Meant as a sly swipe at the inflated hype around artificial intelligence, a billboard at a construction site in Antwerp, Belgium, in June read “Hey ChatGPT, finish this building.”
Artificial intelligence, the technology that powers chatbots like ChatGPT, won’t be assembling apartments or erecting stadiums any time soon, but in construction — an industry stereotypically known for clipboards and Excel spreadsheets — the rapid embrace of the technology may change how quickly projects are finished.
Drones, cameras, mobile apps and even some robots are increasingly mapping real-time progress on sprawling job sites, giving builders and contractors the ability to track and improve a project’s performance.
“Forget about robots building a skyscraper,” said James Swanston, chief executive of Voyage Control, which makes project management software for construction sites. “It’s a more fundamental thing, getting the data you need and then using it better.”
The construction industry has long been considered a digital laggard, but architects regularly use digital tools to design projects and create blueprints. Seeing tablets and drones on the same work sites as hard hats and safety vests is common.
Now helmet-mounted cameras capture footage of a site to orchestrate when new crews or materials should arrive, and precise sensors can detect whether a new window is a few millimeters off the project blueprint and needs to be adjusted. And A.I. is starting to be used in buying and selling real estate: JLL, a global broker, recently introduced its own chatbot to provide insights to its clients.
This expanded analysis of data is laying the groundwork for what many hope will be substantial improvements in accuracy, speed and efficiency by reducing the bloated timelines and waste that have made construction increasingly costly.
“The construction industry is the largest in the world, in terms of dollars spent, yet we are the least productive in terms of technological adoption and productivity gains,” said David Jason Gerber, a University of Southern California professor whose research focuses on advanced technology in construction.
But the industry’s embrace of A.I. technology faces challenges, including concerns over accuracy and hallucinations, in which a system provides an answer that is incorrect or nonsensical.
And further data collection has been a knotty problem, in large part because of the nature of huge construction projects: No two developments are the same, with wildly varying topography and local regulations, and new teams of contractors and subcontractors coming together for each project. It’s akin to starting a multimillion-dollar business for every sizable project.
Coordinating the complex ballet of supplies, labor and timetables remains a daunting task. But start-ups and investors see an opportunity, especially as machine learning models, which ingest enormous amounts of data to discern patterns and predict how similar situations will progress, are used to improve project performance.
The pandemic had already pushed construction firms to adopt more digital tools to allow them to work on site during lockdowns, accelerating the development of new technology, said Sarah Liu, a partner at Fifth Wall, a venture capital firm focused on real estate investments.
“The best companies aren’t touting themselves as A.I. companies,” she said. “They’re touting themselves as problem-solving companies.”
The construction consulting firm nPlan, led by Dev Amratia, who helped draft Britain’s national artificial intelligence strategy, uses complex algorithms to map out the progress of vast infrastructure projects and avoid mistakes or supply gaps. Its machine learning system was trained on a database of more than 740,000 projects.
The firm’s largest project to date, a $11 billion overhaul of railroad infrastructure in Northern England, will use the lessons gleaned from studying that vast array of projects to create detailed, real-time project maps for builders, which is expected to shave up to 5 percent off the total cost.
Buildots, a start-up in Israel that provides project management guidance via wearable cameras that analyze building progress, signed a deal for its first U.S. project, a mixed-use development in Manhattan. The firm commissioned a study of 64 international building sites, and it found that just 46 percent of the average work site was being used at any time, evidence of poor organization and scheduling.
“At the best construction site we’ve studied, progress varied by 30 percent each week,” said Aviv Leibovici, the firm’s chief product officer and a co-founder. “I think there are massive inefficiencies in this industry.”
Construction firms have also made significant investment in their in-house technology. Avison Young’s Project Management Services division claims its proprietary software and management programs can, on average, cut development time 20 percent.
Suffolk, a large construction firm in Boston, invested $110 million to fund construction start-ups, and it has a team of 30 data analysts collecting and scrutinizing information from job sites. At a construction site for South Station Tower in Boston, a 51-story development by Hines, cranes have cameras that document and label steel being used on the building’s frame, creating a data set expected to be used on other projects in the future. Additional programs are being used to track progress and even predict accidents.
“We have zero unemployment in the industry; technology is just going to help existing workers do more,” said John Fish, chairman and chief executive of Suffolk. “A.I. is just going to replace the companies that don’t use A.I.”
There is trepidation about A.I., and its reported issues with accuracy, being used in an industry where safety is so important. Programs like ChatGPT have an unfortunate tendency to occasionally make up answers based on incorrect predictions, said Julian Moutte, chief technology officer at Bentley Systems, a construction software firm.
“In infrastructure, this is something we can’t afford,” he said. “We can’t have A.I. hallucinate the design of a bridge.”
But the purported ability to work faster and cheaper has proved attractive. Dusty Robotics, a tech firm in Mountain View, Calif., develops autonomous devices to trace building blueprints on construction sites, a job typically done by hand. While researching the industry, the company’s chief executive, Tessa Lau, observed workers measuring out plans with chalk and tape; some workers had even tried taping pens to Roombas.
Ms. Lau was worried about the reaction that laborers would have to robots and A.I. encroaching on their job site. But in an industry desperate to attract younger workers, offering potential apprentices the ability to use drones and robots can help with recruitment and retention.
Tony Hernandez, a union carpentry trainer in Northern California who teaches apprentices to use drones and Dusty robots, sees these technologies as “just another tool.” He prefers the robot to trace lines instead of having to bend down and trace himself, meaning less wear and tear on his knees.
“This is a great retention tool,” he said. “It’s brought in kids who grew up on Xbox and can figure out these tools in a five-hour class.”
Dusty has 120 units on sites across the United States, but that is just the beginning. Ms. Lau calls the units, which can collect gigabytes of data, “Trojan horses to train the A.I.s of the future.”
Reducing risk may ultimately be where this technology makes its mark. Depending on the location and nature of work, insurance can make up as much as 10 percent of the cost of a single project, which can easily be hundreds of millions of dollars. Now, with A.I. providing better ways to keep on task, there is less risk and cheaper insurance options.
Shepherd, an insurance start-up, uses construction data to provide contractors with cheaper premiums. Wint, an Israeli start-up that uses proprietary sensors and algorithms to eliminate water damage, which leads to roughly a third of damage claims on construction sites, has been used on roughly 2,500 projects. A study by Munich Re found Wint can cut the loss rate 90 percent.
“Insurance costs can be the difference between whether or not projects are able to be sustainably financed,” said Justin Levine, a co-founder and the chief executive of Shepherd.
