Startups applying LLMs to insurance claims and underwriting, pre-Series B

Sourcing list — LLMs × insurance claims/underwriting, pre-Series B

1. Sixfold AI — Purpose-built generative AI for underwriters that ingests submissions and surfaces risk signals against carrier appetite. Raised a Series A (~$15M, Salesforce Ventures participation); strong design-partner logos among mid-market carriers.
2. Kyber — LLM-drafted claims correspondence (denial letters, reservation-of-rights, status notices) with compliance guardrails. Seed-stage (YC-adjacent ecosystem); attacks a painful, regulator-sensitive document workflow no one else focuses on.
3. Qumis — "AI policy expert" that reads insurance policies and answers coverage questions for brokers and claims teams. Early seed out of Chicago; founded by an ex-coverage attorney — rare domain-founder fit for policy interpretation.
4. Assured Insurance Technologies — Conversational AI-driven FNOL and claims intake for P&C carriers. Series A; live with national carriers, and FNOL is the natural wedge into full claims-file automation.
5. Liberate — Voice + LLM agents automating FNOL calls, claims status, and servicing end-to-end. Seed/A stage; voice-native claims automation is differentiated vs. the document-only crowd.
6. Curacel — Claims automation and fraud detection infrastructure for insurers in Africa, increasingly LLM-powered. YC-backed, seed stage; geographic white space with little competition and high fraud pain.
7. Inaza — AI data pipeline for auto insurers: document ingestion, fraud checks, and straight-through claims processing. Early-stage (Ireland/US); focused wedge in high-volume personal auto claims.
8. Simplifai — Norwegian company deploying "AI employees" (now generative-AI based) for claims handling at European insurers. Growth beyond seed but pre-large-round; strong Nordic/EU carrier traction where US competitors are absent.
9. Charlee.ai — Predictive + NLP claims intelligence flagging litigation risk and severity early in the claim lifecycle. Still early/bootstrapped-lean; litigation-propensity prediction is a high-ROI, underexploited niche.
10. Ledgebrook — E&S MGA using LLM-driven submission ingestion to quote specialty risks in minutes. Series A; interesting as a full-stack bet — owns the underwriting result rather than selling software.

Caveat: stage data moves fast; verify current round status (esp. Assured, Ledgebrook, Simplifai) before outreach.

White-space gaps noticed:

1. Subrogation and recovery — Almost everyone attacks intake (FNOL, submissions) or correspondence; nearly no early company applies LLMs to identifying and pursuing subrogation opportunities, a directly monetizable, success-fee-friendly workflow buried in claim files.
2. Reinsurance and treaty workflows — LLM tooling for treaty wording review, bordereaux ingestion, and cedent submission analysis is essentially empty at seed stage, despite reinsurers being document-drenched and highly concentrated (short, high-ACV sales list).