Question 1

What does Vortic actually do for an underwriter?

Accepted Answer

Vortic takes a broker submission — PDF, email, spreadsheet — and returns a fully cited underwriting memo in under 30 seconds. Insured details extracted and reconciled, postcode pulled to flood zone, loss history surfaced, TIV sanity-checked across line items, pricing benchmarked, treaty utilisation projected, compliance red-flags raised. The underwriter reads the memo and clicks bind, decline, refer, or query. Judgment stays with the human. The lookups go to the system.

Question 2

What underwriting lines does Vortic support?

Accepted Answer

Today: commercial property and casualty across the US and UK, including US E&S coastal, retail commercial, program business, and Lloyd's coverholder books. Specialty lines (cargo, professional liability, construction, cyber) are supported with custom agent prompts during onboarding. We do not currently target retail personal lines.

Question 3

How long does onboarding take?

Accepted Answer

Two weeks from contract to first live submission for a typical MGA. Week one: connect data sources (broker email inbox, existing book of bound risks, treaty programme). Week two: tune the risk agent against your appetite, adjust the memo template, set role permissions, train the team on Command Center.

Question 4

How does Vortic handle data we cannot send to third-party LLMs?

Accepted Answer

Three layers. PII redaction before any model sees the submission — insured names, broker emails, postcodes are masked in the audit log by default. You choose which ground-truth lookups (flood, postcode, FEMA NFIP) join the prompt per-agent. Enterprise customers can route the platform to their own private LLM deployment so no inference traffic leaves the VPC.

Question 5

How do you ensure the AI does not hallucinate flood zones, loss history, or pricing?

Accepted Answer

Ground-truth blocks. For factual claims — flood zone, ZIP/postcode, sanctions, regulatory thresholds — Vortic injects verified data from the source (FEMA NFIP, NOAA HURDAT2, OFAC SDN, postcodes.io, your loss-history database) directly into the prompt and instructs the model to cite or refuse. The memo carries explicit citations back to the source. If a lookup fails, the memo says "no data" instead of guessing.

Question 6

What audit trail does the platform keep for compliance?

Accepted Answer

Every model call is logged with the prompt version, model used, tokens consumed, latency, and structured output. Every memo cites its sources. Every decision (bind / decline / refer / query) is captured immutably with the underwriter, the rationale, and the agent outputs at that moment. Designed against NAIC model governance bulletins and Lloyd's coverholder oversight from day one. The full audit table is exportable for state DOI or PRA submission.

Question 7

How does Vortic integrate with our existing systems?

Accepted Answer

Inbound: a webhook endpoint where brokers (or your own broker portal) can POST submissions directly into your queue, authenticated with a bearer key. Outbound: HMAC-signed event webhooks fire on every bind / decline / refer to your policy admin system, treaty platform, or finance ledger. CSV import is available for bringing existing bound books into Vortic for portfolio analysis.

Question 8

How does the team and permissions model work?

Accepted Answer

Four roles per organisation: owner, admin, member, viewer. Owners manage billing and ownership; admins manage members and integrations; members run pipelines and bind; viewers see decisions but cannot execute. Email-invite flow with a 14-day token. A user can belong to multiple organisations.

Question 9

How do credits and pricing work?

Accepted Answer

Pay-as-you-go. $5 free on signup. $0.10 per credit, $20 minimum top-up, credits never expire. A full underwriting pipeline (parse + eight specialists + memo) costs around 18 credits ≈ $1.80 per submission. Enterprise pricing available with SSO, dedicated CSM, and routing to private LLM deployments.

Question 10

What happens if a regulator asks why a specific risk was bound?

Accepted Answer

Open the submission history page. It shows every agent that ran, the inputs they received, the outputs they produced, the prompt version active at that moment, and the underwriter's final decision with timestamp and rationale. The full chain reconstructs in seconds. Built for the moment of regulatory scrutiny, not against it.

Question 11

How does Vortic compare to ChatGPT or general AI tools?

Accepted Answer

A general AI tool can read a PDF and summarise it. Vortic runs specialist agents in coordinated parallel, grounds each one against verified data sources (EA, FEMA, postcodes.io, your loss-history), produces an underwriter memo with structured citations, sits behind a human bind gate, and logs every call for the regulator. Different problem, different posture.

Question 12

Does Vortic replace underwriters?

Accepted Answer

No. Vortic automates extraction, ground-truth lookups, specialist analysis, and memo drafting. Underwriters retain judgment on bind, decline, refer, conditions, and broker conversations. Agents propose; humans dispose; the audit trail captures everything.

Question 13

How fast can Vortic analyze a broker submission?

Accepted Answer

A typical full agentic pipeline run—including PDF parsing, parallel specialists (risk, flood, pricing, compliance, treaty, portfolio), and memo synthesis—completes in roughly 20–30 seconds end-to-end with streaming updates. Actual latency varies by document complexity and model routing.

Question 14

Who should request access to Vortic?

Accepted Answer

Teams that bind commercial risks on delegated authority—MGAs, Lloyd's coverholders, binding-authority underwriters, wholesale brokers needing faster quote-back, carriers overseeing delegated partners, and reinsurers monitoring treaty utilisation. Early access is invite-only; use the signup form to describe your book and workflows.

Question 15

Why use agentic underwriting instead of a single AI chatbot?

Accepted Answer

Specialist agents each have a narrow contract, dedicated prompts, and traceable outputs; they can run in parallel before synthesis into one memo. That yields faster bind-ready decisions, clearer citations to external data (flood, firmographics, market benchmarks), and an audit trail regulators expect—versus one opaque assistant thread.

Guides, research & case studies

Why parallel agents beat sequential workflows in commercial underwriting

How a US E&S MGA cut submission-to-bind from 4 days to 35 seconds

2026 Commercial Property AI Underwriting Benchmark Report