Can Two Versions of AIG Be Better Than One?
During a recent earnings call with analysts, American International Group (AIG) CEO Peter Zaffino shared how the company is integrating generative AI (GenAI) to create a “digital twin” of its operations — a strategic move aimed at enhancing decision-making, operational efficiency, and client service.
This digital replica will represent key data, processes, business logic, and a network of interrelationships across AIG’s business units and functions, Zaffino explained.
“The foundation of this initiative is our AIG ontology,” said Zaffino, referring to the structured framework AIG has been developing since it began its AI journey. “You’ll be hearing much more about this in the coming quarters as we continue our rollout.”
What Is Ontology in This Context?
In philosophical terms, ontology refers to the study of being and existence. In a business and AI context, it involves defining the elements a system needs and how those components interact. In AI specifically, ontology helps machine learning systems understand data and use it effectively.
Zaffino noted that the AIG ontology will serve as the backbone for technical reviews, provide enhanced tools and insights for claims teams, and improve overall service to clients.
He also emphasized that AIG’s ontology will log all actions taken, shaping business logic and enabling auditable transparency across workflows.
Strategic Tech Partnerships and Decision-Centric Architecture
Zaffino highlighted AIG’s ongoing collaborations with Palantir, Anthropic, and AWS as key partners in this AI-driven transformation.
A blog post by Palantir’s chief architect underscores the significance of ontology in enabling AI to work in real-world enterprise contexts. The post explains that most traditional data systems lack the ability to capture reasoning behind decisions or to tie those decisions to actual outcomes — limiting the effectiveness of AI.
Palantir argues that a decision-centric architecture is essential in today’s fast-moving world: “Conventional analytic architectures do not contextualize computation within lived reality, and therefore remain disconnected from operations.”
With a digital twin, AIG aims to model and test various operational scenarios and forecast outcomes — improving speed, insight, and performance.
Early Use Cases: Underwriting and Claims Processing
AIG has already begun leveraging GenAI in meaningful ways. One early use case is underwriter support for nonprofit business products, which began late in Q1.
“The early results have been very encouraging,” Zaffino said. The technology will expand to middle market accounts at Lexington Insurance Company in Q3, with broader rollout planned across North America commercial lines in 2026.
The company also showcased its AI-driven underwriting platform during its recent Investor Day. (See: AIG: Turning One Human Underwriter Into Five, ‘Turbocharging’ E&S)
Zaffino added that AIG is accelerating AI deployment in the claims process as well. Initial testing on notice-of-loss handling has shown significant efficiency gains — cutting processing time from days to hours.
“We’ve trained large language models to automatically extract and organize key insights, enabling claims adjusters to make faster, better-informed decisions,” Zaffino said. “It’s all part of fulfilling our commitment to support clients when they need us most.”
