The AI advantage isn’t the algorithm
AI might make insurers faster and more sophisticated at analysing risk, but the technology itself is unlikely to provide a lasting competitive advantage, according to Ascot Group’s CIO.
Instead, as AI tools become more widely available, the differentiators could increasingly be the things insurers cannot simply buy: access to attractive business, strong distribution relationships and the accumulated judgement of their underwriters.
That is the view of Owen Williams, group CIO at Ascot Group, who argues that the industry’s focus on increasingly sophisticated models can obscure a more fundamental question: what risks does an insurer actually have the opportunity to write?
“The leading tools are available to everyone at similar cost, and a capability you can buy by subscription is not a moat,” Williams told Bermuda+ILS ahead of ILS Bermuda Convergence, held October 12-14.
Even the idea that proprietary data will separate carriers deserves scrutiny, he said. Insurers might possess large quantities of private information, but that does not necessarily make it a valuable proprietary dataset.
“Unstructured, unlabelled and disconnected data can be a liability, not an asset,” Williams said. Instead, he believes advantage is more likely to come from governance, including a consistent data model, clear lineage and the ability to connect individual underwriting decisions with their eventual outcomes.
But Williams sees another factor as still more important in specialty insurance: submission flow. “Selection can only work on the risks you’re shown, and distribution and broker relationships determine that pool,” he said. “The best model in the market can’t beat an average one on a thin flow.”
That could become more significant as AI improves the speed at which insurers process submissions. Williams expects the broker-to-carrier interface to move increasingly away from document-based exchanges over the next five years, with submissions arriving as structured, machine-readable exposure data.
Much of the current back-and-forth required to obtain missing information could consequently disappear. But while enabled by technology, Williams sees that development fundamentally as a change in distribution. “The carrier that can quote credibly in hours gets first look, and first look is worth more than a slightly better model,” he said.
Technology, in other words, does not remove the importance of insurer-broker relationships. It could amplify their value. Carriers that brokers already trust because of service, claims handling and established relationships might be better positioned to benefit from faster underwriting than competitors using much the same technology. “As AI raises the floor of underwriting proficiency, distribution is what separates carriers above it,” Williams said.
Nor does greater automation necessarily diminish the importance of the underwriter. Williams argues it could make expert judgement more valuable, provided insurers capture it. “Every expert decision is a training label,” he said.
The distinction is between recording what an underwriter decided and capturing why that decision was made. An insurer that systematically preserves the reasoning behind risk selection and pricing decisions can connect that judgement with subsequent outcomes, gradually building a body of knowledge derived from its own underwriting experience.
“Carriers that capture the reasoning behind decisions, not just the outcomes, build an asset no one else can copy. Carriers that automate without recording that reasoning give it away,” Williams said.
That points to a different kind of AI competition from a race to acquire the most sophisticated model. If broadly available technology raises underwriting capabilities across the market, differentiation might depend increasingly on what sits around the model: the quality of business coming through the door, the relationships that generate that flow and the institutional knowledge being captured from those deciding which risks to take.
The technology might become easier to obtain. The harder task will be building – and retaining – the advantages it can put to work.
Did you get value from this story? Sign up to our free newsletters and get stories like this sent straight to your inbox.
