“Firms don’t need to radically overhaul their systems to realize the benefits of AI,” said Meghan Anzelc (pictured), global leader of transformation solutions at Aon’s Strategy and Technology Group. “They often already have the tools, but just need some guidance and to be made aware of the use cases.”
That mental hurdle - more than cost or infrastructure - is what’s holding some re/insurers back. While the industry is beginning to integrate AI into core functions like underwriting, claims and risk modeling, adoption is uneven.
Anzelc pointed to three key areas where re/insurance firms were making real progress. The first was layering in additional data sources to refine underwriting and claims processes. The second was unlocking unstructured data that had been previously ignored. “Every organization has documents that are filed and forgotten - submissions, bordereaux, client files,” she said. “The cost of extracting value from them is much lower now than it was even a few years ago.”
The third area of traction was use of AI features already embedded in existing tools - CRMs, ERPs, and productivity software. “Many companies are already paying for these AI capabilities through their existing contracts. It’s just a matter of activating them,” she said.
Despite that, many firms were still stalling. Anzelc said it was often due to outdated assumptions about what deploying AI involves. “There’s a misconception that it requires a large investment or a massive team to get started,” she said. “That’s just not true anymore, but not everyone realizes the technology has evolved and that it is also user-friendly.”
That means that the benefits of AI are now available to all. “I have seen an organization implement a single AI tool into underwriting and achieve a $7 million improvement in profit in under a year,” Anzelc said. “For decades, forward-thinking companies have quietly been using AI for measurable and sustained value – improving loss ratios, reducing expenses, and achieving profitable growth. Those same benefits are now within reach of everyone in the industry.”
Anzelc was blunt about the limitations of even the most advanced AI tools when data is poor. “Regardless of how good the tech is, and to adopt a well-used phrase: it’s still garbage in, garbage out,” she said. “If your bordereau has no information, it just has no information.”
Where she saw real promise was in using AI to improve data assessment processes - flagging inconsistencies, identifying gaps and helping teams prioritize remediation. “You can build a robust and repeatable process to surface the quality of data at different points, and let people focus on fixing what matters,” she said.
The notion that legacy systems automatically stall AI transformation is another myth Anzelc dismissed. She’s worked with firms operating outdated and fragmented infrastructure, yet still found ways to deliver value. “Every organization I’ve worked with, no matter the standard of the systems there was always something you could do to make things better,” she said.
This might mean bolting on tools to extract insights from unstructured data and serve that intelligence to users without disrupting the underlying systems. It requires what Anzelc called a “small increments” mindset, rather than the pursuit of sweeping transformation. “If you’re going to make improve outcomes, something has to change. And with that comes at least some level of disruption,” she said.
The path forward, she added, is building a continuous improvement culture. That means looking for tactical wins, not just long-range moonshots.
With AI increasingly used in underwriting and pricing, regulatory scrutiny is sharpening. But Anzelc pushed back on the assumption that AI must be a black box. “That’s a design choice,” she said. “Transparency and precision are on a spectrum rather than an absolute.”
She pointed to the insurance sector’s long-standing use of generalized linear models in pricing - transparent, explainable and regulator-friendly. Today, more complex models can be layered with those same transparent techniques to make outputs understandable. “It’s not a black box if you’ve built it with explainability in mind,” she said.
With generative AI tools widely available, the playing field is more level than ever, Anzelc argued. “These capabilities are available to everyone. We can all go to openai.com or gemini.google.com and start using the tools,” she said. “It’s not about access anymore - it’s about what you’re doing with the readily-available tools.”
The firms that will pull ahead are those able to align technology with culture, operations and talent. “The greatest competitive advantage will come from those that bring these three elements together and execute at scale,” she said.
This has sparked deeper conversations within firms, beyond basic upskilling. Organizations are now looking at what structural changes are needed to support AI over the long term. “They’re not just asking how to upskill people today, but where they want to be as an organization in a few years’ time,” she said.
That includes talent planning, workforce strategy, and process design - not just AI investment. “It’s not enough to have a roadmap for your AI tools,” she said. “You need one for your people too.”
