AI in Construction: Potential Use Cases and Challenges to Adoption

The year is 1968 and a glowing red light tells an astronaut named Dave that it can’t open the pod bay doors. It’s 1984 and Arnold Schwarzenegger is wearing dark shades and telling us he’ll be back. It’s just 11 years ago and Joaquin Phoenix is falling in love with a computer voiced by Scarlett Johansson.

Humans, for some reason, are obsessed with intelligence that isn’t our own. There are countless depictions of AI dating back decades, from the introduction of the Turing Test in 1950 to Isaac Asimov’s publication of his I, Robot stories shortly after. Enemies, overlords, lovers, helpers, friends - there’s no shortage of the variable ways AI is represented in media. But one portrayal is still missing from the catalog. Consider:

We open in media res. A construction project manager sits at his computer. He is in a field trailer parked close to an active job site. Men and women in high-vis vests move around the interior, carrying tablets and small, styrofoam cups of coffee. Despite the frantic pace of the job site, the man is calm. He opens his project management software and types a simple command - “generate report.” An AI assistant pulls data from multiple sources, incorporates factors like weather delays, and produces a comprehensive progress report. Our hero reviews the report and sends it off to the other project stakeholders. Credits roll.

Cue a 10-minute standing ovation at Cannes, a promotional press circuit, more Oscars and Golden Globes than any shelf can possibly hold.

Okay, not really. The more mundane uses of AI are not the ones that are frequently captured on the big screen. But as we consider the potential use cases of AI in infrastructure construction, you may still find yourself tempted to applaud the potential efficiency and accuracy gains. And of course, as with any new technology, there are obstacles to consider. Let’s dive in.

Potential Use Cases of AI in Construction

Predictive analytics can prevent cost overruns

AI works by processing massive amounts of data to draw reasonable conclusions. Thus, it stands to reason that if you gave artificial intelligence historical access to every project you’ve ever managed, it could use that information to generate projected costs and timelines based on the history of project and team size, past overruns, weather, and other factors. This information could be used to both predict potential risk areas for overruns and create a realistic project timeline and cost estimation.

Data analysis from multiple sources can automate and deepen reporting

Speaking of vast quantities of data, the digital job site revolution is widening the availability, amount, and quality of data on most infrastructure projects. Projects are bid, managed, and inspected digitally. Rovers capture precise measurements and tie them to item quantities automatically. GIS integrations are increasingly common. Even notes from inspectors can be digitally collated into distinct observations about personnel, equipment, potential concerns, etc.

With all of this data available, project reporting may become increasingly automated, where AI analyzes data from these combined sources to produce comprehensive reports on project progress. And since AI can draw conclusions based on this data, project managers may be able to ask questions like “what is the time elapsed on items versus proposed?” and instantly get an answer.

Photo analysis can improve job site safety

“May” is a bit of a misnomer here - this technology is already in use. It’s not widespread yet, but some companies are already combining drone photography and AI in compelling ways. AI can scan and analyze drone photography of a job site to point out geographical features that may impact safety. Similarly, that combined technology can be used during construction to pinpoint safety violations - a missing hard hat or high-vis jacket here, absent signage there. With over 20,000 annual injuries to workers in road construction alone, it’s no wonder this technology is already being leveraged to protect the construction workforce.

AI-powered efficiencies can address workforce shortages

A 2022 study by AGC America found that 91% of construction firms with open positions reported difficulties in finding professionals to fill those positions. It isn’t just a lack of people that is affecting the workforce - it’s a lack of industry knowledge. “Brain drain” is a common phrase used to describe our most knowledgeable and experienced workers leaving the industry due to retirement or other factors - the “silver tsunami” is another. Not only can AI make the workforce we do have more efficient by helping with project planning and labor distribution - it may also be able to fill those knowledge gaps being left behind.

What if you could program an AI to draw on the experiences of our retiring experts and act as a responsive knowledge base? What if one of those dog robots created by Boston Dynamics could be programmed to conduct inspections? Of course, there may be concerns about human jobs being replaced by machinery - more on that later.

Vast databases can support asset management post-project

We touched on the value of AI for project planning and ongoing project management. But what about after the project is finished? The use of AI can tie closely into BIM and asset management practices. If you’re capturing every aspect of a construction project digitally, you have access to a massive repository of data. Things like material types, weather during installation, issues that popped up during construction, etc. Just as with predictive analytics for project planning, AI can use this data to predict when assets will degrade over time and suggest preventative maintenance actions. By leveraging AI, not only can we build a better asset - we can maximize its use long-term by preventing problems before they develop.

Challenges to Adoption and Implementation

The sustainability concern

An oft-cited statistic - and I know this because my partner mentions it to me every time I use AI to generate a picture of my dog as the Mona Lisa - is that a single ChatGPT query uses something like seven cups of water to cool down servers as they process data. We may soon see increased regulations on AI usage as we understand more about the energy consumption involved. Some developers have dubbed this “the energy problem” - the demand for AI electricity requirements is projected to be greater than what is available. Of course, it should be noted that people far, far smarter than me are working on improving hardware efficiency and overcoming this challenge.

How do we get all of this data?

In previous issues of Deconstructor, we’ve talked about how a common data environment would require the implementation of open data standards. Open data standards support the exchange of data and the interoperability necessary to pull information from multiple sources into one database for AI to parse and process.

For an industry that’s still lagging on the digital adoption curve, the potential unavailability of this data presents an obstacle to AI-driven efficiency that relies on large quantities of quality data. Engineering firms and local governments need to continue embracing advanced digital construction management systems to capture data electronically. An equal burden falls on the developers of these systems to ensure they play nicely with others.

To quote Infotech’s Jim Ferguson:

“As organizations pursue artificial intelligence, the quality of the data in those systems is a crucial component. When data quality meets the standard for its intended use, it ensures that the data used for reporting, decision-making, and analysis is trustworthy and reliable. High data quality ensures that information is suitable for analysis, decision-making, risk management, reporting, and other data-driven activities.”

The robots are coming for my job

Replacing humans with technology is not a new concept. Textiles were spun by hand until the Spinning Jenny reduced the need for manual spinners. Henry Ford popularized the assembly line in factories to reduce the need for laborers in automobile production. The concerns about technology taking over human jobs are often overstated - unless you’d rather still be spinning your clothes by hand, of course. Each time there is a technological leap like this, it frees up humans to take on more important roles than require distinctly human judgment.

More importantly, as stated above, there are more jobs available in the infrastructure construction industry than there are humans to fill them. While there are genuine, real concerns about the implementation of AI technology and the various aspects of life it may impact, job security should not be one of them.

The robots are coming for us all

In 2015, over 150 scientists and entrepreneurs, including Stephen Hawking and Oren Etzioni, signed an open letter warning about the long-term concerns of artificial intelligence. It’s not a particularly fun read, with passages like “we could one day lose control of AI systems via the rise of superintelligences that do not act in accordance with human wishes – and that such powerful systems would threaten humanity.”

I’m not going to call this fear-mongering, as many genuine experts in the field of artificial intelligence contributed to this letter. That said, I will note that one of the signatories, Elon Musk, is currently doing everything he can to sell us on “Grok” an AI-powered search assistant that is under fire for inaccurate summaries and the spread of misinformation. Clearly, the expert community’s feelings on AI have evolved since 2015 as they are increasingly incorporated into our everyday tools. Will there ever be a Skynet-like uprising that pushes us to invent time travel to save humanity? Perhaps- but for now, I imagine most AI engines are more focused on putting the right amount of fingers on a human hand.

What’s Next?

Construction tech providers are exploring new ways to incorporate AI into their solutions. At Infotech, we’re working to see how AI can improve reporting and research in our Appia platform for construction administration and inspection. Some of those potential use cases are explored here in a blog about compensable delays and claim avoidance. Right now, we’re in the listening phase. What aspects of your job would you love to be automated by AI? What practical use cases are you dreaming up? Let us know- we promise not to accidentally create any T-800s.