Generative AI tools such as ChatGPT, which once seemed far off on the horizon, have arrived and are beginning to take shape. Industries, including construction and insurance, are contemplating whether, how, and to what extent they will incorporate this technology.
For those involved in construction claims and litigation, these tools offer potential solutions to challenges such as document management and analyzing loss events and trends. However, professionals need to weigh these benefits against potential risks. In a survey of CLM’s construction professionals, members and fellows sound off on how generative AI tools may impact construction claims, what some of the risks and benefits could be, and whether the industry will be an early adopter of these tools or take a wait-and-see approach.
How Generative AI Can Benefit the Industry
Some professionals ranged from optimistic to bullish about how generative AI tools might be used to help better evaluate and resolve construction claims.
Nick Bilski, director of risk management, BE&K Building Group—“I think it could be a life saver for insureds on claims. You simply put in information on a claims scenario, and it should spit out what coverage lines should respond, and also should spit out information on contractual indemnity as well. It should take the place of a claims role on behalf of a broker in the next 10 years.”
Michael Rodriguez, casualty claims specialist, Tokio Marine HCC—“Handling construction claims, we are dependent on the analysis of defects, damages, and the repairs required to correct these. Collating information and observations is of high importance to understand and to offer a solution. AI can help identify and track trends sooner. Examples would be asphalt bleed, underperforming waterproofing products, and CPVC failures. Understanding how materials and components fail, and under what conditions, can lead to better understanding of improper uses of such materials or installation standards that may ensure failure.”
William A. Wheatley, CEO, Wheatley US Limited—“A task that is often done poorly on construction projects that experience claims is document management. Sorting, categorizing, and linking related documents in a database can be done using software; a well-educated AI could do that, and at the same time scan for hints of problems that are about to surface. That would give the project manager the opportunity to intervene and head off the problem. I offer that service as a professional, but it could be automated using an AI.”
Caution Ahead
Other CLM professionals expressed concerns about using generative AI tools, with one saying outright, “I oppose ChatGPT and think it should be illegal.” While conversations have arisen about whether and how to regulate AI, a ban seems unlikely. But perhaps some caution is in order for a technology that, as one member puts it, appears to be a “good pretender.”
Joseph Herbert, partner, Hall Booth Smith, PC—“I believe that it may help parties work more efficiently and may help expedite construction processes. However, as with all technology there is the possibility that it can be exploited for improper purposes.”
Ryan Williams, partner/attorney, Wilson Elser Moskowitz Edelman & Dicker, LLP—“One concern I have is that construction professionals will rely on generative tools to create complex contracts without the assistance of counsel. Nuanced provisions, such as risk allocation and insurance requirements, may not get the scrutiny from knowledgeable professionals that is required in a given situation. If litigation arises, construction professionals may find that an AI-generated ‘one size fits all’ approach to drafting fails to offer the necessary protection, leading to unnecessary exposure.”
Patrick Kenny, partner, Armstrong Teasdale LLP—“The role of AI tools like ChatGPT is still a long way off from playing any meaningful role with respect to construction claims. I had been considering attempting to use ChatGPT as a first line of initial research, especially on items for which there seems to be no on-point law. I therefore tried out ChatGPT several times on what I consider 'softball' legal issues.
“ChatGPT is a good 'pretender.' It does generate short analyses and arguments on law related topics. However, its accuracy is way off. You can demonstrate this to yourself by picking a few legal issues that might come up in a claims setting and asking ChatGPT to generate a summary of the primary arguments with citations for or against some aspect of that legal issue. It will do it, and the argument will appear to be at least adequate, until you double check it.
“ChatGPT seems to lack judgment, and considers all of the data that it pulls from who knows where as equal in strength and reliability. The end result is ChatGPT regularly miscites legal authorities and often seems to make things up.
“If its errors were rare and minor, I might have continued to work with it, treating its ‘work product’ like that of a new associate right out of law school, but ChatGPT's reliability and work product does not come close to that of a brand new law school graduate.
“This is a clear case of the hype vastly exceeding the product. Generative AI is nowhere near ready for prime time. I am confident the early adopters of this tech in the insurance arena will end up with their names in the captions of bad faith decisions addressing the role of the technology.”
Early Adoption or Wait and See?
Most survey respondents believe the industry will take a wait-and-see approach to adopting generative AI tools such as ChatGPT. Some, like Williams, cite business reasons: “While current iterations of generative AI may offer tangential benefits—such as the ability to author basic analysis concerning the technical aspect of a claim’s subject matter—more meaningful claim-specific analysis is currently better handled by experienced counsel.”
Others note the traditional inertia in claims departments for holding off on adoption. “Claims departments tend to think in archaic terms and think traditionally,” Rodriguez says.
Still, others anticipate rapid adoption and tangible benefits. Wheatley says, “Claims departments should be early adopters, if they understand the benefits. Someone needs to develop a well-trained, special purpose version of ChatGPT that could be installed on the project computers. Most claims departments are not capable of implementing that. It needs to be made available as an off-the-shelf package for installation on project computer systems.”
The future of generative AI tools shows promise for construction claims professionals. But, as with any emerging technology, risks abound. If the range of survey responses here is any indication, there may be multiple ways forward for the industry when it comes to adoption and uses.