The Claims Adjuster's AI-Assisted Framework for Defensible SIU Referral Documentation: A Standardized Protocol

Free AI Prompts for Adjusters

Close claims faster. Download 3 copy-paste AI templates to speed up your FNOL interviews, vendor assignments, and recorded statements.

Send My Free Prompts

We respect your privacy. Unsubscribe at any time.

Built with Kit

The Problem: Why SIU Referrals Get Rejected—and What It Costs

Insurance fraud costs the U.S. property and casualty industry an estimated $308 billion annually, yet a significant percentage of legitimate SIU referrals never advance past intake—not because the claim is clean, but because the referral narrative is inadequate. Adjusters frequently know why a claim is suspicious before they can articulate how to document that suspicion in language an SIU investigator and outside counsel can act on.

The core workflow bottleneck is structural: most adjusters are trained to investigate claims, not to write evidentiary referral packages. The distinction matters enormously. A referral narrative is not a diary note—it is a threshold document that must stand alone, communicate risk with precision, and avoid language that could later be characterized as prejudging the insured. Under the NAIC Model Unfair Claims Settlement Practices Act and most state equivalents, an adjuster's handling decisions—including when and how SIU is engaged—can be scrutinized in bad faith litigation. Failure to refer and failure to refer properly both carry exposure.

Compounding this, a 2025 Deloitte survey found that 35% of insurance executives rank fraud detection among their top priorities for generative AI deployment—yet most front-line adjusters have received no standardized training on using AI to improve referral quality. The gap between carrier-level AI strategy and adjuster-level execution is where defensible claims handling breaks down.

SIU Referral Red Flag Classification Matrix

Step-by-Step Protocol: Writing a Defensible SIU Referral Narrative

Step 1 — Assemble the Factual Chronology Before Drafting

Before opening ChatGPT, build your raw timeline: FNOL date, reported loss date, policy inception date, first contact date, RS date, and any IME or EUO scheduling milestones. A referral narrative that lacks a precise chronology will fail intake review regardless of how well-written it is. Pull your ISO ClaimSearch results and prior claim history into one summary before you begin.

Step 2 — Categorize Each Red Flag by Type, Not Conclusion

Map every suspicious element to one of the established referral trigger categories (see matrix above). Do not write "the claimant appears to be lying." Instead write: "Claimant's recorded statement dated [DATE] states the vehicle was parked at [LOCATION A]. FNOL documentation dated [DATE] places the vehicle at [LOCATION B]." The referral must present facts that lead the SIU investigator to conclusions—not present conclusions that the SIU investigator must validate.

Step 3 — Use ChatGPT to Structure and Pressure-Test the Narrative

With your raw facts compiled, use a structured prompt (see examples below) to have ChatGPT draft the referral narrative. Your role is to provide the factual inputs; ChatGPT's role is to organize them into referral-appropriate language, ensure each red flag is tied to a specific document reference, and flag any conclusory language you may have inadvertently included.

Step 4 — Review for Compliance with State SIU Reporting Requirements

Most states mandate that carriers maintain a Special Investigative Unit compliant with the NAIC Insurance Fraud Prevention Model Act (Model Act #680). Several states—including California (10 CCR §2695.8), Florida (§626.9891 F.S.), and Texas (Ins. Code §701.051)—impose specific timelines and documentation standards on fraud referrals. Confirm your referral satisfies any state-specific threshold language and mandatory reporting windows before submission.

Step 5 — Diary the File for SIU Response and Reserve Impact

Once submitted, immediately diary the file for SIU acknowledgment (typically 5–10 business days carrier-dependent) and document the referral submission in the claim notes with timestamp. If the SIU investigation is expected to affect reserve adequacy, update the indemnity reserve with a documented basis. Undiarized SIU referrals are a common source of claims handling complaints during DOI market conduct examinations.

Step 6 — Document That Human Judgment Drove the Referral Decision

As AI-assisted workflows become standard, regulators in multiple states now require documentation that a qualified human professional—not an automated system—made the investigative determination. Add a brief claims note affirming your independent review of the file and your professional judgment as the basis for the referral. This single step protects the carrier from AI-overreliance arguments in subsequent bad faith litigation.

Common Mistakes That Compromise SIU Referrals

1. Using conclusory language that predetermines the outcome.
Phrases like "it is clear the claimant staged this loss" or "this appears to be a fraudulent claim" give the SIU team nothing actionable and create a documented record of bias that opposing counsel will exploit in bad faith litigation. Every assertion must be anchored to a specific, dated document in the file.

2. Submitting a referral without an ISO ClaimSearch pull or prior claim history review.
Prior claim patterns are among the most actionable intelligence an SIU investigator has at intake. A referral that omits this—even when the prior history is clean—signals an incomplete investigation and weakens the overall credibility of the referral package.

3. Failing to document the timing relationship between policy inception and loss date.
The policy inception-to-loss window is a recognized red flag indicator in every major SIU training framework, yet adjusters routinely omit this from referral narratives. Include the exact number of days between policy issuance and reported loss in every referral.

4. Neglecting to diary the file after referral submission.
An undiarized SIU referral creates a file management vacuum. If the SIU team does not acknowledge within your carrier's standard window, the adjuster is responsible for following up—and a file showing weeks of silence after referral is a liability during market conduct examinations.

5. Confusing AI-generated analysis with independent professional judgment.
As of 2025–2026, multiple state regulators have issued guidance requiring that adverse claims decisions—including SIU referrals—be driven by documented human judgment, not algorithmic output alone. Adjusters who treat AI-generated red flag summaries as a substitute for their own documented analysis expose their carrier to regulatory risk and bad faith claims.

Closing: The File You Build Today Is the Defense You Have Tomorrow

SIU referral documentation is not administrative paperwork—it is a legal record that may be produced in litigation, examined in a DOI market conduct review, or used to establish (or undermine) the carrier's good faith in handling the claim. The adjuster who writes a precise, fact-anchored, properly categorized referral narrative at day one of an investigation is the adjuster whose file survives scrutiny at year three of litigation. Caseload pressure is real, and the temptation to write a three-sentence referral and move on is understandable—but the downstream exposure from an inadequate referral far exceeds the fifteen minutes it takes to do it correctly.

AI-assisted documentation does not replace adjuster judgment. It amplifies it—enabling you to produce referral packages that are more thorough, more consistently structured, and more legally defensible than the average diary note written under caseload pressure. Used within the compliance constraints outlined in this protocol, ChatGPT is a force multiplier for exactly this kind of precision documentation work.