FOR INSURANCE CLAIMS AND UNDERWRITING

The demand package was drafted by AI. Verify it before you pay on it.

AI now drafts the demand letters, medical chronologies, provider narratives, and broker submissions that arrive on a desk. The document reads as the other party's position, it reads with full confidence, and every number in it is doing work on your reserve.

A treatment timeline that does not hold, a policy provision quoted to say something it does not say, a provider whose identity does not match the record, damages arithmetic that does not compute. On claims touching Medicare, Medicaid, or other federal programs, the False Claims Act adds treble damages plus per-claim statutory penalties. Asking another AI to check the draft is not a fix. In our pre-registered study of twenty company profiles, three frontier checkers disagreed by as much as 9.6 to 1 on the same document, and the harshest one returned zero errors on documents containing sixteen verifiable ones.

The AI is the witness. The math is the judge. GauntletScore verifies each claim against primary sources, then computes the trust score with deterministic math. The agents gather the evidence; no model sits in the verdict. The same evidence always produces the same score.

What GauntletScore checks

Policy and regulatory references against the governing text, so a demand or a determination does not rest on a provision quoted to say something it does not say.

Benefit and damages arithmetic with a deterministic math verifier, so specials, limits, deductibles, and reserve math hold.

Named providers and entities against the public and corporate record, so the parties in the file are who the file says they are.

The causal story of the claim because a demand is a chain of reasoning, and a chain that does not hold is what moves the number.

GauntletScore runs a dedicated analysis of cause-and-effect claims, testing injury-to-treatment attribution, timeline consistency, and logical structure. A demand whose reasoning fails the test counts heavily against the document. The engine earns its value by catching the story that reads soundly and does not hold.

What the first run looks like

Upload the document at gauntletscore.com. Minutes later you are reading the Gauntlet Report: a trust score with a 95% interval, every flagged claim with its verdict and the primary source behind it, and a cryptographically signed, tamper-evident certificate. The demand that does not hold loses its number before it moves your reserve, and the file carries an audit trail that it was independently checked, available to compliance, regulators, and any later review. In a live production run, the engine examined nineteen claims in one fluent, credible document and returned two debunked against the court record, each with the source that contradicts it. Every voting agent independently recommended against proceeding.

LIMITS

It does not make the coverage decision and it does not verify what no authoritative source can confirm. A claim it cannot check is reported as unverifiable, not as false. The decision stays yours; the record that it was checked is the product.

Where it fits

Demand review.

Run the package against the policy text and the record before it moves the reserve or drives the settlement number.

Underwriting review.

Check broker submissions and AI-drafted memos against filings and primary sources before they bind risk.

Before a denial issues.

Run the determination and the letter against the record while the decision can still change, and keep the signed certificate for the bad-faith file.

There is a demand package on your desk right now whose number you do not quite believe. Run it. Either the specifics hold and you negotiate knowing that, or one does not, and the number just changed.

Analyze a Document

Three free credits. No card. No call.

Validation study pre-registered on the Open Science Framework, March 2026; in progress, manuscript in preparation.