AIG: AI efforts
AIG has made AI a centrepiece of its strategy under CEO Peter Zaffino, framing it as a growth engine for commercial underwriting. Its flagship system, built with Anthropic (Claude) and Palantir, compressed business-review time more than 5x and lifted underwriting data accuracy from ~75% to over 90%. Part of InsureBench's overview of AI in insurance; every figure is sourced.
Where AIG ranks on AI
AIG is featured in the 2026 Evident AI Index for Insurance and ranks notably high — 6th on the Leadership pillar — reflecting Zaffino's prominent public commentary on bringing AI into AIG's core.6
The flagship is an AI-assisted underwriting system built with Anthropic and Palantir that Zaffino says compressed review timelines more than 5x while improving data accuracy from 75% to over 90%.1
NA/Europe insurers
The AI build-out
From an underwriting-accuracy breakthrough to a multi-agentic underwriting build.
Accuracy breakthrough disclosed
On its Q3 call, AIG reports underwriting data accuracy improving from ~75% to over 90% in early pilots.2
Investor Day: AIG Assist
AIG details "Underwriting by AIG Assist," built with Anthropic and Palantir, and 2030 E&S growth goals.3
Anthropic partnership
Anthropic names AIG a Claude customer; Zaffino cites the 5x speed-up and 75%→90% accuracy.1
Scaling + "Auto extract"
AIG Assist scales to Lexington and more lines; a patent-pending "Auto extract" structures documents with LLMs.4
Q4: beyond expectations
370k+ submissions processed; Lexington 55% faster time-to-quote, +40% binding; a multi-agentic solution underway.5
Atlanta innovation hub
AIG plans an Atlanta AI hub for 1,000+ colleagues and 600+ new roles over five years.2
Inside the programme
A partner-built underwriting assistant, evolving toward purpose-built agents.
AIG Assist
An AI underwriting assistant built with Anthropic's Claude and Palantir's Foundry, lifting accuracy from 75% to 90%+.1
Multi-agentic build
Purpose-built agents for submission ingestion, risk evaluation, pricing benchmarking and synthesis.4
Auto extract
An LLM capability that converts unstructured submission documents into structured data.4
Data foundation
About $300M in AI/data/talent over two years, on a $1B+ five-year foundational data investment.2
"We have been able to compress the timeline to review business by more than 5x in our early rollouts while simultaneously improving our data accuracy from 75% to over 90%."
Peter Zaffino, Chairman & CEO, AIG — Jul 20251AI to grow, not to cut
AIG frames AI as a way to write more business with the same underwriters — processing far more submissions "without additional human capital," with underwriters kept in the decision. Zaffino has called the early results "beyond expectations."5
Whether those AI-assisted underwriting decisions are right is exactly what an insurance AI benchmark measures.
- Built with Anthropic (Claude) and Palantir (Foundry)
- ~$300M AI/data investment; $1B+ five-year data foundation
- 370k+ submissions processed in 2025
- Atlanta AI hub: 1,000+ colleagues, 600+ new roles
Why this matters for benchmarking
AIG is putting AI at the heart of commercial underwriting — the exact work InsureBench scores. The more those decisions are model-assisted, the more it matters whether they're right. See the methodology.
Sources
- Anthropic — Claude for Financial Services (AIG; Jul 15 2025). anthropic.com
- CIO Dive — AIG leans on generative AI to speed underwriting. ciodive.com
- Carrier Management — AIG's AI underwriting (Anthropic + Palantir; Apr 28 2025). carriermanagement.com
- Reinsurance News — Speed drives growth as gen AI accelerates underwriting (Q3 2025). reinsurancene.ws
- The Motley Fool — AIG Q4 2025 earnings call transcript (Feb 11 2026). fool.com
- Evident Insights — AI Index for Insurance 2026 report. evidentinsights.com