Travelers' Patent Lets an LLM Read the Injury File First

The USPTO granted The Travelers Indemnity Company US 12,700,485 B2 on August 4, 2026, for a system that runs a medical-record large language model against a claimant's file, extracts discrete injury events, and renders them on a scaled, interactive 3D body model. Claim 3 names the intended use directly: an "insurance claim evaluation tool."

The 3D imagery is the output. The extraction is the mechanism, and it sits exactly where casualty actuaries set the first number on a bodily-injury file.

Key Takeaways

Nothing in claim 1 requires a human reviewer between the model's extraction step and the body model that gets displayed. The LLM decides what counts as a discrete injury and where it belongs.
Colossus assigns severity points across roughly 600 to 720 coded injury types under more than 10,000 internal rules, but an adjuster still enters the facts. This mechanism removes that data-entry layer.
Bodily injury paid claim frequency rose 11% over two years while physical-damage frequency fell 7.6%, and bodily injury now accounts for 52.4% of total liability dollars paid.
The average third-party medical demand rose from $24,300 in the first quarter of 2023 to $32,300 in the first quarter of 2026, on roughly flat delta-v figures.

What Claim 1 Actually Recites

The patent, filed July 10, 2024 and issued more than two years later, runs 19 claims and assigns entirely to Travelers (Google Patents). Its own background section states the problem it was built to solve: "medical record review and analysis is accordingly still prone to human errors and still consumes a great deal of time to process." The specification frames the fix as letting untrained personnel navigate a claimant's medical history spatially rather than requiring a specialist to read a stack of records line by line.

The independent claim recites a 3D human body model controller, an augmented-reality output device and "a non-transitory data storage device... storing a medical record Large Language Model (LLM)." The controller identifies medical records tied to a claim, executes the LLM to compute discrete medical event data elements, populates a data structure mapping each event to a specific body portion, generates a proportionally scaled model, and outputs interactive overlays at the corresponding anatomical location.

The claim as written puts the generative model first in the chain. The adjuster, nurse case manager or defense counsel encounters the claim only after the LLM has already decided what counts as a discrete injury and where it belongs on the body.

It lands five weeks after TravelersLLM, a proprietary model trained on millions of internal documents, launched July 1, 2026 (Travelers Investor Relations), against technology spend of more than $1.5 billion last year with roughly half directed at cloud, analytics and AI (Coverager). The build-versus-buy case for that model is the same one this patent extends into claims.

The Case Reserve a Model Sets Before Anyone Adjusts the Claim

Initial case reserves on bodily-injury claims have historically been an adjuster's judgment call, informed by injury type, treatment codes and comparison to internal severity tables or a licensed evaluation tool. Colossus, the best-known incumbent and in use across much of the industry including Travelers, is a rules-based point system: it assigns severity points across roughly 600 to 720 coded injury types under more than 10,000 internal rules, but an adjuster still answers the questions the software asks and feeds it the facts.

This mechanism removes that data-entry step. The LLM reads the file directly and computes the discrete medical event data elements on its own, so an adjuster's first exposure to the claim can be the 3D model the LLM has already built.

The capacity case for that is real. Bodily injury paid claim frequency rose 11% over the two years through mid-2025 while physical-damage claim frequency fell 7.6%, a divergence CCC's Crash Course 2026 report calls unprecedented across coverage lines (Claims Journal, August 2026). Bodily injury paid severity climbed 10.3% in the year through the second quarter of 2025 and 32% over four years, and the line now accounts for 52.4% of total liability dollars paid, a record high.

Dimension	Rules-based (Colossus-style)	LLM extraction (US 12,700,485)
Human role	Adjuster enters coded facts; software scores them	LLM reads the file directly; adjuster reviews the output
Auditability	Point table and rules can be inspected line by line	Extraction logic is model-internal and version-dependent
Consistency source	Depends on adjuster's input accuracy	Depends on model consistency across record formats
Discovery profile	Established, decades of case law	Novel; Lokken-style rulings still setting precedent

The reserving consequence is that two effects move the same diagonal in opposite directions. One is a genuine severity signal: a nerve impingement in a radiology note that a manual read would have under-reserved, correctly raising the case reserve. The other is pure timing and format: a claim whose full injury inventory is computed and coded in days rather than the weeks a manual review took shows a stronger, denser first-diagonal case reserve purely because the coding happened sooner.

Comparing pre-adoption and post-adoption accident quarters on standard age-to-age factors reads the second as if it were the first. The correction is a reporting-pattern curve fitted to the post-adoption diagonal and held separate from any severity-trend selection. The Casualty Actuarial Society's work on individual claims reserving makes the same point from the other direction: claim-level granularity is exactly what aggregate triangle methods were built to summarize away (Casualty Actuarial Society).

Auditability Is Where the Rules Engine Was Better

Standard model risk management assumes a relatively stable benchmark: back-test predicted loss cost against actual development, hold the model's inputs and logic fixed, measure drift over time. A generative model extracting injury events from unstructured medical text does not offer that stability. Its extraction behavior is sensitive to how a record is phrased, which physician wrote it, which EHR system it came from, and which model version is running in production.

Two claimants with clinically identical injuries, documented in different note-taking styles, can plausibly get different discrete-event extractions, and there is no stable correct severity table to validate against the way a rules engine's point assignments can be audited line by line. The NAIC's Model Bulletin on the Use of Artificial Intelligence Systems by Insurers, adopted in December 2023 and taken up by more than 20 state insurance departments by mid-2026, requires ongoing monitoring rather than a one-time validation at deployment.

The consistency gain also converts into a discoverability cost. More than 70% of major carriers already use some form of claims-valuation software, and a Minnesota federal court in the Lokken case against UnitedHealth Group ruled that plaintiffs are entitled to discovery of documents describing how an AI system works, its development goals, and whether it was designed to supplant human decision-making (National Law Review, March 2026). A carrier's own patent application, which by design explains what the system does and why it was built, sits near the center of that kind of request.

The body model compounds the exposure rather than diluting it. A rules-based point total is dry and hard to dramatize. An anatomically rendered map of a claimant's injuries, generated by a system the carrier designed and patented, is a demonstrative exhibit plaintiff counsel could request in native form. Verdicts of $10 million or more rose 52% from 2023 to 2024, and Swiss Re attributes a 33% rise in liability costs from 2020 to 2024 to social inflation, which is the same severity environment behind umbrella and excess ILF repricing.

Travelers' Patent Lets an LLM Read the Injury File First

Key Takeaways

What Claim 1 Actually Recites

The Case Reserve a Model Sets Before Anyone Adjusts the Claim

Auditability Is Where the Rules Engine Was Better

Further Reading

Sources