Construction Insurers See AI’s Value in Better Risk Decisions, Not Cost Cuts

For construction firms and insurers, AI’s clearest business case may be fewer costly errors and claims, rather than lower operating expenses. At the International Risk Management and Insurance Construction Risk Conference in National Harbor, Maryland, panelists argued that the technology should help people make better decisions and catch risks earlier.

“If I have a 10% reduction in expense, that doesn’t hold a candle to a one-point basis reduction in loss ratio,” said Jay Bell, chief operating officer at Zurich North America. “It’s not even close. But the majority of the focus is still on efficiency.”

Put AI to work earlier in the risk cycle

Bell sees potential in using AI during preconstruction, underwriting and active project monitoring to find signals that could precede a claim. He said he’s interested in vendors working with insurers and contractors to identify those signals. Preventing a fatality, he added, would make the effort worthwhile.

The stakes are rising as construction activity shifts toward large, complex projects, including data centers, advanced manufacturing facilities and power generation, said Michelle Sartain, president of Marsh Risk U.S. and Canada. With risks becoming more interconnected, she said, AI could support planning, track project progress and help teams recognize emerging problems sooner.

That focus also changes how companies measure productivity. Bell warned that treating efficiency as the main objective can turn AI into a headcount reduction exercise, since cutting staff is a common way to capture savings. He argued that the better use is to empower employees to improve project outcomes.

Keep human judgment in the workflow

Cheri Hanes, head of construction innovation and sustainability at Axa XL, described AI as a check on professional work. An engineer who reviews five subcontractor materials estimates each day could use AI to fact-check the analysis and help catch expensive mistakes. Hanes said that approach is preferable to increasing the engineer’s workload to 10 estimates while assigning them to verify the machine’s output.

“We’ve already established that the human can’t thoroughly look at 10 of them a day, right?” Hanes said. In that setup, she said, the worker is being “used up by the machine” and knows they aren’t producing their best work.

AI can also serve as another set of eyes during safety reviews. Hanes, who is based in Dallas, said a chatbot can identify hazards in a jobsite photo with reasonable success, and may perform better when it can draw on Occupational Safety and Health Administration data or a company safety manual.

Data quality will shape the payoff

How firms adopt the technology will depend partly on their size, said Hugh Seaton, Austin-based managing director of consultancy AI Edge. Smaller businesses may use chatbots and assistants; larger firms may connect AI to existing systems; and the most sophisticated may develop their own tools. Those larger efforts require dedicated internal technology teams, Seaton said, but can make company expertise available to employees who have permission to access it, supporting better decisions, estimates and problem analysis.

That potential depends on the quality and organization of the underlying information. Panelists said structured, reliable data helps AI combine knowledge from across a business and produce more useful results.

Bell said fabricated AI answers, or hallucinations, have become less common. Errors may instead stem from prompts that omit necessary context or from conflicting material in a company’s knowledge base. Zurich is reviewing and restructuring its SharePoint files to improve the reliability of AI answers drawn from them. For construction firms, the practical test is whether AI helps teams recognize risk and act on sound information before a problem turns into a claim.