VorticVortic
◑ Solutions · P&C Carrier Underwriting

Carrier-grade triage
velocity. Without the replatform.

Vortic is the AI workforce for P&C carrier underwriters writing direct authority: submission triage, portfolio concentration surfaced pre-bind, CAT response in minutes, and an audit-grade memo for every state DOI review — all inside SSO + VPC + BYOLM your security team will sign off on. No deck. No pilot. Working software on the systems your UW team already uses.

  • SSO · SAML / OIDC
  • VPC / on-prem deploy
  • Bring-your-own LLM
  • SOC 2 · NAIC · state DOI
  • PII redacted at trace
BROKER · SUBMISSIONPOLICYHOLDER · FNOLAGENT · ENDORSEMENT REQREGULATOR · QUERYVORTICagentic rosterSTRUCTURED CASE FILECLAIM FILE · TRIAGEDENDORSEMENT DRAFTBORDEREAU · FILEDONE AGENT FABRIC · ONE AUDIT CHAIN
◑ Carrier UW workflows

Four surfaces a P&C carrier UW lives in every day.

Built around the carrier UW decision sequence: triage the submission, surface portfolio impact, respond to CAT events in minutes, and hand the regulator an audit-grade memo. Each surface is a live product, not a roadmap promise.

Submission to quote

Submission-to-quote in under 30 seconds.

Multi-agent triage runs across the whole submission in parallel — parse, risk, flood, pricing, compliance, treaty, portfolio. Every cited finding lands in a memo your underwriter signs off on, never an LLM-only auto-bind.

  • Parse + cite every field from PDFs, ACORD, broker email
  • Specialist agents run in parallel, not in chain
  • Audit-grade memo synthesised before the UW opens the case
See the agent fabric →
Portfolio concentration

Concentration + CAT cap, surfaced before bind.

Every submission carries a marginal-impact widget: "If bound, Manhattan 100 goes 35.8% → 38.4%. Cap 25% — already in breach." Your UW sees the portfolio consequence before clicking bind, not at end-of-quarter when treaty calls.

  • Live ZIP3 / postcode aggregation of the bound book
  • Capital + PML estimates (OEP 1-in-100, 1-in-250, AEP 1-in-10)
  • Rebalance proposal: declines + headroom + treaty cession
Open Concentration →
CAT live response

Re-score the entire book against a declared event. 18 minutes.

A hurricane is declared. Vortic streams per-policy scoring (wind damage curve × distance × surge) across every active bound risk in parallel. Worst-loss policies, treaty cession draft, and cedent / insured letters land before your CAT calls go out.

  • Storm-track + live NOAA / NHC / USGS feed
  • Per-policy SSE: gross loss · ceded · net retained · top exposures
  • Treaty cession draft + cedent letters — drafted, never auto-sent
Open CAT mode →
Audit-grade memo

NAIC + state DOI + internal review trail. One artifact.

Every agent call is logged with prompt, model, output, and trace id. State DOI rate-filing review, NAIC market-conduct, internal UW audit — all answered from the same memo + trace log. No more "let me pull that file" delays.

  • Every LLM call cited with prompt + model + trace id
  • Decision audit: rationale, conditions, override history
  • Exportable for NAIC market-conduct + state DOI rate-filing review
See the audit trail →
Dynamic sub-agents

Reflexive mini-agents the pipeline spawns when a specialist needs a second opinion.

Define focused sub-agents the pipeline spawns dynamically: verify a postcode against FEMA NFIP, corroborate loss history against ISO ClaimSearch, look up a clause cited in a coverage memo, summarise concentration impact. Five customisable categories — verification, research, synthesis, data fetch, custom. A pure-function reflexive suggester matches trigger patterns against parent agent JSON and either auto-spawns or surfaces a click-to-spawn chip in the run UI. Every spawn writes a sub_agent_invocations row with parent/child trace lineage.

  • 5 categories · 5 trigger kinds (equals/contains/regex/array_includes/truthy)
  • 7 seed templates ship on first visit — fully editable
  • Auto-spawn or click-to-spawn, both audited the same way
See the sub-agent builder →
Straight-Through Processing

Auto-bind the 60–80% of submissions that don't need a human.

Every submission that finishes the pipeline with all greens, within concentration cap, within pricing band, and outside hard exclusions is auto-bound — bound_risks row created, audit trail stamped, no human touch. Configurable TSI cap, required-green agents, excluded occupancies / states / perils, pricing rate band, post-bind concentration limit. Three terminal outcomes: auto_bind / auto_decline / refer. Every decision carries the full reason list.

  • Configurable per customer — no code
  • Hard exclusions auto-decline; everything in-between refers
  • Every STP decision lands on the submission with reasons
See the STP rules →
◑ Claims & adjudication

One platform for bind and claims. Same audit trail. Same agent fabric.

Top-level workbench for adjusters and supervisors. FNOL parser, Coverage Analyst, Claim Memo writer — three agents that move a claim from broker email to adjudication memo in minutes. Reserve revision history, payment authority gating, statutory letters with state-specific UCSPA language, supervisor queue, and book insights.

FNOL intake

Intake from email, voice, portal — auto-acknowledged in minutes.

Broker email, claimant voice transcript, portal form, or mobile photo upload — all into one structured FNOL. The acknowledgment letter dispatches 5 minutes after intake with state-specific UCSPA language; cancellable from the claim detail page during the undo window.

  • Multi-channel FNOL parser
  • State-specific acknowledgment (CA / FL / NY / TX + generic)
  • 5-minute undo window before letter sends
Coverage analysis

Joined against the policy — covered, partial, not covered. Every clause cited.

Coverage Analyst agent joins the FNOL against the linked bound risk and returns a decision with cited exclusions, deductible, limit remaining, and fraud flags. Refers rather than denies under uncertainty — denial without basis is bad-faith exposure under most state UCSPAs.

  • Linked to bound_risks policy summary
  • Fraud flag detection (rapid-post-bind, round losses)
  • Conservative: refers under uncertainty
Adjudication + authority

Pay / deny / partial / refer. Per-adjuster authority gating, full audit trail.

Live authority preview as the adjuster types the pay amount: $50k for adjuster / $250k for senior / $1M for supervisor. Over-authority disables the pay button and routes through refer-to-senior. Every decision lands on a claim_decisions row with snapshot of the agent outputs at decision time.

  • Per-adjuster bind-equivalent limits
  • Refer-to-senior with proposed pay attached
  • Audit trail with agent_outputs snapshot per decision
Fraud / SIU lane

Rule engine + LLM + ISO ClaimSearch + NICB Forewarn. One fraud score.

Six deterministic rules (rapid-post-bind tiers, round-loss, prior-similar by frequency, no-police-report by cause, high-value-low-doc, inception-to-loss-mismatch) plus an LLM narrative agent for witness contradictions and suspicious phrasing. Wired to real ISO ClaimSearch + NICB Forewarn API clients with mock fallback for preview. Cross-claim pattern matcher catches repeated claimants, attorneys, and repair-shop clusters across your own book.

  • Aggregate fraud score 0–100 with severity-weighted math
  • Refer-to-SIU workflow with status + audit row
  • Subrogation lead detector + demand letter draft generator
Statutory letters

Acknowledgment, RoR, denial, payment. UCSPA-compliant. State-aware.

Four high-volume states have cited statutory language baked in: Cal. Code Reg. § 2695.5(b) + § 2695.7(j), Fla. Stat. § 626.9541, 11 NYCRR 216.4(a), Tex. Ins. Code § 542.055 — each with the right DOI complaint phone number. RoR letters include the non-waiver language. Denials include appeal rights. Every letter is reviewable before send.

  • CA / FL / NY / TX statute-aware language
  • Acknowledgment auto-schedules, others land as draft
  • Reviewable + cancellable before dispatch
◑ Across the rest of the carrier

Submission intake. Claims FNOL. Policy comparison. Regulatory reporting.

The same agent fabric your UWs use already extends to claims ops, policy admin, and compliance — so an enterprise rollout doesn't fork into four separate vendors.

emailPDFformPARSECASE FILEINSUREDPOSTCODETIVOCCUPANCYLOSS HX
Submission intake

Convert mixed-format submissions into clean case files.

Broker emails, PDFs, ACORD forms, web-form intake. The pipeline extracts every required field, validates against your appetite rules, flags missing info with citations, and routes the case to the right underwriting team.

  • Handles email + PDF + form + portal feeds
  • Cited extractions; human can verify any field
  • Routes via your appetite rules, not generic ones
VOICE FNOLTRIAGEFNOLCLAIM FILENON-CLAIM
Claims FNOL & document intelligence

Triage first notice of loss and structure every claim document.

FNOL voice and email intake. Distinguish claims from non-claims. Extract coverage facts from estimates, police reports, medical records, and photos. Build a structured claim file the moment the call ends.

  • Auto-routes non-claims (roadside, billing) elsewhere
  • Parses estimates, police reports, medicals, photos
  • Surfaces fraud indicators + coverage-impact early
EXPIRINGRENEWAL+8.5%
Policy comparison + bind verification

Catch coverage drift before bind. Compare renewals in seconds.

Compare expiring vs renewal across coverage, limits, deductibles, exclusions, endorsements, premium, period, and conditions. Reconcile bind orders against quoted terms — never bind a coverage that wasn't agreed.

  • Side-by-side dimension diff with severity
  • Premium delta callout with currency
  • Bind verifier prevents E&O exposure
BOUND BOOKBORDEREAUGENERATORBORDEREAU FILE
Bordereaux + regulatory reporting

Produce bordereaux and filings without manual reconciliation.

Generate US program-business filings, NAIC and state DOI queries, plus Lloyd's coverholder bordereaux directly from the bound book. Validate every line. Surface blocking issues before submission.

  • Standardised bordereau format per program
  • Validation flags: blocking / warning / info
  • Auditable source row → bordereau line
◑ Outcomes

Three numbers carrier UWs keep score on.

30%
UW capacity unlock

Existing underwriting headcount writes more book — submission triage + cited memo lift the marginal capacity per FTE.

$14M
CAT-cap save

Concentration breach surfaced before bind on the breaching ZIP3. Real number, real customer, fewer mid-quarter treaty calls.

18 min
CAT response

A hurricane is declared. The whole book is re-scored, treaty cession drafted, cedent letters templated — before the CAT call goes out.

◑ Personas

UW first. The same agent fabric, four functions inside the carrier.

Underwriting

Direct-authority bind decisions, submission triage, appetite gating, portfolio concentration pre-bind, CAT response, treaty cession, audit-grade memo.

Claims

FNOL voice + email triage, document intelligence, fraud flagging, coverage decisioning.

Policy administration

Endorsement drafting, plain-language coverage summaries, bind verification, renewal automation.

Regulatory & compliance

NAIC + state DOI rate-filing review pack, bordereaux generation, sanctions screening, KYC + UBO verification, auditable trace logs.

◑ How it works

Five steps. Same shape every time.

  1. 01

    Submission / claim / endorsement / regulator query lands in your environment.

  2. 02

    Vortic's parse agent extracts structured fields with cited confidence.

  3. 03

    Specialist agents run in parallel — risk, flood, treaty, sanctions, compliance, claims-intel.

  4. 04

    Synthesis agent writes a memo / claim file / endorsement / bordereau line.

  5. 05

    Your underwriter, adjuster, or compliance reviewer confirms — every action is on the audit chain.

SSO · SAML / OIDC + SCIM
VPC · on-prem deploy option
Bring-your-own LLM (Anthropic · Bedrock · Vertex)
SOC 2 prep · NAIC + state DOI audit pack

See your carrier book run on Vortic.

Send us a sanitised submission, claim, or renewal pack. We'll run it through Vortic and walk you through what the agents found — typically a 30-minute call with a solutions engineer who's worked the carrier UW desk.

Talk to a solutions engineer Book a demo →

founder@getvortic.com · we reply within one business day