Seven systems. One engine. Built for insurance.

Viola runs as seven systems on one engine. They are Multi-Agent Intelligence, the Rating Engine, Document Intelligence, Payments and Billing, Real-Time Operations, Viola Concierge, and Security and Governance. Together they cover 30 named capabilities. Each system works on its own. The platform is the way they work as one. Multi-Agent Intelligence runs specialised agents for underwriting, claims, documents, and research in parallel. The Rating Engine treats rates as governed, versioned assets. Document Intelligence ingests and extracts from anything that arrives. Payments and Billing collects premium the moment a policy binds. Real-Time Operations keeps every workspace live. The Concierge is the conversational front door for customers and admins. Security and Governance applies enterprise SSO, role-based access, and audit logging across all of it.

Rates are governed assets, not settings. Every rate version is effective-dated. It moves through draft, review, and approve before it can price a policy. That means you can always show which version priced which policy. Actuaries import rate tables in bulk from the spreadsheets they already use. Base rates, factors, and coverage rules are set per state. Rating runs on the server at every quote, save, and bind. Server-side rating matters because it means the premium cannot be altered client-side between quote and bind; the figure is recomputed against the governed version each time. A version stays frozen once approved, carrying the base rates, factors, and rules for one product in one state. Months later the approval trail stands up alongside the rates themselves.

Viola reads PDF, DOCX, images, spreadsheets, and email. It runs OCR and structured extraction on police reports and claim documents. Each document is triaged on arrival for severity, coverage match, and completeness. It also generates policy documents, acknowledgements, and filings. No one re-keys a field. Structured extraction is the part that matters operationally: rather than storing a scan and leaving a person to read it, the system pulls the parties, dates, and damages out as fields and validates them against the file already on record. A mismatch surfaces as a discrepancy instead of propagating silently. Generation runs the same way in reverse, producing ID cards and state filings, including SR-22, from the policy record rather than from a separate document system.

Viola takes credit card and ACH payments inside the bind and renewal flows. Payment methods are kept in a secure vault for recurring billing. Customers can pay in full, in installments, or by auto-pay. Voids, refunds, and transaction lookup all happen in one console. Every transaction keeps a full audit trail, down to the raw gateway response. Collecting inside the bind flow rather than after it removes the reconciliation step: the payment is attached to the policy it paid for, so there is no batch to match back. Renewals run through the same billing setup as the original bind, including installments and auto-pay, so a continuing policyholder is not asked to re-enter details Viola already holds in the vault.

Access runs on enterprise SSO with token-based authentication. Roles are fine-grained and apply across modules. Privileged actions are logged in full. The architecture is cloud-native SaaS, aligned to industry security standards. The AI agents are gated too. People approve the decisions that carry risk, and every one of those is auditable. Fine-grained and cross-module means a role is defined once and holds everywhere, rather than a person accumulating different permissions in underwriting, claims, and billing. Comprehensive audit logging of privileged actions is what makes the approval trail usable after the fact: it records who approved a rate version or moved a reserve, not merely that the record changed. The same gates apply to agent actions as to human ones.

Viola is multi-line by design. Personal auto is live today. Commercial lines are next. Products, rates, and rules are configured rather than coded. A new product or state is therefore a configuration change routed through approval, not a vendor project. Multi-line by design means the data model does not assume one product shape, so adding a line is a configuration exercise rather than a schema migration. The practical test is whether a change needs a vendor ticket and a release window. In Viola it does not: a new product, a new state, or a mid-term rule change is authored in the platform, routed through draft, review, and approve, and scheduled with an effective date. The compliance guardrails apply throughout.

The system that keeps every screen current. Quotes, claims, assignments, and approvals stream to customer, agent, and admin workspaces as they happen. There is no refresh, no stale queue, and no waiting for an overnight job. Assignments, approvals, and status changes raise instant notifications. Operational dashboards show live performance, and the platform surfaces proactive recommendations on cost and growth. The reason this is a named system rather than a UI detail is that batch-driven operations shape how a team works: queues get worked on yesterday’s picture, and two people act on the same file because neither saw the other take it. Streaming state to all three workspaces at once removes that class of problem rather than managing around it.

Your own policy and claims knowledge. The agents retrieve against it by meaning rather than by exact keyword, so a question finds the right prior file even when the wording differs. It feeds underwriting, claims, document, and research agents. Every transaction also feeds back in, so retrieval keeps improving as the book grows. The distinction from keyword search is practical rather than technical. A keyword index finds a prior claim only if someone happened to describe it in the same words; retrieval by meaning finds it when the wording differs but the situation matches. That matters most on the files where precedent is worth the most: an unusual loss, a coverage question that has come up before, a risk that resembles one already priced.

A support chatbot answers questions about the product. The Concierge does the work. A policyholder can quote, buy, and service a policy end to end in conversation. An admin can ask a portfolio question in plain English and get real figures back, not a link to a dashboard. It is role-aware and account-scoped, and it asks for confirmation before it acts. Account-scoped means it can only see and act on records the person asking is entitled to, so the same assistant is safe to expose to a policyholder and to an operations lead. Confirmation-gating means it never commits a change on an inferred intent. It also includes a built-in guide that knows every screen and walks users through tasks.