The efficiency case for artificial intelligence (AI) in insurance often arrives as a cost argument. James Vlasveld, senior underwriter for professional and financial risks at specialty insurer Markel in Australia, makes a different one.
"Leveraging the technology should allow us to be out in the market more as opposed to focusing on how we're going to take the operational costs out of the business," Vlasveld said.
Vlasveld described work to Insurance Business that is now done by AI and the time it is saving him. Asked whether he is finding more time for face-to-face contact, he pointed to financial institution submissions, the kind that arrive with a dozen underlying funds and require a consolidated view before any risk judgment can be formed.
"I'm finding myself extremely more efficient," he said. "That's a button press away, as opposed to having to sit down for several hours, if not days, in the old world, pre-AI, to go through those with a fine-tooth comb to find those specific numbers."
Vlasveld said the models Markel uses return the reference within the source document alongside the figure, so the underwriter can see where the number came from. Compounded across a book of business, he said, that saves hours of work. His account of what those hours now buy is specific. He said the change allows him to be more collaborative with broking partners and clients, and to focus on the risk itself rather than on detail that anybody could assemble given enough time.
The distinction Vlasveld drew is about what happens to the time, not whether the time is saved.
"I don't think of it as a cost-cutting exercise," he said. He allowed that cost is a portion of it and that the wider business community talks about the technology in those terms. The internal conversation at Markel, on his account, runs elsewhere. He said the question the company keeps returning to is how the technology can produce better underwriting outcomes.
"It's how can the underwriters be more present in their job and make better underwriting decisions," he said.
He extended that to the profit line, arguing that better decisions should produce better underwriting outcomes and, in turn, better underwriting profit outcomes.
Read next: From missiles to missed payments: How the Iran crisis is rippling through specialty lines
"I think that's where we should be focusing on, where that cost is being alleviated, as opposed to whether that's headcount or operational or whatnot."
Vlasveld also described significant investment attached to deployment rather than savings extracted from it. He said the internal communication has been that deploying the technology carries substantial investment and that he personally has access to four or five large language models because the organisation has been keen to give underwriters access to as much as possible.
He was not uncritical about what heavy reliance on the models can do to an underwriter. He described a risk of cognitive offloading, where the answer is an easy click and the thinking behind it stops being built and said Markel runs internal training that emphasises judgment, critical thinking and knowing how to challenge the outputs.
Whether the argument holds across the market is a separate question. Vlasveld suggested that specialty is well insulated from some of the pressures colleagues in the commercial insurance space are facing at the moment.
He acknowledged that parts of the specialty market will become more platform reliant, nominating small and medium enterprise segments and certain product subsets. Long-tail specialty placements, where the information volume is high and the risk is complex, he put in a different category. The reason he gave was accountability rather than capability.
"But I think ultimately brokers, boards, executives, they have to trust in that human element," he said.
He framed accountability as the key message for a complex judgment call and as a chain running in each direction. Underwriters, he said, need to be accountable in terms of their decision-making.
"Brokers are accountable in terms of having that human interaction and being confident in what service they're providing to their clients," Vlasveld said.
Clients carry accountability in the same fashion, on his account, and for large clients he tied it to their own governance obligations rather than to service preference.
"They can't be purchasing insurance that is just chucked down a pipeline somewhere that hasn't had a human," he said.
Watch next: Where AI meets PI: Coverage questions the industry is still answering
IB asked Vlasveld if he thought that, in the near future, specialty lines, despite their complexity, could still be at risk from an AI takeover?
"I don't want to be that story of the guy who sat there in the 19th century looking at the steam engine and the hot air balloon and said everything that's ever going to be invented has been invented," he said.
But for now at least, the complexity and accountability requirements of specialty lines may have safely ring-fenced the human components that run from insurer to broker to client. And the AI benefits Vlasveld is seeing are testable. If the time recovered from submission processing is genuinely being redirected outward, it should show up in response times, in how the risk conversation runs at renewal and in how often the underwriter turns up in person.
