The construction sector is the largest industry in the world, accounting for 14.2% of global GDP. But as other industries have steadily grown their productivity in the last few decades, construction productivity has been stuck in the mud.

Between 1970 and 2020, as aggregate productivity for the U.S. economy doubled, in fact, labor productivity in the U.S. construction sector declined an average of 1% a year. Some estimates put this at $30 billion to $40 billion in losses. Meanwhile, schedule and cost overruns are the norm. Just 8.5% of megaprojects ($1 billion or more) meet or exceed their time and budget expectations, according to one research study. On top of all of that, skilled labor is growing scarce as older workers leave the industry and fewer young people enter its ranks.

While other industries have embraced the digital world to improve efficiency and performance, construction has historically been slow to adopt technology, says Dr. David Jason Gerber, director of the MS in Advanced Design and Construction program at the University of Southern California (USC).

It’s not hard to understand why. After all, you can’t digitalize concrete. Construction is a low-margin, materials-based, physical labor-intensive field.

But interest in AI could be changing all that. The industry has accumulated decades’ worth of data that, with the help of AI-driven analysis, could be tapped to boost productivity, prevent schedule overruns, improve cost-effectiveness, bridge labor gaps, and reduce risks.

Specifically, applying generative AI (GenAI) could be a turning point for the industry—particularly as construction projects grow more complex and demand continues to increase. Many of the largest construction firms are exploring GenAI use cases and adopting GenAI capabilities in software they already use. And the global market for GenAI in construction is set to grow at a CAGR of 35% between now and 2032, ultimately totaling $3.3 billion.

"We're sitting on 40 years of construction data," Kelsey Gauger, Suffolk Construction’s national director of operational excellence, told Business Insider. The $6-billion company has already deployed AI successfully to improve safety and now plans to use generative AI for operational efficiencies and, more strategically, to fuel better design decisions.

Productivity and GenAI in construction, by the numbers

U.S. labor productivity across all sectors, 1950–2020: +230%

U.S. labor productivity in construction, 1970–2020: -1% per year

Global market growth for GenAI in construction: 35% compound annual growth rate, 2022–2032

GenAI market value by 2032: $3.3 billion

Sources: Chicago Booth Review, Precedence Research

However, while construction firms that identify, test, and implement GenAI applications may be in a position to solve problems that have plagued construction for decades, they’ll need to overcome the industry’s traditional reluctance to spend on technology. They’ll also need to ensure top leadership support, prepare their data, identify priority use cases, create responsible AI policies, and encourage employees to accept GenAI value.

As Warren Kudman, senior vice president and CIO of international construction giant Turner Construction, told the Wall Street Journal, “I think we’re all looking for some proof of concept that we can go after fairly quickly to prove to ourselves that there is real impact. There is a certain degree of urgency because we don’t want to be left behind.”

Paving the way for GenAI in construction

Generative AI isn’t the only AI tool to solve the construction industry’s challenges. However, its ability to use large language models (LLM) to create new text and images through a natural language interface makes it a good fit for construction, from planning to jobsite execution.

Jose Luis Bianco, who leads McKinsey's engineering and construction work in North America, estimates that GenAI and AI could open value of up to $18 billion for home builders alone, which is about 10% of industry revenues.

Some prime areas for GenAI application include the following:

Supply chain and procurement. Construction firms manage vast, complex supply chains and networks of trade partners. By analyzing data on supplier performance, quality control, and project requirements, GenAI can identify potential bottlenecks or inefficiencies so construction companies can address issues before they occur, streamline operations, and ensure timely project completion. When facing material shortages or delayed deliveries, for example, GenAI might suggest alternative suppliers based on past collaborations or supplier quality, to keep work flowing.

At its annual Innovation Summit in November 2023, Turner Construction demonstrated a GenAI contract drafting tool that uses natural language processing (NLP) and machine learning algorithms to understand requirements, automatically generate text for statements of work or master services agreements, and speed up the overall procurement process. Tested by the firm’s procurement team, this is a huge productivity tool for a company with 30,000 annual deals with contractors who supply labor and materials for projects.

According to McKinsey, GenAI and AI could open value of up to $18 billion for home builders alone, which is about 10% of industry revenues.

Digital assistants/copilots. Nearly one in four construction workers in the U.S. are 55 or older, according to the Associated Builders and Contractors (ABC) trade association. Meanwhile, an estimated 1.9 million construction workers are likely to leave the industry in 2024, according to historical U.S. Census Bureau information. The U.S. construction industry will need an estimated 501,000 additional workers on top of the normal pace of hiring in 2024 to meet demand (and another 454,000 in 2025), according to a proprietary model developed by Associated Builders and Contractors. The UK will need 937,000 new construction recruits over the next decade, according to the UK Trade Skills Index 2023.

GenAI could help get new hires up to speed much more quickly, through the deployment of digital assistants for junior project managers. By interacting with existing systems and data, these assistants can answer queries, offer suggestions, and provide step-by-step instructions in real time. Suffolk Construction, for instance, plans to roll out the copilot built into its collaboration software to find information and answer questions for workers in the field.

Construction’s labor problem

Number of construction workers in the U.S. who are 55 or older: 1 in 4

Number of construction workers in the U.S. likely to leave the industry in 2024: 1.9 million

Number of additional workers needed in 2024, on top of the normal pace of hiring: 501,000

Sources: Associated Builders and Contractors , U.S. Census Bureau

Copilots could also take on tedious tasks, boosting productivity among leaner staffs. Gilbane Building Company, one of the largest privately-held real estate development and construction companies in the U.S., has been conducting GenAI proofs of concept that position its people for the “best use of their time.” Last November, Gilbane began piloting an LLM that enables project teams to retrieve up-to-date construction documents “within seconds,” says Kelly Benedict, Gilbane’s Head of Innovation and Transformation in an interview with Building Design + Construction.

“GenAI can become that assistant or intermediary for any persona on a job site, from superintendent to trade foreman, to do their job better,” says Michael Colapietro, CEO and president of construction and operations automation platform SmartApp. When scheduling a sequence of work, for example, users can ask the assistant to “show me my critical path items at risk” or “help me figure out my best two-week look-ahead for tasks for all my trades.”

Project management and collaboration. Construction projects involve multiple parties engaged in a wide range of discrete activities. As a result, keeping everyone on the same page – literally – remains a challenge. It’s not uncommon, in fact, to use paper blueprints, drawings, and specs, and when plans change, not everyone may have access to the most recent information.

Many in the industry have shifted to using building information modeling (BIM) systems, whereby a digital twin of the built asset is created and managed. However, updating the digital twin can be laborious, as can pulling information from the BIM, which may contain hundreds of terabytes of data.

GenAI can help by updating the digital twin with data gathered from AI-powered cameras and sensors and comparing it with original plans. For example, it might find important information within the data lake of the digital twin, or even send alerts when, say, piping is scheduled for the following week, but the materials are running behind. Once the project is complete, construction firms could use GenAI to take all the data in the BIM and package it up as a digital manual, including links to warranties and troubleshooting information.

Dynamic scheduling. Scheduling has historically been the purview of industry pros using traditional project planning tools—and years of experience—to build out their schedules in a linear fashion. Typically, a single schedule is produced, as exploring alternatives would be too time-consuming and costly.

Project scheduling could be done more quickly and accurately using AI algorithms to analyze historical company data, project requirements, and resource availability. GenAI tools could quickly develop multiple scheduling options and run what-if analyses to explore how changes in variables (e.g., number of cranes used) would affect timelines and costs. Such automation and optimization could yield significant time and cost savings.

Building the schedule, though, is only half the battle. Adjusting it based on changing conditions is the other. LLMs can be used to assess project risks, considering factors like weather, labor availability, and supply chain disruptions. These models can also continuously monitor a project's progress and make changes to schedules in real time. If a task is delayed or completed early, the algorithm can automatically reschedule dependent tasks to improve project timelines.

LLMs can be used to assess project risks, considering factors like weather, labor availability, and supply chain disruptions. These models can also continuously monitor a project's progress and make changes to schedules in real time.

Quality control and compliance. Keeping on top of complex, confusing, and frequently changing building codes is another challenge GenAI is particularly well suited to solve. Regulations vary by jurisdiction, year, and project type. They often include exceptions and can be challenging to interpret.

Some construction technology vendors are introducing AI-powered building code copilots that can provide answers to building code queries: How many exits are required for a restaurant? What’s the minimum ceiling height for a habitable room? Gen AI could also be used to analyze building schemes and blueprints to identify elements that don't meet standards, enabling early corrections and adjustments.

Reshaping construction with AI

Discover how integrating AI and machine learning can improve planning, safety, efficiency, cost control, and informed decision-making in engineering, construction, and operations. See demo video.

Hammering out the details

Construction businesses face multiple hurdles when it comes to taking advantage of GenAI. Most of these are centered on the industry’s relatively low levels of technology investment to date. This is combined with these businesses’ propensity to favor short-term cost concerns over long-term tech investment benefits. “The contract values may be relatively large, but the margins are not,” says USC’s Gerber. “There can be some myopia and worry about how this will affect the meeting of deliverables on this project today.”

Nonetheless, with some of the biggest industry players moving forward with GenAI, it’s increasingly less of an option to sit on the sidelines. To prepare for GenAI, construction businesses will need to focus on the following:

Make GenAI a business priority. Construction companies need to coalesce around a vision and commitment to using GenAI, determine how much they are willing to invest, and authorize an accountable senior leader to take charge of the project, advises McKinsey’s Bianco. Large construction firms can build their own teams of data and AI experts to work with business leaders to define and pursue use cases, while smaller firms may lean more heavily on external partners to drive strategy and adoption.

The important thing is to get started now rather than wait to explore these technology innovations, which is what construction firms have historically done.

Get serious about data. Contractors tend to think each project is unlike any they’ve done before. And, indeed, for many construction businesses, each project actually is a one-off.

“You only build a building once,” Gerber says, “and the project teams are always different.” This way of thinking, however, results in siloed construction data, often owned by different partners, who are reticent to share their information because of competitive concerns.

So, while the industry sits on an enormous cache of data, many are just at the starting point of harmonizing the data and providing consistent access to it. Tasks like data cleansing, normalizing, and transforming will be essential to remove noise, standardize formats, and make the data suitable for training generative AI models.

As part of its investigation into GenAI, Clark Construction is creating an enterprise data warehouse and building a data fabric to connect various local, ad hoc, and enterprise systems, says Akhil Mathur, senior technology director at the large commercial and civil contractor in the U.S. The company plans to use GenAI to inform decisions and optimize the acquisition, planning, and delivery of construction projects.

Contractors tend to think each project is unlike any they’ve done before. And, indeed, for many construction businesses, each project actually is a one-off. This way of thinking, however, results in siloed construction data, often owned by different partners, who are reticent to share their information because of competitive concerns.

Prioritize use cases. Explore specific outcomes (or actual prompts) that GenAI could help with beyond cost cutting or productivity improvements. Examples include Turner’s choice of using GenAI to draft contracts for its tens of thousands of annual deals, and Gilbane’s LLM pilot that helps project teams retrieve information in seconds. Validate and prioritize those use cases, considering which are not only the highest value but also the most feasible.

Pilot with care. “ It’s all about de-risking the pilots,” says Gerber. Construction companies can do this by choosing projects where GenAI can be explored without adding risk—maintaining parallel processes, for example, and keeping humans in the loop. They can also partner with vendors and academic institutions, and have strategies to measure, monitor, and manage their test cases.

One of the benefits of construction is that—because it’s project-based—businesses can try out a new technology on a single project. Similar to testing out a new material, if it’s useful, they can increase their use of it.

Create responsible AI policies and governance. This space is changing quickly. From the start, organizations need to make sure they’ve fully assessed the risks, such as data privacy, explainability, AI bias, and model hallucinations. They also need to create AI policies and risk mitigation practices before rolling out new capabilities. As Julien Moutte, chief technology officer at construction software firm Bentley Systems told The New York Times: “We can’t have AI hallucinate the design of a bridge.”

Responsible AI frameworks should include ethical guardrails, accountability for managing risks, an organization-wide structure for controls and governance, and ongoing reassessment, Bianco advises. Clark Construction, for example, wants to be able to take advantage of external data for use in AI models but also needs to prevent the sharing of its own data with the outside world, says Mathur.

Bring the culture along. Finally, of course, construction employees will have to be comfortable using GenAI. This is something early adopters are already encountering with other AI initiatives. Suffolk, for example, appointed its first chief data officer in 2017 and has a team of 30 data analysts, but there is still some resistance to digital solutions.

“Holistically, our industry does not always leverage technology and data to the fullest capacity, and historically [it] has not done that," Suffolk’s Gauger told Business Insider. It has been challenging, he continued, to convince people to trust the data and to "buy into the information that the model is providing."

No turning back

GenAI has great potential to usher in a new wave of productivity and efficiency within the construction industry and could well be a key driver in some long-overdue digitalization, data integration, and governance. As more industry players experiment with and adopt the technology, the pressure is on others to adopt, as well.

At Clark Construction, Mathur says company leaders are convinced that active exploration of GenAI is crucial to continued competitive advantage and ability to win work. “The real risk,” agrees Gerber, “is to not invest and be left behind.”