Artificial intelligence (AI) has come to all industries, including construction. And while almost every new technological jump tends to be lauded as “game-changing,” the impact of machine learning and qualitative outputs presented by AI appear to have that potential. But like any tool, the value of AI will be in how companies use and control it.
Accelerating and improving
We are not referring to armies of robot workers putting buildings together when talking about AI. Rather, AI is a broad description of many different computer functions that try to mimic human problem-solving.
Machine learning is the most relevant sector of AI currently being used for construction. In this process, a computer takes vast amounts of data and looks for interrelationships and consequences from how those data points interact. Humans do this regularly. Whether it is a toddler learning that a stove is hot only after touching it the first time or forgetting to start a friendly reminder with “per my last email.” The difference is that computers can process a volume of data points and interactions in seconds, while that same amount would take a human their entire lifetime.
Many core construction functions have a role based on reviewing plan data and then figuring out the best way to bring those plans together. This is true for design, bidding, procurement, safety and scheduling. AI streamlines these functions because, unlike BIM, which tries to model plan interactions, machine learning can potentially use BIM data and the physical building of millions of other prior jobs to model the construction process. AI can also repeat those modeling efforts hundreds or thousands of times, looking for problems and developing preferred schedules well before a permit is pulled or excavation begins.
Facilitating the trades
The construction trades are starting to see AI benefits as well. With market barriers to this technology being lowered to the point where AI is a smartphone app, AI is becoming more accessible. Automation of takeoffs, code compliance reviews, energy efficiency reviews based on a prescriptive path and performance modeling save thousands of hours. AI is beginning to mirror the value that seasoned experts have on jobs: they see the issues before they happen because they’ve encountered them before.
Learning its limits
While valuable, AI is not perfect. Like most technology, the issues and problems with AI do not lie with the technology but with the people using it. For example, a lawyer was recently sanctioned for filing a court document written by AI containing completely fabricated references and arguments. The core failing was not that the machine prepared something incorrectly, but that people failed to ask the right questions to create the proper output, and its results were not validated. Those concerns also rest with construction projects.
Where project-critical decisions are left to artificial means, problems arise without proper oversight. For that reason, it is not unreasonable to foresee scope of work agreements that contain express limitations on the use of AI in certain functions or compel specific datasets be used if AI is a tool being relied on during construction.
Insurance limitations are not far behind this technology as well. While errors and omissions and professional services policies may protect against mistakes made by well-intentioned but erroneous humans, the scope of coverage becomes more complicated when the error results from bad programming and data input. A developer’s use of AI for an RFI review may present coverage problems because the computer conducting the review may not have up-to-date data about the actual job, compromising its ability to respond. The insurance coverage question then becomes whether the cause of the damage was the AI response or failure to maintain the data—with the availability of insurance policy proceeds hanging in the balance.
Also, consider the real potential for injury. Poor sequencing from AI could jeopardize limited safety resources onsite. Improper data could lead to code compliance violations, and the potential failure to properly assess required performance metrics could lead to latent safety hazards.
Using powerful tech safely
Like any tool, AI has benefits and dangers of misuse. So, what are companies to do?
Start with planning ahead. Consider whether AI is a tool to be introduced or something to steer clear of on projects. If it will be used, what data validation and ongoing update protocols are implemented to ensure the results of AI responses are timely, valid and accurate. Stay current with vendors and their dataset offerings that can be used on jobs. And, perhaps most basically, get to the site. Human intuition and creativity are unlikely to be replaced soon.
Roles may change with AI’s use, but that’s true with all technological innovations. Getting ahead of the AI curve will allow early adaptive use of this emerging technology.