use it to improve the deliverable and improve the profitability of everyone,” he says. “Business needs drive innovation a lot of times, and there’s still a lot of business need around making sure that you are holding people accountable, that people are delivering what they’re supposed to.”