If artificial intelligence (AI) were a person working alongside us in the world of construction-defect and general liability claims, it would not show up as the seasoned adjuster who has handled thousands of files, the forensic expert with decades of field experience, or the defense attorney who has memorized every procedural maneuver. Instead, AI today would look unmistakably like a 17-year-old. It would be sharp, inconsistent, bold, quick to learn, easily tripped up, occasionally brilliant, often misunderstood, and developing faster than the adults around it can fully appreciate.
In the claims and litigation sector in 2026, we see a system capable of analyzing data at superhuman scale but still vulnerable to misinterpretations. We see flashes of insight emerging from models that learn from enormous troves of text, images, and technical information; yet we also see the occasional error that makes headlines and fuels public anxiety.
Feeling anxiety is understandable: AI’s failures tend to be loud and very visible. A chatbot produces an odd answer, a hallucinated court proceeding makes its way into a legal filing, or an automated summary misses an important nuance, and the moment spreads quickly across social or professional networks. It becomes easy for observers to perceive these incidents as signs that the technology is unreliable or overrated. But this interpretation misses the larger and far more important truth: AI is still developing and, like any adolescent, its mistakes are part of its rapid growth rather than signs of decline. Despite occasional setbacks, the technology is rapidly progressing in areas such as reasoning capabilities, contextual comprehension, multimodal learning, and workflow integration.
AP Construction-Defect Claims
Nowhere is this developmental phase more striking than in construction claims and litigation. If we imagine AI as a 17-year-old, then the construction claims environment is the toughest academic program imaginable, akin to an Advanced Placement (AP) class. This is not a classroom of simple, labeled examples with clean inputs and tidy outputs. The environment is extraordinarily complex and challenging, characterized by uncertain information, technical and legal intricacies, emotionally invested parties, conflicting expert opinions, and documentation that is always changing. Introducing AI into this setting is like sending a teenager into a combined program of civil engineering, law, contract interpretation, forensic science, and insurance coverage analysis, all at the same time, with new material arriving daily and with no option to fall behind.
AI entering this environment is naturally prone to missteps, especially given the wide variety of applications. Depending on the type of program and software used, AI may misinterpret a crack pattern in a photograph or overgeneralize from an ambiguous sentence in a deposition transcript. AI may overlook subtle sequencing issues in a repair scope or misread the obligations of a subcontractor in a wrap up program. AI also has a habit of wanting to give you solutions, even when you have asked it not to.
But these mistakes are not technological failures; they are learning moments. They are signs that the system is beginning to apply adult-level reasoning to real-world problems and discovering where its early assumptions do not yet align with the complexity of the territory.
AI gets an A+
Just as with human teenagers, the moments of brilliance can be impressive, making both software developers and parents very proud. AI can read thousands of pages within minutes, cross reference information across documents instantly, and identify patterns in large data sets that humans would not have the time or cognitive bandwidth to notice. For instance, AI can spot inconsistencies between expert reports, highlight repeated defect patterns across buildings, or link sequences of events in a timeline that might otherwise remain buried in the noise. It can also boil down and simplify lengthy reports into bite-sized information, key for claim handlers to identify the specific information they need to make decisions. These are glimpses of an emerging adult intelligence.
The common claim that most AI projects fail is often misinterpreted. A widely cited study performed by MIT noted only a 5% success rate with AI applications, meaning 95% will fail. When organizations adopt AI without preparing data workflows, without integrating the technology into daily operations, or without setting realistic goals, the outcomes are predictably disappointing. And such disappointments are often noted in the lessons learned dialogues shared by digital transformation and innovation experts.
Expecting an AI model to deliver immediate practical and/or financial return without a supporting structure, or plan, is the equivalent of handing an unlicensed teenager the keys to a car without teaching them how to drive, then blaming the teenager for crashing the car. The root cause of failure is almost always a lack of planning, training, supervision, and systems integration. These cases fill the 95% bucket. And the cases where AI is rolled out thoughtfully, with attention to the workflows it must support, the results are often impressive. The 5% of successful integrations are meaningful and game changing.
AI in 2026
In 2026, AI begins transitioning out of its experimental phase and into a more functional and embedded role. This shift is driven by several major developments across the broader AI landscape. Systems are more efficient, more context aware, and more capable of interpreting multiple types of information at the same time. They are better at retaining context across long files, more confident in connecting evidence from different sources, and more comfortable working inside complex workflows rather than simply reacting to isolated prompts.
AI and Carriers
At the carrier level, AI’s maturing intelligence becomes even more powerful, especially given the amount of data and claims history in archives. AI can identify recurring defect patterns across a service line or book of business, highlight common subcontractors associated with certain issues, spot unexpected inflation in repair costs in particular regions, or detect repeated leakage in wrap up programs. It also identifies larger patterns based on claims data, such as a class of business that was thought to be profitable but actually is not (or vice versa), which allows underwriters to determine if that class is something to continue writing. These insights allow carriers to make proactive decisions rather than reacting case by case.
AI and Claims
In the claims industry, AI’s evolution manifests in several practical ways. Document review, for example, becomes more interpretive. Rather than generating broad summaries, AI connects a building envelope report with thermal imaging and photographic evidence, making inferences about moisture pathways or defect patterns. The documents are independent, but AI programs recognize the relationship between the two. AI can read a 12-building HOA defect file and begin to map the storyline in a way that feels almost like a junior associate who has read every page and retained the key insights. And just like with a junior associate, if the content is checked for accuracy and citations, it can be relied upon as being solid work product.
Claims analysis also becomes more relational. AI starts to recognize how documents relate to each other. It identifies contradictions between two experts who addressed the same detail differently several years apart. It detects that water intrusion problems across multiple buildings share a common directional pattern or material failure. It also differentiates between installation errors and manufacturing defects by recognizing repeating signatures across multiple files. Such analyses go beyond the expectations of a junior associate and bring tangible value and insights to ongoing cases.
AI and Expert Reporting
Expert reporting changes as well. AI cannot and will not replace an expert, nor should it. But it will increasingly support the analytical work that precedes expert opinions, specifically the extensive time in preparing analyses, calculations, and reporting. Using prompts with purpose, AI highlights missing evidence, aligns or contrasts various expert statements across drafts, identifies repair scopes that conflict with code requirements, and/or calls attention to sequencing assumptions that lack documentation. This level of analysis allows experts to focus their time on driving home the points that matter rather than on the time consuming process of retrieving and comparing background data.
AI and Litigation
Litigation support enters a new phase too. AI has become something like a mapmaker. AI tracks how testimony evolves across drafts and depositions. It identifies contradictions between fact witnesses and experts. It shows how small shifts in a timeline may affect notice requirements or limitations analysis. It weaves together causation threads that cross multiple building systems and identifies subrogation opportunities buried within layers of contractual relationships. As AI gains multimodal reasoning, it correlates photographs with witness statements, invoices with weather records, and scopes with code citations.
Maturation of AI
AI’s adolescence will not last much longer. The signs of adulthood are already visible. It creates enormous volumes of information, retains everything it has seen, and improves rapidly with exposure. Its limitations are shrinking, and its strengths are expanding. But just like any developing professional, AI still needs guidance, supervision, mentorship, and thoughtful integration.
Construction claims is not the place where AI will replace people. It is the place where AI is learning to become a professional itself, providing its colleagues with a 24/7 resource of information. The experts, adjusters, and attorneys who learn how to guide and shape this emerging intelligence today will be the leaders defining how the industry evolves over the next decade. They will be the ones who understand how to harness AI’s strengths, support its weaknesses, and ensure its growth serves the needs of clients, projects, and the built environment.
The teenager, AI, is growing up. It is time for the claims industry to decide what kind of adult we want it to become.
Terence Kadlec, P.E., is senior vice president, technical services and civil/structural engineer, MC Consultants. terence.kadlec@mcconsultants.com
Steve Lokus is vice president, casualty construction claims, Sompo. slokus@sompo-intl.com