v2.0

GoRules Version 2 is here - redesigned, now with managed cloud.GoRules Version 2 is here!

Watch the launch videoWatch
gorules
Book a demoGet started

An automated underwriting system
your underwriters own.

Appetite, knockouts, referral triggers, and authority limits live in visual decision tables your underwriting team edits, not in the PAS release backlog. Every submission comes back bind, refer, or decline in milliseconds, with the exact rules that fired.

From broker inbox to bound quote,
start to finish.

Watch a morning on a commercial property desk: a clean sprinklered submission stamped REFER because it sits $400K past the appetite ceiling, a one-cell change, a replay over 1,140 referred submissions, a signed release - and a straight-through rate that goes from 54 to 71 percent.

scroll to play
Monday, 08:52

A clean risk lands in the inbox.

A broker emails SUB-4451 with the ACORDs attached: a sprinklered bakery, $1.4M total insured value, protection class 3, five years without a loss. It goes into the referral queue behind 46 others.

The submission

Nothing wrong with it. Still referred.

Joisted masonry, full NFPA 13 sprinklers, hydrant 260 feet out, zero losses in five years. It sits outside the appetite band for one reason: the auto-quote ceiling for protected risks stops at $1.0M.

One cell

The chief underwriter moves the ceiling.

The rule that referred it opens the live appetite table. For protection class 5 and better, the auto-quote ceiling goes from $1.0M to $2.0M. On a branch, nothing else touched.

Proof before production

1,140 referrals re-run.

Ninety days of referred submissions replay against the draft while you watch: 380 come back as auto-quotes, projected loss ratio holds at 48 percent, coastal wind exclusions never move.

Sign-off

Underwriting approves. v6.4.0 ships.

Jane Cooper signs the release, it promotes through staging to production with the full trail recorded. Rollback stays one click away.

Same morning

The band moves. The queue drains.

The green zone runs past $1.4M, SUB-4451 stamps QUOTED at $18,400 and goes back to the broker at 08:53. Behind it: 47 referrals down to 12, straight-through rate 54 to 71 percent. The grain elevator at protection class 9 stays exactly where it is.

Your appetite is a PDF.
Your engine disagrees.

Ask a carrier or MGA what its risk appetite is and you get a PDF. Ask the systems and you get three different answers: eligibility logic hard-coded in the policy administration system, a spreadsheet of class limits the product team maintains, and a set of referral habits that live in senior underwriters' heads. An automated underwriting system is supposed to make those three the same thing. In most carriers they drift apart quarter by quarter, and the underwriting queue absorbs the difference.

That drift has a number attached: straight-through processing rate. Personal auto and home bind most clean risks without a human. Mid-market commercial rarely comes close, because a single risk carries far more variables - class code, total insured value, protection class, loss history, schedule credits - and every rule that handles them is another ticket in the PAS backlog. Submission-triage vendors commonly report that roughly a quarter of incoming submissions fall outside the carrier's own stated appetite and are still screened by hand. The risks that could have bound in seconds wait behind them.

Then come the questions a spreadsheet cannot answer. Which version of the eligibility rules bound this policy last March? Who raised the TIV ceiling, who approved it, and what did it do to the referral mix? Supervisors now ask sharper versions of the same thing. The NAIC Model Bulletin on the Use of Artificial Intelligence Systems by Insurers, adopted in December 2023 and issued in roughly half the states, expects a documented governance program covering underwriting and rating. NYDFS Circular Letter No. 7 (2024) expects New York insurers to test and document how automated underwriting and external consumer data affect protected classes. In the EU, AI systems used for risk assessment and pricing in life and health insurance sit in Annex III of the AI Act as high-risk.

If your underwriting rules are a vendor config export and a folder of change-request emails, every one of those conversations is harder than it needs to be.

Appetite in tables.
Referrals with reasons.

GoRules is the decision layer inside your underwriting flow: submission in, bind, refer, or decline out, with the rule that produced it named.

01

Knockouts, appetite, and authority as one decision graph

Underwriting is a sequence, not a single lookup, so model it as one. Hard knockouts run first as Policies, ordered blocks that stop a submission on a prohibited class code, a coverage you do not write, or a state you are not admitted in. The appetite matrix follows as a decision table: class code by total insured value by protection class on commercial property, or face amount by age band by build and nicotine status on life. Referral rules then compare the risk against the binding authority of the underwriter who owns it. ZEN expressions handle derived values like loss ratio, TIV per square foot, or years since the last large loss, and typed TypeScript function nodes cover anything custom. One structured decision comes back: bind, refer with a reason code and a routing target, or decline with grounds.

02

Underwriting owns the rules, with a release trail

Product and underwriting edit decision tables directly, and natural-language rendering makes each rule read like the appetite guide it replaces. Git-like version control records who changed which threshold and when, approval workflows keep risk and compliance in the loop, and any release rolls back in one click. Dev, staging, and production stay separate environments, so a new wildfire referral threshold is proven against last quarter's submissions before a live quote ever sees it.

03

Scores and third-party data stay inputs; the decision stays readable

Hazard scores, MVR, MIB and prescription results, loss runs, telematics, and whatever your intake tooling extracts from ACORD forms and statements of values all arrive as plain inputs to the rules. The model estimates risk; the rule decides what to do about it. That separation is what lets you show an examiner a readable rule and a version number instead of a model output nobody can explain, and it gives the human oversight regulators expect something specific to review.

Keep the PAS.
Move the judgment.

GoRules does not replace your policy administration system, your rating engine, or your underwriting workbench. It owns the decision those systems keep asking for.

01

Connect

Your PAS, underwriting workbench, broker portal, or submission-intake service calls GoRules over REST, or embeds the MIT-licensed open-source ZEN Engine in-process with SDKs for Node.js, Python, Java, Go, C#, and Rust. Carrier stacks are usually already on the JVM, so in-process evaluation is a normal option rather than a rewrite.

02

Model

Rebuild the appetite guide as decision tables and graphs: knockouts, the appetite matrix, referral triggers, and authority limits each become a block you can point at in a review. Start from the Insurance Underwriting Risk template and the Policy Eligibility Analyzer rather than a blank canvas.

03

Simulate

Replay last quarter's submissions through a branch before anything ships, and read the result in the numbers underwriting is measured on: straight-through rate, referral volume by trigger, and which risks change decision. Stakeholders review the diff in natural language, and promotion through environments records an approval at each step.

04

Operate

Every evaluation returns the outcome plus the rules that fired, so a referral reaches the workbench with its reason attached and a decline carries its grounds. Watch straight-through rate, referral mix, and quote-to-bind weekly, and tune the tables instead of filing tickets. Roll back in one click if a change misbehaves.

Built for production,
not proof of concept.

  • Sub-millisecond evaluation with the open-source ZEN Engine (Rust core), fast enough to run appetite and eligibility inline in a quote-and-bind flow
  • Git-like version control on every underwriting ruleset: branches, commit history, approval workflows, one-click rollback
  • Self-host with Docker or Kubernetes so applicant medical data, loss runs, and PII stay inside your infrastructure, or run on GoRules Cloud
  • An audit trail an exam can follow: which ruleset version decided a submission, who changed the threshold, who approved it. SSO, role-based access, SOC 2
  • No lock-in: the engine is open source under MIT with a Rust core, and 70+ templates, including Insurance Underwriting Risk and Policy Eligibility Analyzer, give you a working starting point

Questions, answered.

What is an automated underwriting system?

An automated underwriting system evaluates an insurance application or submission against the carrier's underwriting rules and returns a decision - bind, refer to an underwriter, or decline - without a human touching the clean cases. It applies knockouts, the risk appetite matrix, referral triggers, and authority limits to data from the application and from third-party sources, in seconds instead of days. The term is also used in mortgage lending for systems like DU and LPA; in insurance it means the rules layer between submission intake and rating or binding. GoRules provides that layer as visual decision tables and graphs your underwriting team edits directly.

What is the difference between an automated underwriting system and a policy administration system?

A policy administration system owns the policy lifecycle: quoting, issuance, endorsements, renewals, and billing. An underwriting workbench owns the human's workspace: the queue, the documents, the notes. An automated underwriting system owns the decision itself - whether the risk is in appetite, whether it can bind straight through, and who it refers to when it cannot. GoRules replaces only that decision logic, which today usually sits hard-coded inside the PAS or in vendor configuration nobody outside the vendor can edit, and your existing systems call it over REST wherever a decision is needed.

How do we raise straight-through processing rate without loosening appetite?

Straight-through rate improves when referrals get more precise, not simply fewer. Make every referral name its trigger, then work the mix: referrals that fire on missing data belong back at intake, referrals that fire on thresholds nobody has revisited in three years belong in a review. In GoRules you branch the ruleset, replay historical submissions through both versions, and see exactly which risks change decision before anything reaches production. The largest gains usually come from segment-specific thresholds, because one global TIV or age cutoff pushes far more cases to a human than a set of tuned ones.

How does automated insurance underwriting hold up in a market conduct exam?

Every GoRules decision traces to the rules that fired and to the ruleset version that produced them, and every rule change records author, diff, approver, and timestamp. Those are the artifacts current supervision asks for: the NAIC Model Bulletin on the Use of Artificial Intelligence Systems by Insurers expects a documented governance program covering underwriting and rating, NYDFS Circular Letter No. 7 (2024) expects testing and documentation where automated underwriting and external consumer data are used, and the EU AI Act classifies life and health risk assessment as high-risk with logging and human-oversight duties. Keeping model scores as inputs and decisions as readable rules also gives that human oversight something specific to review. GoRules supplies the traceability; the obligations, testing, and filings remain yours.

Can we self-host so applicant medical and claims data never leave our infrastructure?

Yes. Deploy the platform with Docker or Kubernetes inside your own network, or embed the MIT-licensed ZEN Engine directly in your services so evaluation happens in-process and no applicant data crosses a vendor boundary. That matters most in life and health underwriting, where submissions carry medical records, prescription histories, and lab results. SSO and role-based access control govern who can view or change rules, and GoRules Cloud is available when managed hosting is acceptable.

The same engine,
next door.

One decision layer serves the whole institution - these use cases run on the same tables, versioning, and audit trail.

Insurance QuotingThe same rules, quote to bindFraud DetectionThe same rules layer for claims and paymentsLoan DecisioningThe same decision layer for lending

Bind the clean risks.
Refer the rest with reasons.

Start from the Insurance Underwriting Risk template, replay last quarter's submissions through it, and see what your straight-through rate could be by the end of the week.

Get started Book a demo