As AI tools seep into every corner of insurance, senior leaders are becoming less worried about the technology itself – and more worried about what it might do to junior talent. Those concerns surfaced during a recent industry event in Toronto.
Asked whether she worries about early‑career staff becoming too reliant on AI, CFC Underwriting CEO Kate Della Mora (pictured centre right) didn’t hesitate. “Yes, 100%,” she said.
For Della Mora, the concern isn’t that AI will replace junior underwriters, but that it will blunt the critical thinking skills they’re supposed to be developing.
“I think that critical thinking is going to be incredibly important for us as leaders to continue to foster within our organizations, because, again, it’s garbage in, garbage out,” she said.
She said she has already seen large language models confidently produce wrong answers – and warned that younger users may not have the experience to spot the errors.
Her concern extends to how AI might handle reputational information about insurers themselves. She noted that if enough people were to publish false or skewed reviews about a competitor online, a naïve AI system might later surface those as “facts” – simply because that’s what the internet shows.
“I’ve asked ChatGPT questions, and it’s wrong,” she said. “It’s going to spit that out to you as fact, and it’s not. It’s been created.”
Frederic Ling (pictured centre), SVP and head of specialty at Liberty Mutual, said he shares the concern.
“I do have a passionate opinion on this. Short answer is, yes, I do,” he said when asked whether junior staff could become too dependent on AI.
Ling’s fear is not just that junior underwriters will lean on AI for first drafts, but that they will be able to present those drafts so persuasively that managers, brokers and clients may not realise how shallow the underlying analysis is.
“My fear is that a junior underwriter can read a Copilot output and can present it in a really convincing manner to the audience and convince you… that is, in fact, true,” he said.
For Ling, the safeguard has to be deliberate “stress‑testing” of AI‑generated opinions.
“To me, it’s really around context,” he said. “So, it’s really stress‑testing the opinion… to see if there’s enough layers behind the comment or the conviction to see if you understand the risk or analysis.”
Without that discipline, he warned, organisations risk falling into a “circular approach” where chatbots reinforce each other’s outputs, and everyone simply assumes they must be right.
Despite those concerns, both Della Mora and Ling stressed that AI can significantly strengthen underwriting – if it is used to augment human judgment rather than replace it.
Della Mora said the real value lies in using AI to compress the grunt work. In the past, underwriters might have spent hours or days combing through complex submissions or documents just to pull out basic facts.
“From an underwriting perspective, it’s not necessary to fully replace underwriters,” she said. Instead, AI can take “really complex submissions or complex information that would have taken us hours or days to go through” and distil it down to the “pertinent points” needed for a proper risk analysis. That, in turn, should free up underwriters to build better relationships, deepen their industry knowledge and spend more time on judgment‑heavy decisions.
Ling echoed that view with a story from his early career, when he underwrote a complex publicly traded banking risk.
“I used to lock myself in a room for three or four days, and [go through] financial statements and notes, effectively doing my own data scrape,” he said. When he finally emerged with a stack of information, his manager asked for his view of the risk.
“I said, ‘Oh, I actually don’t know. I spent three days just gathering information,’” he recalled.
For him, that experience captures what AI should change. The goal is not to have models make the call instead of underwriters, but to “spend a lot less time… arriving at that data set” and much more time forming an informed opinion.
“If you speak with most specialty underwriters, the pride of their role is being an expert in their particular field or subject area. I don’t think that will change,” he said. Data and transformation tools should “take us… 80% of the way”; the remaining “20% is that subjectivity around context [and] the commercial relationship.”
