American International Group (AIG) made substantial progress in artificial intelligence within the last year, and its outcomes and capabilities are more than previously predicted.
The insurer during an Investor Day less than a year ago presented the potential benefits from its use of generative AI. At the time, CEO Peter Zaffino said these thoughts were “aspirational.”
Less than a year later, during a recent conference call with analysts to talk about fourth quarter earnings, Zaffino said, “We see the capabilities are much greater.”
Related: AIG Underwriting Income Up 48% in Q4 on North America Commercial
“We’re seeing a massive shift in our ability to process a significant submission flow way beyond our expectations without additional human capital resources,” Zaffino said. “That has been the biggest surprise.”
Zaffino said AIG in 2025 “made significant progress embedding generative AI across our core underwriting and claims processes, and expending it,” with goals of deploying its gen AI tool AIG Assist in a majority of its commercial lines businesses.
As an example, Zaffino said AIG Assist was used in its excess & surplus company, Lexington Insurance. At Investor Day, AIG revealed the “ambition” of reaching 500,000 submissions by 2030. Zaffino said Lexington has already reached over 370,000 submissions.
“I think the acceleration and the opportunity is greater than I thought at Investor Day,” Zaffino said during the call. He spoke of training needed to manage the data coming in at a fraction of the time, and the “orchestration of [AI] agents” to “be able to scale and be able to analyze that information, that’s not biased in any way, that’s through the entire workflow.”
The front-to-back workflow can be shrunken with the implementation of gen AI and “multiple agents with a proper orchestration,” said Zaffino. Orchestration in AI terms is the coordination and management of AI models and systems. Zaffino said AIG didn’t talk much about AI orchestration at Investor Day but the company has developed this layer “to coordinate AI agents to drive better decision-making and reduce costs across the organization.”
He described AI agents “as companions that operate alongside our teams.” The agents can offer information in real time, advise based on historical use cases, and challenge the underwriter’s decisions. This orchestration can “streamline simple, repetitive, and lengthy processes to support decision-making,” Zaffino said.
Related: AIG’s Zaffino to Step Down as CEO as Aon’s Andersen Steps In
AIG has used its gen AI capabilities in the conversion of Everest’s retail commercial business. According to Zaffino, accounts were prioritized for renewals “in a fraction of the time.” The insurer built out an ontology of Everest’s portfolio, and coupled it with the ontology of AIG’s businesses, which “allowed us to prioritize how the portfolios could blend together,” he added.
The recent launch of Lloyd’s Syndicate 2479 in a partnership with Amwins and Blackstone, marked the first time AIG deployed gen AI for a special purpose vehicle (SPV). With Palantir, large language models were used to determine whether Amwin’s program portfolio aligned with the risk appetite of Syndicate 2479. Zaffino said AIG has a “strong pipeline of SPV opportunities.”
Topics
InsurTech
Data Driven
Artificial Intelligence
AIG
Interested in Ai?
Get automatic alerts for this topic.
