Headquartered in New York, AIG is a global insurer providing commercial and personal property-casualty coverage across Liability, Financial Lines, Property, Global Specialty, Crop Risk Services, Personal Lines, and Accident & Health. As of December 31, 2025, the company employed approximately 22,100 people across 45 countries and underwrote business in more than 200 countries and jurisdictions through its global network.
AIG reported $3.1 billion in net income for full-year 2025. General Insurance recorded $23.7 billion in net premiums written, while its combined ratio improved by 1.7 percentage points year over year, from 91.8% to 90.1%. Global Commercial net premiums written increased 4% on a reported basis to $17.4 billion, supported by 9% growth in new business.
Against this backdrop, artificial intelligence has become a significant part of AIG’s broader transformation strategy. In its 2025 Annual Report, the company identified deploying and scaling agentic AI solutions as a strategic priority to improve processes and decision-making across underwriting and claims. During 2025, AIG expanded its work with technology partners including Palantir, Anthropic, AWS, and Google, embedding AI capabilities more deeply into its insurance operations.
Two initiatives illustrate how that strategy is taking shape. The first, AIG Assist, applies generative AI to underwriting submission triage, helping underwriters process and prioritize large volumes of complex submissions. The second uses portfolio ontologies developed with Palantir to create machine-readable representations of insurance portfolios, allowing AIG and its partners to analyze risk and deploy capacity faster.
Accelerating Underwriting Submission Triage With AIG Assist
Complex commercial underwriting has long presented a capacity challenge for insurers. Adding more underwriters can increase the number of submissions that can be reviewed, but it does not necessarily address the underlying amount of time required to read, reconcile, and assess each submission.
AIG’s investor materials indicate that manually reviewing a complex commercial submission can take three to four weeks. The scale of the challenge becomes particularly clear when looking at Lexington, AIG’s excess and surplus lines business. In 2024, Lexington received approximately 300,000 new-business submissions but bound policies for only about 2% of them, or roughly 6,700 policies, generating approximately $1 billion in new-business premium.
AIG’s ambition for Lexington is substantially larger. By 2030, the company has outlined a goal of reaching 500,000 submissions, increasing its bind rate to 6%, and generating $4 billion in new-business premium. Achieving those targets requires more than simply expanding the underwriting workforce. AIG needs to process more submissions while also improving its ability to identify and convert opportunities that fit its underwriting strategy.
That is where AIG Assist comes in.
Commercial insurance submissions typically arrive as an unstructured collection of broker cover letters, statements of value, loss runs, supplemental applications, and other documents. These materials can vary considerably in format and completeness. Traditionally, underwriters have had to manually review and reconcile the information before determining whether a risk should be quoted and under what terms.
AIG Assist is designed to change that initial process. Its patent-pending Auto Extract capability uses large language models to extract structured information from unstructured submission documents. The system then evaluates submissions against the relevant business’s stated risk appetite, prioritizes opportunities for review, and creates curated summaries for underwriters.
Instead of working through submissions primarily in the order they arrive, underwriters can begin with opportunities that appear more closely aligned with the insurer’s underwriting criteria. The technology effectively moves part of the information-gathering and preliminary assessment process upstream, allowing human underwriters to spend more time on risk selection, pricing, coverage, and policy structure.
AIG has expanded AIG Assist incrementally rather than attempting an enterprise-wide launch all at once. The technology initially entered production within Financial Lines, including the company’s Private Not-for-Profit business. AIG later reported that the system was reviewing 100% of applicable submissions in that operation.
The technology was subsequently introduced into Lexington’s middle-market property and casualty operations. In November 2025, AIG said it expected to complete deployment across Lexington’s remaining wholesale business by the end of the year. The company also accelerated its planned rollout across North American, U.K., and EMEA commercial lines by six months.
The approach is also being explored on the claims side. AIG has piloted similar document-extraction capabilities intended to reduce the time between receiving a first notice of loss and issuing a coverage letter.
The operational impact for underwriters is significant. Rather than starting with an unsorted collection of documents, an underwriter can receive a prioritized submission containing extracted information and an AI-generated summary. This can reduce the amount of time spent on document reconciliation and data entry while allowing human expertise to remain focused on the decisions that require judgment.
AIG is also developing a more advanced multi-agent underwriting architecture with Palantir and Anthropic. On AIG’s first-quarter 2026 earnings call, CEO Peter Zaffino described a proposed system in which specialized agents could handle different parts of the underwriting workflow. One agent could ingest submissions and extract information, another could evaluate risks against underwriting guidelines, another could benchmark pricing against portfolio targets, and another could synthesize the results for an underwriter.
AIG is developing an orchestration layer to coordinate these agents and their handoffs. The company has emphasized that the proposed system is intended to supplement underwriters rather than replace them. Human oversight would remain central to consequential decisions involving risk, pricing, and coverage.
Importantly, AIG has characterized this multi-agent architecture as a system under development rather than a fully deployed production capability.
Measuring the Impact of AI-Assisted Underwriting
AIG’s early disclosures suggest that AIG Assist is affecting both the speed and breadth of underwriting operations.
In one early deployment, AIG reported that submission turnaround time fell from three to four weeks to less than one day. The proportion of applicable submissions reviewed increased to 100%, rather than a filtered subset. Zaffino also reported that data-extraction accuracy improved from approximately 75% to more than 90%, while processing time was substantially reduced.
The company continued to report progress as deployment expanded. By the end of 2025, Lexington had received more than 370,000 submissions, representing a 26% year-over-year increase and moving the business closer to its long-term goal of 500,000 submissions. AIG also reported a 35% improvement in the submit-to-bind ratio for Lexington’s middle-market property business following the rollout of AIG Assist.
On AIG’s first-quarter 2026 earnings call, Zaffino said the technology had helped Lexington’s middle-market property business quote 30% more submissions, reduce underwriters’ time to quote by 55%, and increase the number of submissions bound by approximately 40%.
Taken together, these figures indicate that the technology may be influencing both operational capacity and commercial conversion. More submissions can be reviewed, underwriters can reach decisions faster, and a greater number of opportunities can potentially move through the underwriting funnel.
However, the figures should be interpreted carefully. AIG has disclosed these results through investor materials and earnings calls, and they have not been independently validated. The company has also not disclosed enough underlying volume, measurement-period, or operational detail to isolate the precise contribution of AIG Assist from other changes occurring across the business.
Even with those limitations, the deployment pattern is clear. AIG began with selected underwriting businesses, expanded into additional lines, and has reported improvements in review speed, submission coverage, and conversion as adoption increased.
Using Portfolio Ontologies to Speed Capacity Deployment
AIG’s second major AI use case addresses a different challenge. Instead of focusing on an individual submission, the company is working to create structured, queryable representations of entire insurance portfolios.
AIG refers to these representations as ontologies. Built using Palantir’s Foundry platform, the models can integrate information such as insured risks, exposures, policy limits, attachment points, modeled losses, and underwriting rules. The objective is to create a common digital representation of a portfolio that can be analyzed consistently without requiring teams to manually assemble information from multiple systems for every new review.
This approach becomes particularly valuable when insurers are evaluating large books of business, acquiring renewal rights, deploying third-party capital, or managing capacity through distribution partnerships.
AIG has disclosed several applications of this ontology-based approach, including its work surrounding the Everest renewal-rights transaction, Lloyd’s Syndicate 2479, and its collaboration with McGill and Partners.
In October 2025, AIG agreed to acquire renewal rights for most of Everest Group’s global retail commercial insurance portfolios, representing approximately $2 billion in premium. Everest retained responsibility for the liabilities and claims administration associated with its existing policies, while AIG gained the opportunity to offer coverage to eligible accounts at renewal.
To support the transaction, AIG developed what it called an “Everest ontology,” creating a digital model of the portfolio. The model enabled underwriters to evaluate account limits, attachment points, pricing, and other characteristics to determine how the acquired renewal opportunities could fit within AIG’s existing portfolio.
The approach was subsequently applied to a different insurance and capital structure through Syndicate 2479.
AIG worked with Palantir, Amwins, and funds managed by Blackstone to establish Syndicate 2479, a special-purpose vehicle at Lloyd’s managed by Talbot Underwriting. The syndicate was established to underwrite $300 million in premium from a diversified portion of Amwins’ approximately $6 billion delegated-authority portfolio beginning January 1, 2026.
AIG used Palantir Foundry to validate its portfolio analysis before launch and developed an ontology that enabled large language models to access more than four million industry data points. AIG has described the initial portfolio analysis as completed, while positioning the use of multiple AI agents to retrieve data, evaluate risk characteristics, and test programs against the syndicate’s risk appetite as a longer-term capability.
A third application emerged in March 2026, when AIG announced a collaboration with specialty broker McGill and Partners. Under the arrangement, AIG expects to provide 25% capacity across up to $1.6 billion of McGill’s specialty gross premiums written.
Following analysis of the portfolio, AIG developed underwriting criteria designed to support real-time underwriting through McGill’s digital broking platform. AIG and Palantir also built an ontology intended to provide near-real-time information about exposures, deployed limits, modeled risk, and losses. This gives AIG a framework for monitoring portfolio performance and managing capacity as the underlying risk profile changes.
From Individual Submissions to Portfolio-Level Intelligence
The ontology strategy represents a broader evolution of AIG’s AI operating model.
AIG Assist focuses on accelerating the analysis of individual submissions. Portfolio ontologies operate at the level of an entire book of business. Both approaches seek to transform fragmented insurance information into structured data that can be analyzed more quickly and consistently.
For portfolio managers, the goal is to integrate exposure, limit, modeled risk, and loss information into a shared model rather than repeatedly assembling those details from separate systems. Accounts and programs can then be evaluated against defined underwriting criteria and portfolio-level risk appetite.
The architecture also creates the foundation for AI agents to interact with portfolio information. In the longer term, agents could retrieve relevant data, evaluate risk characteristics, monitor changes in exposure, and help decision-makers understand how additional capacity could affect the portfolio.
That does not mean AIG has eliminated account-level underwriting review or created a system that autonomously allocates capital. The company’s public disclosures instead describe the technology as decision-support infrastructure intended to make portfolio analysis faster, more consistent, and more responsive to changing exposure data.
The three arrangements also demonstrate how the same underlying technology can support different insurance and capital structures.
In the Everest transaction, AIG acquired renewal rights and said it could write qualifying policies on its existing balance sheet without requiring incremental capital. In Syndicate 2479, Amwins and funds managed by Blackstone provide third-party capital for a portfolio managed by AIG through a Lloyd’s vehicle. In the McGill collaboration, AIG provides capacity to qualifying risks distributed through the broker’s digital platform.
Palantir provides the Foundry technology used to organize and analyze portfolio data, while AIG contributes its underwriting criteria, risk appetite, and portfolio expertise. Amwins and McGill contribute distribution relationships and portfolio data, while Amwins and Blackstone provide capital for Syndicate 2479. Talbot serves as the Lloyd’s managing agent for that syndicate.
Building an AI-Enabled Insurance Operating Model
The significance of AIG’s strategy extends beyond individual AI tools. The company is building an operating model in which structured data, AI-assisted workflows, human expertise, and portfolio intelligence work together.
AIG Assist demonstrates how generative AI can address one of the most time-consuming parts of commercial underwriting: understanding and organizing unstructured information. Portfolio ontologies take the concept further by creating a digital representation of risk across an entire book of business.
Together, these approaches point toward an insurance workflow in which AI can help move information through the organization faster while keeping consequential decisions under human control.
The potential business impact is substantial. Faster submission review can increase underwriting capacity without requiring proportional increases in headcount. Better prioritization can help underwriters focus on opportunities that align with risk appetite. Structured portfolio data can make it easier to evaluate new books of business, manage third-party capital, and adjust capacity as exposures change.
But the technology remains at different stages of maturity. AIG Assist has accumulated meaningful deployment experience and publicly reported performance improvements, while the broader multi-agent underwriting architecture remains under development. The ontology-based capacity strategy has demonstrated a repeatable technical approach across several partnerships, but long-term underwriting results have not yet been disclosed.
AIG has not yet reported loss ratios, retention rates, capacity utilization, or other long-term underwriting performance measures attributable specifically to its ontology-based processes. As a result, the available evidence establishes the operating structure and intended model more clearly than it establishes the ultimate financial return.
The Next Phase of AI in Insurance
AIG’s AI strategy illustrates a broader shift taking place across commercial insurance. The focus is moving beyond isolated automation tools toward systems that can understand documents, connect information across portfolios, coordinate specialized AI agents, and provide decision support to experienced professionals.
The company’s work with AIG Assist shows how generative AI can reduce the administrative burden associated with underwriting submissions. Its work with Palantir demonstrates how structured portfolio models can support more sophisticated analysis of risk, capacity, and capital.
The common thread is not the replacement of insurance expertise but the creation of infrastructure that allows that expertise to operate at greater scale.
For AIG, the long-term opportunity is to process more opportunities, respond faster to brokers and clients, deploy capital more efficiently, and give underwriters better information at the moment decisions need to be made. Whether these systems ultimately deliver sustained improvements in underwriting profitability and portfolio performance will depend on how effectively AIG combines AI capabilities with its underwriting discipline.
For now, the company’s deployments provide a clear indication of where large commercial insurers are heading: toward AI systems that do more than automate individual tasks and instead become part of the operating architecture through which insurance risk is evaluated, managed, and ultimately underwritten.