How Argo Group approaches AI integration

Register now

Andy Breen joined Argo Group, a commercial specialty insurer, nine months ago after several years with American Express. His charge as SVP of digital for the company has been to streamline Argo's digital initiatives and help transition the carrier from traditional insurance IT tactics into the new digital-first world.

"Our CEO [Mark Watson] believes that the industry has been insulated from tech disruption due to capital requirements and regulations, but it's coming, and he wants us to be out ahead of that and become the first 'digital carrier,'" Breen says.

Breen, like many industry observers, sees distribution as the ripest section of the insurance value chain for disruption. But for Argo, and its often esoteric product mix, that doesn't mean replacing the company's cohort of brokers with robots. Rather, the goal is to use digital technologies like artificial intelligence and machine learning to supplement the existing distribution staff and make it easier for them to accurately price risks. That means automating certain clerical work, or data-gathering from third-party sources, and giving underwriters the best possible picture of a risk

"We want to be the fastest and the smartest underwriters," Breen says. "We do a lot of algorithmic quoting, and we're using data and machine learning and AI to build models that can do more sophisticated pricing."

Getting from rules-based underwriting technology to truly dynamic machine learning is not an easy process, Breen continues. However, the technology has made great advancements in the past few years, fueled by the increasing number of startup companies putting time and effort into the technology.

Breen says that while he's impressed with the attention and effort flooding into the insurtech arena, comparing it favorably with the fintech boom in other financial services several years ago, that insurers should cast a wide net when looking for partners.

"All the AI startups we work with had nothing to do with insurance" before signing on with Argo, he says. The company looked for partners that had strong AI platforms and were having success in other business, and focused on using their data to help build an insurance use case for the platform. Argo is hiring data scientists and engineers that can help digitalize all the unstructured data coming into the company and feed the AI platforms that come in from outside.

"I don't have to bring in the fundamental science or know the stochastic calculus to build AI from scratch," Breen says. "What we also do is we organize a network across enterprise boundaries, partnerships and investments in the AI field -- we have other people working on things where we don't have the expertise."

AI holds a lot of promise for insurers, Breen concludes, but carriers also have to make adjustments in how they think about technology and relationships with tech providers in order to get the most value out of it. For example, because AI learns dynamically, it's harder to test than a rules-based system.

"How do you test a non-deterministic system? You really have to have people that dive in," he says. But, he notes, "The machine doesn't have biases other than what the human trained it on. You want to take the best peforming elements of underwriting and encode that into the machine."

And, he says, insurers have to be willing to trade the first-mover advantage for the possibility that their data will be used to develop a platform that their peers could use.

"Before, if I was going to buy tech, I would have IP protections. But in the world of AI, I'm going to give [a vendor] a bunch of data, but [they] get the benefit of having one big neural net," he explains.

For reprint and licensing requests for this article, click here.
Artificial intelligence Machine learning Underwriting Customer data Data quality ARGO