Artificial intelligence (AI) has advanced faster than expected over the past nine months, said AIG Chairman and CEO Peter Zaffino, who outlined the company’s next phase of AI deployment focused on developing multi-agent solutions with an orchestration layer that coordinates specialized AI agents.
During AIG’s first-quarter 2026 earnings call, Zaffino emphasized the potential of artificial intelligence to significantly improve performance and provide better solutions for customers and companies.
He said: “Our approach to using AI focuses on three important components. First, you must understand the technology and capabilities of large language models. Second, you must have pattern recognition capabilities to know how to apply AI to your business. Third, you must have a culture and execution record to effectively deploy AI within the organization.
“While we expect this technology to evolve meaningfully over time, we could not have predicted the rapid pace of development over the past nine months or the breadth of potential applications of AI.”
Zaffino highlighted the success of AIG Assist, its AI-powered generation tool, particularly in the Lexington middle-market real estate space, where it helped increase submissions by 30 percent, shorten underwriters’ offer times by 55 percent, and increase the binding force of submissions by approximately 40 percent.
He continued: “Now, with advances in inference models, AI agents can review, challenge and ultimately recommend underwriting observations so that our underwriters can make more informed decisions and provide more reliable insights to complement their experience and underwriting judgment. We are advancing our business model and AI implementation plans to capitalize on this potential.
“To illustrate the scale of recent advances in AI, when we started using Claude 2.0, AI agents could run autonomously for less than an hour. Today, they can run autonomously for up to 30 hours.”
Zaffino noted that this quarter, building on the early success of AIG Assist, AIG began the next phase of its agent AI strategy, working closely with Palantir and Anthropic.
He explains: “Using Palantir’s foundry platform, we extended our ontology, a digital map of our business that includes our underwriting processes, workflows and data relationships. This ontology, combined with orchestration, will allow us to deploy multiple AI agent teams to integrate with our core systems, which will improve decision-making and reduce costs over time.
“As a logical next step in AI deployment, we are creating a multi-agent solution with a powerful orchestration layer that coordinates specially trained AI agents to seamlessly complement our underwriter analytics and should further enhance our underwriters’ ability to assess risk and rates and deliver coverage that adjusts in real-time.
Zaffino explained that at this stage, each AI agent will be built specifically for a specific underwriting function.
“For example, one agent could handle submission ingestion and data extraction. Another agent could do risk assessments against our underwriting guidelines, another could benchmark pricing against our portfolio targets, all through collaborative agents that synthesize input from the other agents’ large language models. These agents would communicate and hand off work to each other to enhance our underwriters, much like a well-functioning underwriting team, but running at machine speed and with inherent consistency,” he said.
However, he emphasized that human oversight is critical to the underwriting process and will continue to be so, noting that AIG will be able to monitor each agent’s activities and intervene in real time if needed.
Zaffino also highlighted the potential for large language models to work alongside underwriting and claims professionals to improve decision-making, provide more timely responses and produce more accurate results due to their intuitive nature and ability to learn from the information available to claims experts.
“Overall, we are very pleased with the progress we have made and we are testing the use of the multi-agent solution to improve our team’s productivity, efficiency, and learning and development,” Zaffino said.