Steps 15, 16, 18 of the FNOL-to-reserve pipeline - fraud triage, SIU referral, subrogation/salvage - are where Shift Technology, ISO ClaimSearch, the National Insurance Crime Bureau (NICB), and Carfax do their work. The artifacts: a claim fraud score with network-analysis visualization that a SIU investigator can read; a SIU referral memo with NAIC Unfair Claims §4.1–§4.4 reason codes; an FCRA §615 adverse-action notice when fraud leads to denial relying on consumer-report-like data; a subrogation triage with liability analysis and police-report cross-reference; a total-loss salvage decision drawing on Copart and IAA data; a UM/UIM stacking analysis on multi-policy households where state law allows or restricts stacking. The worked scenario this lesson uses is a third-party staged-loss pattern in Tampa: 5 plaintiffs, the same medical clinic, the same plaintiffs' attorney firm, a soft-tissue cluster across a 6-week window, all reporting rear-end collisions at the same three intersections. Shift Technology's network analysis surfaces the cluster; ISO ClaimSearch confirms the cross-carrier pattern (the clinic appears in 47 BI claims at 6 carriers across 18 months); NICB pattern databases corroborate; SIU investigation triggers; FCRA adverse-action analysis attaches if denial fires; coverage status remains intact while legitimacy is investigated. This lesson is the artifact-level build of Steps 15-18 with the staged-loss scenario worked end-to-end, the Shift Technology vs. in-house fraud model comparison, the NAIC §4 reason-code framework for SIU referrals, the FCRA adverse-action workflow for fraud-based denials, and the subrogation triage logic that determines recovery potential on legitimate claims. The Atlanta auto BI scenario from Lessons 7-8 reappears here as the subrogation worked example.

Step 15 - Shift Technology Fraud Screen on the Staged-Loss Pattern

Shift Technology runs at intake (within 24 hours of FNOL) and continuously throughout the claim lifecycle. Each claim scores 0-100 against the carrier's fraud model combined with the cross-carrier patterns Shift maintains across its customer carrier portfolio. Network analysis surfaces clusters that single-claim review cannot detect because the cluster signal lives across claims, not within any single claim.

The Tampa staged-loss pattern surfaces over six weeks. Claim 1 arrives 2026-03-15: rear-end collision at the intersection of Dale Mabry Highway and Kennedy Boulevard, soft-tissue BI claim, claimant treated at Tampa Bay Spine and Wellness Clinic, attorney of record Smith & Garcia. Shift scores 24 (below the 35 SIU threshold); the claim routes to fast-track. Claim 2 arrives 2026-04-02: similar pattern at Westshore Boulevard and Spruce Street, same clinic, same attorney firm. Shift scores 38 - above threshold; the clinic-attorney pattern surfaces as a network-analysis hit on the second claim. Claim 3 arrives 2026-04-19: similar pattern at Hillsborough Avenue and Florida Avenue, same clinic, same attorney firm; Shift scores 47 reflecting cluster-density weighting. By Claim 5 (2026-05-08), the network-analysis cluster is unmistakable: 5 BI claims, 6-week window, three intersections (all with prior staged-loss reports in NICB's intersection database), the same Tampa Bay Spine and Wellness Clinic (which appears in 47 BI claims across 6 carriers across 18 months per ISO ClaimSearch), the same Smith & Garcia attorney firm. Shift score on Claim 5: 78.

Shift Technology's Shift Claims agentic addition. The 2026 Shift Claims agentic module - extended to Covéa's 2026 production deployment among other carriers - augments the underlying network-analysis engine with autonomous investigation drafting: the agent assembles the SIU referral memo from cluster evidence, drafts the recorded-statement question set, queues the FCRA adverse-action analysis, and surfaces the comparable-cluster history from Shift's cross-carrier database. SIU investigator engagement remains non-delegable per NAIC §4.3, but the agent compresses the 90-120 minute SIU referral preparation to 15-25 minutes.

The Shift Technology vs. in-house fraud model comparison. Shift: cross-carrier network data Shift maintains across customer carriers + clinic/attorney/repair-shop databases + public records + Shift's proprietary models trained on multi-carrier ground truth. Strength: catches cross-carrier patterns no single carrier can see; the Tampa clinic's 47-claim footprint across 6 carriers is invisible to any one of those carriers. Pricing 2026: $400K-$1.5M annually for a mid-size carrier depending on premium volume and module scope. In-house fraud models: trained on the carrier's own claim data; effective on patterns visible in single-carrier data (intra-carrier repeat claimants, intra-carrier provider density); cannot detect cross-carrier clinic patterns. Most carriers run Shift Technology + an in-house model in parallel: Shift catches cross-carrier patterns, the in-house catches carrier-specific anomalies; cross-validation surfaces high-confidence flags that route to SIU automatically, mid-confidence flags that route to claims supervisor for triage, low-confidence flags that log for downstream monitoring.

ISO ClaimSearch - The Cross-Carrier Confirmation

ISO ClaimSearch is the industry's cross-carrier claim-history database operated by Verisk. Carriers contribute claim data; carriers query for matches at intake and at any point in the claim lifecycle. Shift Technology incorporates ISO ClaimSearch data into its network analysis, but the standalone query also runs at SIU referral as an independent verification source.

The Tampa Bay Spine and Wellness Clinic search. ISO ClaimSearch query on the clinic returns 47 BI claims across 6 carriers across 18 months. Cluster characteristics emerge from the cross-carrier roll-up: all soft-tissue (no objective injury indicators like MRI findings, hospital admissions, surgery, or imaging studies); all closed at $8K-$22K settlement range (the soft-tissue settlement band that triggers minimal carrier scrutiny); all represented by 1 of 3 attorney firms (Smith & Garcia, Rodriguez Law, Thompson & Associates - which share a marketing relationship per Florida Bar advertising filings and public records); all claimants treated for 4-8 weeks of physical therapy at the clinic; all involve rear-end collisions at 7 specific intersections in Tampa. The cluster signal is statistical: the clinic appears at 8x the expected BI-claim rate for clinics of its size and location per a regional baseline Shift computes.

The NICB referral and pattern database. National Insurance Crime Bureau (NICB) maintains additional fraud databases beyond ISO ClaimSearch. NICB query on the cluster returns: the clinic owner is a prior NICB target of interest (2023-2024 investigation closed without charges but pattern noted in the carrier-shared intelligence brief); 2 of the 3 attorneys have prior NICB pattern references in the legal-network database; 4 of 7 intersections appear in NICB's staged-loss intersection database; NICB's recommendation engine returns "full SIU investigation warranted; consider multi-carrier coordination."

Step 16 - SIU Referral with NAIC §4 Reason Codes

The SIU referral memo cites NAIC Unfair Claims Settlement Practices Model Act §4 reason codes explicitly. AI drafts the memo from the Shift + ISO + NICB outputs; the SIU investigator reviews, customizes, and signs. The §4 reason-code structure protects the carrier from later allegations of pretextual investigation and also ensures the SIU process is documented for any downstream FCRA adverse-action notice.

The Claim 5 SIU referral memo - structured sections. (1) Claim summary - claimant name, claim number, loss date, location, alleged facts as stated by claimant in recorded statement. (2) Fraud signals - Shift Technology score 78 with network-analysis visualization showing cluster membership and density; ISO ClaimSearch cross-carrier match indicators (clinic appears in 47 BI claims at 6 carriers across 18 months at 8x expected rate); NICB pattern references on clinic, attorney firms, and intersections. (3) Cluster characteristics - same clinic, same attorney firm, same intersection pattern, soft-tissue only without objective injury indicators, settlement range $8K-$22K. (4) NAIC §4 reason codes - §4.1 indicia of fraud (network clustering above statistical baseline and corroborated by NICB), §4.2 supporting documentation (Shift score + ISO ClaimSearch query result + NICB pattern outputs all preserved with timestamps), §4.3 non-delegable judgment (SIU investigator decides; AI provides signals only), §4.4 audit trail (full reasoning captured with invocation IDs for Shift and ISO queries). (5) Recommended investigation - recorded statements from each of the 5 claimants by SIU investigator with experienced fraud line of questioning; attorney correspondence review for marketing connections; clinic treatment-record review for templated language indicating boilerplate diagnoses; intersection accident-reconstruction analysis (vehicle damage profiles consistent with low-speed staged impacts); NICB collaboration for multi-carrier coordination. (6) Coverage status - coverage remains intact during investigation; the investigation is on legitimacy of the loss, not on coverage applicability; reservation of rights is not triggered at this stage and would be premature. (7) FCRA §615 adverse-action analysis - if investigation produces denial, FCRA notice prepared because the fraud determination relies on consumer-report-like data (ISO ClaimSearch and Shift Technology aggregate public records and cross-carrier data; under FCRA §603(d) the qualification is debatable but the conservative posture is to issue the notice).

The SIU investigator's substantive engagement. The SIU investigator reads the AI-drafted memo, validates each fraud signal against the source data (logs into ISO ClaimSearch directly, runs the NICB query directly, reviews the Shift network visualization), customizes the recommended investigation plan, signs the memo. The §4.3 non-delegable judgment requires this engagement; "rubber-stamping" the AI draft is the most common audit finding in 2026 SIU programs and the most common bad-faith vulnerability in pre-trial discovery.

The FCRA Adverse-Action Workflow

FCRA (Fair Credit Reporting Act) §615 applies when adverse action against a consumer relies on consumer-report or consumer-report-like data. Shift Technology's network analysis incorporates prior-claim history, public records, and cross-carrier data - potentially qualifying as consumer-report-like for FCRA purposes. The conservative posture is to issue the FCRA notice on any fraud-based denial that relied on these data sources.

The FCRA workflow at SIU referral. The Step 16 memo flags whether FCRA adverse-action notice may apply if investigation leads to denial. Decision tree: did the determination rely on consumer-report or consumer-report-like data? If yes, the FCRA notice is required at denial including: right to dispute, source of information (Shift Technology + ISO ClaimSearch + NICB cited), specific reasons for adverse determination, claimant copy of report if requested (Shift and ISO will both produce a consumer-facing report on request), 60-day dispute window. The notice is delivered with the denial letter via certified mail; the carrier maintains response capability for disputes via a dedicated FCRA dispute desk.

The Tampa cluster FCRA posture. If SIU investigation produces denials on the 5 claimants, the FCRA notice prepares for each: right to dispute, sources cited (Shift Technology cluster analysis with score, ISO ClaimSearch 47-claim cross-carrier match, NICB pattern references), specific reasons (network clustering above statistical baseline; soft-tissue only without objective injury indicators; attorney-clinic-intersection pattern matches established fraud schemes per NICB), 60-day dispute window starts. The notices deliver with denial letters; dispute windows start; SIU investigator + claims supervisor + compliance officer review denial timing and FCRA notice content jointly before send. Disputes that arrive within the window route to the FCRA dispute desk for substantive response within 30 days.

The FCRA pre-notice consideration. Some FCRA practitioners argue that pre-notice (advising the claimant before adverse action that consumer-report-like data may be used) is best practice even when not strictly required. The carrier's compliance committee makes the call; in 2026, most carriers running Shift + ISO + NICB stacks issue pre-notice at the SIU recorded-statement stage as a defensive posture, explaining that the investigation may rely on cross-carrier data and inviting the claimant to provide any context or correction.

Step 18 - Subrogation Triage on Legitimate Claims

While the fraud screen runs in parallel with reserve setting on suspicious claims (Steps 13-15 concurrent), subrogation triage at Step 18 fires for legitimate claims with third-party liability exposure. The Atlanta auto BI claim (Lessons 7-8 from this chapter) is the worked subrogation example.

Atlanta subrogation triage. Liability profile: the insured asserts the other driver ran a red light. The police report (received Day 3) confirms the other driver was cited for failure to obey a traffic signal; the police narrative supports the insured's account. The other driver's insurance: identified at scene as Geico; the insured shared the scene exchange information. Vehicle history check via Carfax: the other driver's vehicle (a 2019 model) has a clean history with no salvage, no rebuild, no prior reported claims. Subrogation potential: high (clear liability against the third party). Recovery target: PD $14,200 (from Tractable damage estimate) plus the BI portion the carrier ultimately pays on the insured's medical and indemnity.

The subrogation workflow. AI drafts the subrogation demand letter to Geico citing the police report, vehicle damage estimate from Tractable + CCC line-item, and BI exposure as it develops. Carfax data confirms vehicle ownership chain. The subrogation team reviews and sends. Geico responds within 30-60 days typical for a clean-liability claim. If Geico accepts liability, recovery proceeds via inter-company settlement. If Geico contests, subrogation moves to inter-company arbitration (most states participate in Arbitration Forums Inc. or similar) or litigation if the recovery target is above the arbitration cap. Atlanta scenario: clear liability + cooperative carrier → expected 60-90 day recovery cycle on the PD; the BI subrogation runs longer because the BI is still developing.

The §4 reason chain at subrogation triage. AI subrogation triage produces: liability analysis with source references (police report citation, Carfax vehicle history report, recorded statement transcript), recovery potential estimate with confidence, recommended action (demand letter, IC arbitration filing, or litigation referral), comparable-recovery references from the carrier's subrogation database. Subrogation team reviews and signs. The §4 reason chain captures the analysis for downstream tracking and post-recovery analytics that feed the next quarterly subrogation-strategy review.

Total-Loss Salvage Decision and UM/UIM Stacking

For total-loss vehicle claims, the salvage decision determines whether the carrier sells the salvage vehicle (recovering approximately 25-40% of ACV) or takes a different path such as retained-salvage arrangement with the insured. AI surfaces the salvage market comparable: Copart auction data, IAA salvage data, regional title-status implications, and current scrap-metal pricing trends.

The salvage decision logic. Tractable's repair-vs.-total-loss probability output (74% repair / 26% total in the Atlanta case) feeds the salvage decision when total is declared. For a $28,400 ACV vehicle declared total at the 70% threshold, the salvage estimate at Copart in the Atlanta market may run $7,100-$11,400 depending on the title status, the airbag-deployment status, and the structural-damage profile. The salvage value reduces the carrier's net indemnity exposure to the insured. CCC's total-loss workflow integrates the salvage valuation; Snapsheet's total-loss module provides the comparable-market valuation; the carrier's salvage vendor agreement determines the operational handoff.

The UM/UIM stacking analysis. When the insured has multiple policies in the household (e.g., own auto + family auto + an umbrella), UM/UIM coverage may stack to higher limits than a single policy. AI analyzes household-policy structure, applicable state UM/UIM stacking law, and produces a stacking recommendation. State-specific: California allows stacking under certain conditions per Insurance Code; Texas restricts stacking on most personal auto policies; Florida varies by policy type and the post-2020 limited-stacking statute; New York generally disallows stacking on policies of the same insurer; Pennsylvania allows stacking unless waived in writing with specific statutory language. The stacking analysis affects reserve adequacy and settlement strategy when the third party's liability is partial or uncertain. The AI's recommendation routes to the senior adjuster or coverage counsel for verification because the stacking statutes are intricate and the consumer-friendly interpretation is the regulator-favored default.

The Six-Month Tampa Cluster Resolution

Six months after the first Tampa claim arrival, the cluster status: 5 claims under SIU investigation through the Step 16 referral; 3 closed with denial (FCRA notices delivered, no disputes filed by claimants within the 60-day window - a pattern suggesting weak underlying claims that would not survive substantive review); 1 closed with settlement at the low end of the soft-tissue range after the claimant produced limited supporting documentation that was nonetheless above the inferred-fraud threshold; 1 still under investigation pending a deposition of the treating physician. NICB collaboration produced 2 additional carrier-side investigations of Tampa Bay Spine and Wellness Clinic. Aggregate carrier savings on the 5-claim cluster: $58K-$95K versus unrestricted settlement at industry-average soft-tissue cluster levels of $14K-$22K per claim. The Shift Technology + ISO ClaimSearch + NICB stack produced the savings; without the stack, the claims would have settled in routine workflow at the $8K-$22K range each and the cluster signal would not have surfaced until the post-settlement actuarial analysis at the next quarterly reserve review.

The cross-carrier coordination dimension. NICB's collaborative posture facilitated the multi-carrier review of the Tampa Bay Spine and Wellness Clinic. The 6 carriers contributing the original 47-claim cross-carrier match data each received a NICB intelligence brief; 2 launched their own SIU investigations; the cumulative law-enforcement referral built the case for a coordinated Florida Division of Insurance Fraud action that initiated in late 2026. The Tampa cluster moved from a single-carrier 5-claim defense into a coordinated industry-and-regulator action that benefits the entire market.

Key Takeaways

• Shift Technology fraud screen runs at intake and continuously throughout claim lifecycle. Score 0-100 against carrier model + cross-carrier patterns. SIU threshold typically 35; above threshold triggers SIU referral. Tampa Claim 5: score 78 with network-analysis cluster visualization. The 2026 Shift Claims agentic module compresses SIU referral preparation from 90-120 minutes to 15-25 minutes while preserving the SIU investigator's non-delegable judgment under NAIC §4.3.
• ISO ClaimSearch (Verisk) confirms cross-carrier patterns no single carrier sees. Tampa Bay Spine and Wellness Clinic appears in 47 BI claims at 6 carriers across 18 months - 8x the expected rate for a clinic of its size and location per Shift's regional baseline. The pattern surfaces from cross-carrier aggregation, not from any single carrier's data.
• NICB collaboration adds public-record and fraud-database references and enables multi-carrier coordination. Clinic owner is a prior NICB target of interest; 2 of 3 attorney firms have prior pattern references; 4 of 7 intersections appear in NICB's staged-loss intersection database. NICB's recommendation engine returned "full SIU investigation warranted; consider multi-carrier coordination."
• SIU referral memo at Step 16 cites NAIC §4 reason codes explicitly. §4.1 indicia of fraud (network clustering above statistical baseline and NICB corroboration), §4.2 supporting documentation (Shift + ISO + NICB outputs preserved with timestamps and invocation IDs), §4.3 non-delegable judgment (SIU investigator decides; AI provides signals), §4.4 audit trail (full reasoning captured). The SIU investigator's substantive engagement is the load-bearing element against the rubber-stamping audit finding.
• FCRA §615 adverse-action workflow attaches when investigation produces denials relying on consumer-report-like data. Notice includes right to dispute, sources cited (Shift, ISO, NICB), specific reasons (clustering, soft-tissue only without objective indicators, pattern matches per NICB), 60-day dispute window. Delivered with denial letter via certified mail. Pre-notice at SIU recorded-statement stage is becoming a defensive best practice in 2026.
• Step 18 subrogation triage fires on legitimate claims with third-party liability. Atlanta scenario: clear liability against Geico-insured other driver per police report (citation for failure to obey signal) + Carfax vehicle history confirmation. AI drafts subrogation demand; recovery target PD $14,200 plus the BI portion as it develops; expected 60-90 day cycle on PD; BI runs longer as it develops.
• Total-loss salvage decision uses Copart + IAA salvage data for ACV recovery (25-40%). Tractable's repair-vs.-total probability feeds the salvage decision; CCC's total-loss workflow integrates the valuation; Snapsheet provides comparable-market valuation. UM/UIM stacking analysis varies sharply by state - California allows under conditions, Texas restricts on most personal auto, Florida limits per the post-2020 statute, New York generally disallows same-insurer stacking, Pennsylvania allows unless waived with specific statutory language.
• Tampa cluster 6-month resolution: $58K-$95K aggregate savings on 5 claims vs. routine settlement. 3 closed with denial and no FCRA disputes; 1 settled low-end with limited documentation; 1 still investigating. NICB collaboration enabled 2 additional carrier-side investigations and built the case for a coordinated Florida Division of Insurance Fraud action in late 2026. The Shift + ISO + NICB stack produced the savings; the cross-carrier coordination produced the regulator action that benefits the entire market.