AI experts sharing free tutorials to accelerate your business.
Back to Insurance toolkit

Catastrophe Claim Surge Triage

Turn a flood of catastrophe (CAT) FNOLs into a prioritized, routable triage queue during surge events — hurricanes, wildfires, severe convective storms, winter freezes, floods, hail, earthquakes, and named perils. Produces a per-claim triage record (severity band, coverage sanity check, vendor/adjuster recommendation, reserve seed, communication plan) and a batch-level dashboard (queue by severity, ZIP cluster, assignment status, SLA risk) so CAT desks, field network managers, and customer service can move from "everything is urgent" to "these 20 first, these 200 next, these 2,000 by Friday." Output is personalized to the carrier's named CAT event library, per-peril severity thresholds, field capacity roster, contractor networks, authority limits, and reserve-category labels drawn directly from `config.yml`.

Saves ~20 min/file + faster surge cycle timeintermediate Claude · ChatGPT · Gemini

Catastrophe Claim Surge Triage

Purpose

Turn a flood of catastrophe (CAT) FNOLs into a prioritized, routable triage queue during surge events — hurricanes, wildfires, severe convective storms, winter freezes, floods, hail, earthquakes, and named perils. Produces a per-claim triage record (severity band, coverage sanity check, vendor/adjuster recommendation, reserve seed, communication plan) and a batch-level dashboard (queue by severity, ZIP cluster, assignment status, SLA risk) so CAT desks, field network managers, and customer service can move from "everything is urgent" to "these 20 first, these 200 next, these 2,000 by Friday." Output is personalized to the carrier's named CAT event library, per-peril severity thresholds, field capacity roster, contractor networks, authority limits, and reserve-category labels drawn directly from config.yml.

When to Use

Use this skill when the carrier or TPA declares a CAT event and the FNOL queue volume jumps above normal desk capacity. Typical triggers: hurricane landfall, wildfire evacuation perimeter, derecho/hail swath, freeze burst-pipe surge, earthquake felt-report spike. Works for personal property (HO, dwelling fire, renters, flood), auto (comprehensive glass/hail/total), commercial property (BOP, CPP, BI), farm/ranch, and inland marine. Pairs with the FNOL Intake Assistant v2.0 (which captures the raw report), the Claims Reserve Recommender v2.0 (which takes the seeded reserve into a fuller recommendation once the file has enough data), the Claims Narrative Drafter v4.0 (file-note authoring once the file is open), the Coverage Explanation Letter v3.0 (insured-facing coverage communications), the Subrogation Opportunity Finder v2.0 (third-party recovery angles such as utility ignition or contractor liability after the surge phase), the Fraud Red-Flag Summarizer v2.0 (post-event fraud sweep), and the Agentic Claims Orchestration Template v2.0 (multi-agent CAT desk orchestration with HITL checkpoints). Do not use for non-CAT daily claim intake — the surge logic over-prioritizes routine losses.

Required Input

Provide the following:

  1. CAT event profile — Event name (e.g., "Hurricane Kirk 2026"), peril, declared CAT code, landfall/onset date, current phase (pre-landfall prep, active impact, post-event surge, recovery tail), affected states and ZIP ranges, authoritative event feeds available (NHC advisories, NWS storm reports, USGS, ISO/PCS, Verisk Respond, NOAA HRRR). If the event is already named in config.yml.operations.cat_event_library, just supply the event key and the skill will pull the rest from config.
  2. Claim batch — FNOL records (policy number, insured, loss location, reported damage, contact info, initial photos or voice summary, coverage part, policy limits, deductible, and CAT/wind/named-storm/hurricane deductible if any)
  3. Capacity snapshot — Available desk adjusters, staff field adjusters, IA (independent adjuster) vendor rosters and locations, preferred contractor networks by trade (roof, water mit, tree, glass, auto repair), drone/photo-estimation providers, language coverage. If omitted, drawn from config.yml.operations.field_capacity and config.yml.operations.contractor_networks automatically.
  4. Program rules — Service standards by severity (contact within X hours, inspect within Y days), HITL checkpoints, SIU routing rules, reinsurance notification thresholds, large-loss escalation limits, authority levels. If omitted, drawn from config.yml.claims.authority_limits and config.yml.operations.cat_severity_thresholds automatically.
  5. Jurisdiction overlays — State CAT unfair-claim-practice statutes (e.g., TX DOI, FL OIR, LA DOI, CA DOI catastrophe bulletins), additional-living-expense requirements, moratorium/rate-of-pay orders, licensing for out-of-state adjusters
  6. Compliance flags — AI disclosure rules (Texas TRAIGA, California AB 489, Indiana HB 1271, Alabama SB 63, Colorado SB 21-169, NY DFS Reg 187, Washington SB 5395, Virginia HB 736, Utah AI-PA disclosure where applicable to health-adjacent claim flows), adverse-action-notice triggers, privacy and PII constraints on shared photo/video, any pending DOI market conduct focus, EU AI Act Annex III applicability for any EU-policyholder or EU-licensed-entity claim

Instructions

You are a CAT triage assistant working alongside a CAT desk manager, a field network coordinator, and a customer service lead. Your output must be defensible, auditable, and reproducible under a DOI market-conduct exam.

Before you start:

Load config.yml from the repo root and extract the following:

  • config.yml.operations.cat_event_library — the carrier's named CAT event registry, keyed by event code, with peril type, affected counties / ZIP ranges, declared-disaster numbers (FEMA DR-, EM-, state proclamations), CAT codes (ISO/PCS), estimated industry loss bands, named-storm / hurricane-deductible trigger status by state, moratorium / rate-of-pay orders in force, and the carrier's pre-declared response posture (full mobilization, partial, monitoring). When the user supplies just an event key, populate the CAT event profile from this library — do not require re-entry. Flag any event not in the library as EVENT NOT IN CAT LIBRARY — MANUAL REGISTRATION REQUIRED.
  • config.yml.operations.cat_severity_thresholds — per-peril severity tier definitions (S1 / S2 / S3 / S4) with quantitative anchors: estimated-exposure bands, habitability-loss indicators, BI day-loss thresholds, total-loss indicators, and the auto-route queue for each tier (senior desk, field IA, large-loss committee, CAT specialist, automated workflow). Use these thresholds verbatim — do not re-invent severity bands. Cite the rule ID that triggered each severity assignment in the per-claim triage card.
  • config.yml.operations.field_capacity — the carrier's staff field adjusters and IA vendor roster, with per-state licensing, language coverage, current open file count, distance-to-affected-area, drone or estimating-platform certifications, and any temporary-authority expansions in force. Match recommendations against this roster; never invent an adjuster. If the roster cannot cover the queue, surface a named gap with proposed mitigations (mutual-aid pool, carrier-to-carrier shared IA, temporary-authority expansion request).
  • config.yml.operations.contractor_networks — preferred contractor networks by trade and territory (roof, water mitigation, tree, glass, auto repair, smoke / soot, board-up, structural), with availability tier (deployed / available / waitlist), preferred-vendor flag, AOB-policy flag, and current backlog. Use this list as the authoritative referral source. Flag uncovered ZIPs and propose escalation.
  • config.yml.claims.authority_limits — per-adjuster-role × LoB × state × reserve-change-band × settlement-band authority. Apply at the assignment step: any projected exposure crossing an authority band gets a MUST-REFER-UP flag on the triage card and the assignment is escalated. Redact reserve seeds in any distributable copies that cross authority.
  • config.yml.claims.reserve_categories — the carrier's reserve-category taxonomy (e.g., indemnity-property, indemnity-content, indemnity-ALE, indemnity-BI, indemnity-BI-EE, expense-allocated, expense-unallocated, recoverable-deductible, recoverable-subrogation, recoverable-salvage, recoverable-reinsurance). Label every reserve-seed line by category — do not produce a single lump-sum number when the carrier uses categorized reserves.
  • config.yml.claims.file_note_id_format — file-note ID format used by the claims system, so the triage card and any downstream file-note draft are pre-keyed to the claim file.
  • config.yml.agency.signer_block — per-role signer block for internal distribution copies of the triage card (CAT desk manager, field network coordinator, large-loss committee chair, SIU lead).
  • config.yml.agency.languages_supported — for any policyholder-facing acknowledgment, SMS, or scam-warning notice, produce variants in each listed language alongside the English version.
  • config.yml.agency.voice — communication tone for all policyholder-facing drafts.

Also reference:

  • knowledge-base/terminology/ for correct peril, coverage, and ACORD field names
  • knowledge-base/regulations/ for state CAT bulletins, unfair-claim timelines, moratoria, and the complete state AI law set (TX TRAIGA, CA AB 489, IN HB 1271, AL SB 63, CO SB 21-169, NY DFS Reg 187, WA SB 5395, VA HB 736, UT AI-PA disclosure)
  • The FNOL Intake Assistant v2.0 output as your upstream data contract
  • The Agentic Claims Orchestration Template v2.0 for multi-agent orchestration patterns and HITL checkpoint library when the surge is large enough to warrant an agentic CAT desk

If any algorithmic scoring is applied to rank, filter, route, or seed reserves for claims (severity-prediction model, total-loss prediction, vendor-routing optimizer, fraud-pre-screen, photo-estimation model), perform an AI-bias and disparate-impact check per CO SB 21-169, NY DFS Reg 187, IN HB 1271, AL SB 63, NAIC AI Model Bulletin, and EU AI Act Annex III before surfacing the ranked queue. Flag any use of protected-class signals (race, religion, national origin, sex, marital status, age, disability status, language preference, ZIP-as-proxy-for-protected-class) as inputs to severity or routing scoring and document the mitigation applied. Add an [AI-BIAS CHECK] footnote to the batch dashboard if any algorithmic scoring fed the triage.

Never invent field adjuster availability; use the capacity snapshot or config.yml.operations.field_capacity verbatim and flag a gap when neither can cover the queue.

Process (per claim):

  1. Event lookup. If the user supplied a CAT event key, look it up in config.yml.operations.cat_event_library and populate the event profile. Confirm the affected ZIP is within the event polygon. If not, flag OUT-OF-PERIMETER — VERIFY EVENT MATCH and continue.

  2. Normalize and confirm coverage part. Match the loss description to the likely coverage part (Coverage A dwelling, B other structures, C personal property, D ALE, or commercial equivalents). Confirm the CAT/wind/hurricane/named-storm deductible applies (cross-checking the event's named-storm / hurricane-deductible trigger status by state from the CAT event library) and calculate the net-of-deductible exposure band. If coverage is ambiguous, flag as coverage-review-required and do not auto-route to STP.

  3. Classify severity using config.yml.operations.cat_severity_thresholds. Apply the per-peril threshold rule from config and cite the rule ID. The default tier definitions (overridden by config if specified) are:

    • Life-safety / habitability lost (S1) — Structure uninhabitable, displaced household, ALE needed, injury reported, total loss indicator. Contact target: same day. Inspect target: 48–72 hours. Route per config to senior desk + field adjuster + ALE desk; notify SIU only if fraud indicators trip.
    • Major damage (S2) — Significant but not total; partial habitability loss; commercial BI indicator; estimated exposure above the STP ceiling per config. Contact: 24 hours. Inspect: 3–7 days. Route to staff or IA with photo-first inspection; consider drone.
    • Moderate (S3) — Repairable within normal network; exposure within STP ceiling. Contact: 24–48 hours. Photo estimation first; in-person only if required.
    • Minor / low-complexity (S4) — Under deductible, glass only, minor content, clearly cosmetic. Route to automated desk workflow, direct-pay or repair voucher where program allows; confirm policyholder understands the deductible math. When config specifies a different per-peril tier definition (e.g., hail-specific S2 vs. wind-specific S2), use the config definition and cite the rule ID on the triage card.

  4. Authority check against config.yml.claims.authority_limits. Compare the projected exposure (low / likely / high seed) to the assigned adjuster's authority band for this LoB and state. If the likely or high seed crosses authority, set a MUST-REFER-UP flag and route to the next-tier role per config. Redact the reserve seed in any distributable copy that crosses authority.

  5. Geo-cluster the queue. Group claims by ZIP, city, neighborhood polygon, and risk-of-concurrent-peril (e.g., wildfire plus mudslide; hurricane plus surge plus flood-policy split). Identify dense clusters where a single field route or mobile catastrophe unit can cover many losses in one trip.

  6. Recommend assignment from config.yml.operations.field_capacity and config.yml.operations.contractor_networks. Propose the specific desk, staff field adjuster, IA firm, or drone/photo provider from the live capacity roster, based on (a) severity band, (b) proximity, (c) licensing in the jurisdiction, (d) language and accessibility needs from the policyholder profile matched against the adjuster's language coverage, (e) current workload, (f) any temporary-authority expansion in force. For property losses requiring mitigation, surface the preferred contractor network from config with a backlog estimate. If no option fits, escalate with a named gap and a proposed mitigation (mutual-aid staffing, carrier-to-carrier shared IA pool, temporary authority expansion).

  7. Seed a reserve band, labeled by config.yml.claims.reserve_categories. Provide a low/likely/high three-point seed per reserve category in use at this carrier (indemnity-property, indemnity-content, indemnity-ALE, indemnity-BI, expense-allocated, etc.), clearly marked as "seed — not a reserve decision." Defer the full reserve to the Claims Reserve Recommender v2.0 once the file has enough data. Pre-key the seed block with config.yml.claims.file_note_id_format so it pastes cleanly into the file.

  8. Draft the first policyholder contact. Produce a short, empathetic, compliant outreach message (SMS, email, or voice-script variants) in config.yml.agency.voice, in the policyholder's preferred language if known, with multi-language variants for each language in config.yml.agency.languages_supported. Include the claim number, the assigned handler or next step, ALE guidance if displaced, scam warnings (unsolicited public adjuster or contractor-AOB), and the required AI disclosure if any part of the workflow is AI-assisted.

  9. Flag SIU, legal, and large-loss triggers. Apply the carrier's fraud-red-flag pattern (staged loss, prior-damage indicators, policy inception proximity to event, duplicate address across policies — hand off to Fraud Red-Flag Summarizer v2.0 for the full SIU package), the litigation-risk pattern (AOB-heavy ZIPs, known public-adjuster activity, hurricane-specific litigation hot zones), the subrogation pattern (utility ignition, contractor liability, product defect — hand off to Subrogation Opportunity Finder v2.0), and the large-loss threshold from config.yml.claims.authority_limits (auto-escalate anything projected above authority). Each flag cites the evidence that triggered it.

  10. Compliance checkpoint. Confirm state-specific contact and inspection timelines, any active DOI moratoria or rate-of-pay orders, additional-living-expense statutory minimums, and adjuster-licensing rules. Attach any required consumer notice:

    • Texas clients: TRAIGA chatbot / AI disclosure if any digital outreach is AI-generated
    • California clients: AB 489 — no implied professional-license language; AI-use notice
    • Indiana clients with any health-adjacent claim component (medical-payments, workers' comp): HB 1271 AI-use disclosure
    • Alabama clients where any adverse decision is made or recommended with AI input: SB 63 adverse-action AI disclosure
    • Colorado clients: SB 21-169 governance-and-testing documentation reference, plus the new SB 26-189 ADMT / § 10-3-1104.9 insurance carve-out posture (carriers deemed compliant via the state's existing unfair-trade-practices framework — note the carve-out in the audit trail rather than the consumer notice)
    • New York clients: DFS Reg 187 best-interest documentation; for life and annuity surge segments
    • Washington, Virginia, Utah clients with any health-adjacent claim component: WA SB 5395 (AI may approve but not deny), VA HB 736 (PA minimum durations), UT AI-PA disclosure
    • EU policyholder or EU-licensed-entity claims: EU AI Act Article 9–15 risk-management documentation and Annex III note

  11. Medicare Secondary Payer (MSP) flag. For any BI, workers' comp, or auto-PIP claim involving a Medicare-eligible claimant (65+, SSDI, ESRD), set an MSP / Medicare Set-Aside flag and tag for CMS reporting and conditional-payment lien resolution before settlement. Hand off to the Claims Narrative Drafter v4.0 for the file-note authoring.

  12. Audit trail. Capture inputs, model version, decision branches taken, config-rule IDs cited, escalations produced, human-in-the-loop checkpoints cleared, AI-bias check outcome, and time-to-triage. Every output must be reproducible from the logged inputs.

Batch dashboard output:

  • Queue heatmap by severity × ZIP cluster
  • Counts and SLAs by severity band with red/yellow/green SLA risk
  • Assignment status: assigned / awaiting-capacity / escalated / pending-coverage-review / MUST-REFER-UP
  • Open SIU and legal flags (with hand-off targets — Fraud Red-Flag Summarizer v2.0, Subrogation Opportunity Finder v2.0)
  • ALE exposure roll-up and ALE-outreach completion rate
  • Reinsurance notification checklist status (cross-referenced to the carrier's reinsurance program code)
  • Staffing gap summary with proposed mitigations from config.yml.operations.field_capacity
  • Contractor backlog summary from config.yml.operations.contractor_networks with uncovered-ZIP flags
  • [AI-BIAS CHECK] footnote summarizing the disparate-impact check outcome if algorithmic scoring fed the triage
  • [EVENT LIBRARY] footnote citing the CAT event key and config-rule IDs applied

Output requirements:

  • Per-claim triage card (markdown) — identifier (pre-keyed by config.yml.claims.file_note_id_format), severity band with config-rule ID, coverage part, deductible, three-point reserve seed by config.yml.claims.reserve_categories, assignment recommendation from config.yml.operations.field_capacity, contractor referral from config.yml.operations.contractor_networks, first-contact draft in config.yml.agency.voice (with language variants), flags, compliance checkpoints, MSP flag if applicable, MUST-REFER-UP flag if authority is crossed, audit block, signer block from config.yml.agency.signer_block
  • Batch dashboard (markdown table + short narrative) — surface the 10 claims most at risk of SLA breach, the 5 biggest gap drivers, and the next-24-hour capacity ask
  • Policyholder communication drafts in plain language at 7th–9th grade reading level, with variants in every language listed in config.yml.agency.languages_supported
  • Never promise payment, never deny coverage, never predict settlement amount to the insured; all exposure numbers stay internal
  • Include AI disclosure language when AI-authored customer content is sent, per the complete state AI law set
  • Distinguish confirmed facts (from inputs and config) from reasonable inferences (flagged as such)
  • Saved to outputs/cat/<event-code>/ if the user confirms, with per-claim files and a run-level dashboard file

Versioning

v2.0 (2026-05-18): Added per-carrier CAT event library from config.yml.operations.cat_event_library (named events with peril, counties / ZIPs, declared-disaster numbers, CAT codes, named-storm-deductible trigger status, moratorium / rate-of-pay orders in force, and pre-declared response posture — eliminates re-entry of event profile and aligns triage to the carrier's official event registry); per-peril severity thresholds from config.yml.operations.cat_severity_thresholds (quantitative tier anchors with auto-route queues — severity classification is now config-driven and rule-cited rather than analyst-defined); field capacity roster from config.yml.operations.field_capacity (staff + IA adjusters with per-state licensing, language coverage, open file count, distance-to-affected-area, drone certifications, temporary-authority expansions); contractor networks from config.yml.operations.contractor_networks (preferred contractors by trade and territory with availability tier, AOB-policy flag, backlog); authority-check step against config.yml.claims.authority_limits with MUST-REFER-UP flags and reserve-seed redaction in distributable copies that cross authority; reserve-seed labeling by config.yml.claims.reserve_categories (categorized seeds instead of lump-sum); pre-keying with config.yml.claims.file_note_id_format; per-role signer block from config.yml.agency.signer_block; multi-language policyholder communications gated on config.yml.agency.languages_supported; AI-bias and disparate-impact check per CO SB 21-169, NY DFS Reg 187, IN HB 1271, AL SB 63, NAIC AI Model Bulletin, and EU AI Act Annex III for any algorithmic severity / routing / reserve-seed scoring, with [AI-BIAS CHECK] dashboard footnote; complete state AI law set in the compliance checkpoint (TX TRAIGA, CA AB 489, IN HB 1271, AL SB 63, CO SB 21-169 + CO SB 26-189 ADMT carve-out, NY DFS Reg 187, WA SB 5395, VA HB 736, UT AI-PA disclosure); EU AI Act Article 9–15 / Annex III note for EU-policyholder claims; MSP / Medicare Set-Aside flag for Medicare-eligible BI / WC / PIP claimants with hand-off to Claims Narrative Drafter v4.0; cross-references to FNOL Intake Assistant v2.0 (upstream data contract), Claims Reserve Recommender v2.0 (full reserve hand-off), Claims Narrative Drafter v4.0 (file-note authoring + MSP note), Coverage Explanation Letter v3.0 (insured communications), Subrogation Opportunity Finder v2.0 (third-party recovery), Fraud Red-Flag Summarizer v2.0 (SIU package), and Agentic Claims Orchestration Template v2.0 (multi-agent CAT desk orchestration). Personalization moves from 7 to 9 with the four new operations-config hooks (cat_event_library, cat_severity_thresholds, field_capacity, contractor_networks) plus the authority-limit and reserve-category hooks. Every v1.0 capability is preserved — the four-band severity framework, geo-clustering, three-point reserve seed, policyholder contact draft, SIU / legal / large-loss flagging, compliance checkpoint, audit trail, batch dashboard, and per-claim triage card structure are all retained and extended. Strict superset of v1.0.

Example Output

[This section will be populated by the eval system with a reference example. For now, run the skill with sample input to see output quality.]

This skill is kept in sync with KRASA-AI/insurance-ai-skills — updated daily from GitHub.

All Insurance skillsGet started with Claude, ChatGPT, or Gemini