Fraud Triage by Lumensic

A second read of every claim file.

Fraud Triage reads a claim's documents, flags potential inconsistencies, and shows the evidence behind each flag. Your claims and SIU teams decide what happens next.

Fraud Triage is in pilot.

AI review · synthetic claim
Fraud Triage flag F3, Cross-document contradiction, High, from AI review, on a synthetic claim. The narrative says an ambulance took the claimant from the scene to the hospital emergency department, but the claim document says neither driver requested or was transported by ambulance. Quoted side by side: from the claim narrative, An ambulance took me directly from the collision scene to Briar Glen General Hospital; from the claim PDF, Neither driver requested an ambulance, and neither was transported from the scene by ambulance.

The detail that matters is rarely on page one.

A single claim file can hold a loss notice, a police report, repair estimates, photos and letters. A date that doesn't line up, or an amount that changes between documents, can sit anywhere in it. Reviewers find it when they have time, and they rarely have time.

See a finding

The flag, its reason, and the evidence behind it.

Two Fraud Triage flags from AI review on a synthetic claim, both Cross-document contradiction, High. F4: the narrative says the car was not drivable and was towed, while the claim document says the driver drove the sedan home and no tow truck was used. F5: the narrative says the claimant felt severe neck pain immediately after impact, while the document says the claimant reported no pain at the scene. Each flag quotes the claim narrative and the claim PDF side by side.
AI review · synthetic claim, fictional people
  1. 01

    Both passages, side by side

    The claim narrative and the document are quoted next to each other, with the document named, so a reviewer can check the contradiction without hunting through the file.

  2. 02

    Labeled by source and severity

    Each flag says whether the AI review or a rule check raised it, and how much it matters: Low, Medium or High.

  3. 03

    A reason to look, not a conclusion

    A contradiction can have an innocent explanation. The flag asks for a closer look; it doesn't decide anything.

Fraud Triage claim page for a fictional demo claim, DEMO-VIDEO-1042. Risk level Medium, worked out from one flag. Flag F1, Late reporting, Medium, from the rule engine: claim filed 14 days after the incident, above the typical 7-day window for auto collision. Decision buttons Clear claim and Refer to SIU, and a referral summary with a Copy summary button.
Claim page · fictional demo, no real policyholder
  1. 04

    Rule checks state their reason

    A rule flag names what was found and why it matters, such as a claim filed 14 days after the incident, and links to the evidence.

  2. 05

    A risk level that explains itself

    "One flag, none high, so Medium." The level is worked out from the current flags each time the claim loads.

  3. 06

    A decision only a person makes

    Clear the claim or refer it to SIU. The choice is recorded as your adjuster's, with a referral summary ready to copy.

What it checksIn pilot

Rule checks and an AI review, each with its reason.

Fraud Triage is built for claims teams deciding whether to refer a claim, and for the SIU teams who receive those referrals. Every result says where it came from.

Rule checks
  • Late reporting: a loss reported later than the usual window for its type
  • Policy timing: a loss soon after a policy began, or outside the policy period
  • Repeat claims: several claims from the same policyholder in a short period
  • Claimed amount: an amount above the ceiling for its loss type
  • Treatment vs. damage: treatment out of proportion to the damage claimed
AI review of the documents
  • Contradictions, quoted: where the documents and the claim narrative disagree, with the passages behind each finding
  • Every quote checked: each quote is checked against its document before it's shown
Every claim gets
  • A Low, Medium or High risk level that states its reason
  • A referral summary ready to paste into your claims system
  • A recorded decision: refer to SIU or clear
How Fraud Triage works

Every flag shows its source.

  1. 01

    Upload the claim

    Add the claim's details and up to 10 documents, 20 MB each, as PDF, JPEG or PNG. Scans and photos are read too.

  2. 02

    Review the evidence

    Rule checks and an AI review of the documents each flag what they find and say why. Document findings quote the passage they rest on.

  3. 03

    Your team decides

    Refer to SIU or clear. Fraud Triage never refers, clears or denies a claim on its own. It records the decision your adjuster makes.

Indicators, not verdicts

Every finding sourced. Every decision yours.

A flag means something deserves a closer look, not that a claim is fraudulent. Fraud Triage surfaces potential inconsistencies and the evidence behind them. Referring, clearing or denying a claim is always a person's decision.

Your files, kept like evidence.

Encrypted, with our own key

Uploaded documents and claim text are encrypted by Lumensic, on top of our providers' encryption.

Separated by organization

Every claim belongs to one organization, and the server checks it on every request.

Invitation only

No public sign-up. An administrator sets up each organization and invites its people.

Activity logged

Opening, analyzing and deciding a claim is recorded with the person and the time.

We don't have a SOC 2 report yet.

See exactly what we do today →
Pilot

A six-week pilot on claims you've already closed.

Every claim in the pilot has a known outcome, so your team can compare each flag with what actually happened.

Six weeks

One week of setup, four weeks of use, and one week to review the results with your team and decide.

Free

The pilot conditions and the post-pilot price are agreed in writing before we start.

Closed claims only

30 to 50 closed auto-collision claims with their documents. No open claims.

Up to five people

Named users from your organization, invited by your administrator.

Deleted afterward

All pilot data is deleted within 30 days of the end, and we confirm it in writing.

About Lumensic

Insurance first. Not insurance only.

Lumensic builds evidence-first software for decisions that rest on documents. Our tools read the file, point to the passage that matters, and leave the decision to a person. Fraud Triage is our first product. We'll take on other document-heavy work one clearly defined problem at a time.

  • FounderElijah Melis
  • Based inHighland Park, Illinois
  • CompanyLumensic LLC, an Illinois limited liability company
  • Contacthello@lumensic.com

Evaluate it on representative claim files.

Start with 30 to 50 closed claims your organization approves, and we'll review the findings with your team. Real claim data is used only under agreed security and data-processing terms, and deleted within 30 days of the pilot's end.