Peter Zaffino, Chairman and CEO of AIG, said artificial intelligence (AI) has advanced at a faster pace than expected over the past nine months, as he outlined the company’s next phase of AI deployment, focused on developing a multi-agentic solution with an orchestration layer coordinating specialised AI agents.
AIG Peter ZaffinoDuring AIG’s First Quarter 2026 earnings call, Zaffino highlighted that AI has the potential to materially improve performance and deliver better solutions for both clients and the firm.
He said, “Our approach to using AI has been focused on three important components. First, you have to have an understanding of the technology and capabilities of large language models. Second, you have to have pattern recognition in order to know how to apply AI to your business. And third, you have to have a culture and a track record of execution in order to effectively deploy AI within an organisation.
“While we expected the technology would develop meaningfully over time, we could not have predicted the rapid pace of advancement over the last nine months, or the breadth of AI’s potential application.”
Zaffino underscored the success of AIG Assist, its generative AI-powered tool, particularly in Lexington middle market property, where it has helped deliver a 30% increase in quoted submissions, reduced time to quote for underwriters by 55%, and increased binding of submissions by approximately 40%.
He continued, “Now, with advancements in reasoning models, AI agents can review, challenge, and eventually recommend underwriting observations so that our underwriters can make more informed decisions and provide more robust insights to supplement their experience and underwriting judgment. We’re advancing our business model and AI implementation programmes to leverage this potential.
“To illustrate the magnitude of recent advancements in AI, when we began our work with Claude 2.0, AI agents could operate autonomously for less than an hour. Today, they can run autonomously for as long as 30 hours.”
Zaffino noted that this quarter, in close collaboration with Palantir and Anthropic, AIG has begun the next phase of its agentic AI strategy, building on the early success of AIG Assist.
He explained, “Using Palantir’s foundry platform, we expanded our ontology, a digital map of our business that included our underwriting processes, workflows, and data relationships. This ontology, coupled with orchestration, will enable 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 our AI deployment, we’re creating a multi-agentic solution with a strong orchestration layer that coordinates specialised and trained AI agents to seamlessly supplement our underwriters analysis and should further augment our underwriters ability to assess risks and rate, and provide coverage with real time alignment.
Zaffino explained that in this phase, each AI agent will be purpose-built for a specific underwriting function.
“For example, one agent may handle submission ingestion and data extraction. Another may perform risk evaluation against our underwriting guidelines, and another could benchmark pricing against our portfolio targets, all with a collaboration agent to synthesise input from other agent large language models. These agents will communicate and hand-off work to each other to augment our underwriters, just like a well functioning underwriting team, but operating at machine speed and with inherent consistency,” he said.
However, he stressed that human oversight is, and will continue to be, essential to underwriting processes, noting that AIG will be able to monitor each agent’s activity and intervene in real time if needed.
Zaffino also emphasised the potential of large language models to work alongside underwriting and claims professionals to improve decision-making, deliver more timely responses, and produce more accurate outcomes, due to their intuitive nature and ability to learn from the information available to claims experts.
“Overall, we’re very pleased with the progress we’re making, and we are beta testing the use of multi agentic solutions to enhance our team’s productivity, efficiency, and learning and development,” said Zaffino.
