The problem

What the customer was up against.

A specialty commercial insurer was receiving 180,000 broker submissions a year, with 60% arriving as PDFs or Excel attachments to email. Underwriters spent the majority of their time on data extraction and triage rather than risk evaluation. Average submission-to-quote was 4 business days; brokers were defecting to faster carriers on standard risks.

The solution

What xyner built.

Deployed xyner with a Submission Intake Agent, Risk Evaluation Agent and Quote Drafting Agent — full integration with the policy admin system, the rating engine, treaty rules and the broker portal.

The outcomes

Measured impact.

Submission-to-quote dropped from 4 days to 4 hours on agent-handled submissions; underwriter time on data entry fell by 75%; broker NPS jumped 22 points; loss-ratio impact neutral-to-positive (no quality degradation).

Executive summary

At a glance.

Situation

A specialty commercial insurer was receiving 180,000 broker submissions a year, with 60% arriving as PDFs or Excel attachments to email. Underwriters spent the majority of their time on data extraction and triage rather than risk evaluation. Average submission-to-quote was 4 business days; brokers were defecting to faster carriers on standard risks.

Intervention

Deployed xyner with a Submission Intake Agent, Risk Evaluation Agent and Quote Drafting Agent — full integration with the policy admin system, the rating engine, treaty rules and the broker portal.

Outcome

Submission-to-quote dropped from 4 days to 4 hours on agent-handled submissions; underwriter time on data entry fell by 75%; broker NPS jumped 22 points; loss-ratio impact neutral-to-positive (no quality degradation).

Industry

Insurance · Commercial / specialty

A specialty commercial property & casualty insurer (Lloyd's of London market)

Scope

United Kingdom, Bermuda, North America

Submission intake, risk evaluation, quote generation, broker comms

Duration

9 weeks pilot, 6 months full rollout

From contract signature to full rollout.

Architecture

What the deployment actually looks like.

Specialty underwriting is a judgment-heavy workflow that depends on the underwriter's ability to look across the submission, the risk and the broker relationship. The deployment automates intake and standard rating but leaves underwriting judgment with the underwriter.

Submission Intake Agent

Parses incoming submissions (PDF, Excel, email body); extracts risk attributes; flags missing information; routes to the right underwriting desk by line of business and territory.

Risk Evaluation Agent

Pulls historical loss data, third-party risk data, prior policy history; produces a risk packet for the underwriter with key drivers cited.

Quote Drafting Agent

Generates the indicative quote using the rating engine and treaty constraints; produces a draft for underwriter review on standard risks, full quote for routine business.

Broker Comms Agent

Communicates with the broker through the existing broker portal; acknowledges submissions, requests missing info, sends quotes and follow-ups.

Underwriter workbench

Underwriters review agent output, override or augment, sign off; the workbench captures every override with rationale to feed back into the risk model.

Reinsurance & treaty integration

Quotes respect inwards treaty terms automatically; outwards reinsurance facultative considerations are surfaced for underwriter attention.

Implementation timeline

How the rollout sequenced.

The pilot covered two specialty lines (Marine Cargo and Professional Indemnity) before expanding to the broader commercial book.

Weeks 1-3

Foundations & rating-engine integration

Deploy data plane; integrate policy admin, rating engine, treaty rules and broker portal; configure RBAC inheritance from the insurer's identity provider.

Weeks 4-5

Agent configuration

Configure the four agents against the two pilot lines; load underwriting guidelines and prior loss data into RAG; complete bias-testing baseline.

Weeks 6-7

Shadow mode

Agents process real submissions in shadow; underwriters review outputs; thresholds calibrated.

Weeks 8-9

Pilot live

Live in Marine Cargo and PI with underwriter approval on every quote; broker satisfaction surveyed weekly.

Months 3-4

Commercial book expansion

Add Property, GL and Cyber lines; introduce per-line variations; expand to Bermuda and North America hubs.

Months 5-6

Routine-business autonomy

Autonomy thresholds calibrated; standard renewals and routine new business flow with underwriter sample review.

Governance & controls

How the deployment is governed.

Commercial underwriting carries Solvency II governance, fair-treatment requirements, model-validation expectations, and growing AI-specific oversight from PRA, FCA and equivalent regulators.

Model validation

Each agent version is independently validated by the insurer's actuarial function before deployment; loss-ratio impact monitored continuously.

Underwriter authority

Underwriters retain authority over every bound quote; the platform produces inputs and drafts, the underwriter produces the decision.

Fair treatment

Quote logic is tested for fair-treatment outcomes across broker segments and customer attributes; results reviewed by the insurer's conduct function.

Audit-grade trail

Every submission, every agent action, every override and every approval is captured to a tamper-evident audit trail.

Treaty discipline

Inwards and outwards treaty terms enforced at quote time; treaty breaches structurally impossible by design.

What other enterprises can learn

Three transferable lessons.

Three lessons for commercial insurers deploying agentic AI in underwriting.

1

Automate intake and triage before automating judgment

The biggest cycle-time win came from intake automation. Underwriters did not feel displaced — they felt unblocked.

2

Loss-ratio neutrality is the only convincing argument

Brokers are happy when quotes are fast; CFOs only sign off when loss ratios hold. Continuous loss-ratio monitoring is non-negotiable.

3

Treaty discipline is a one-time architectural win

Encoding treaty terms in the platform meant quote-against-treaty breaches stopped happening. Underwriters got their attention back.

We used to compete on price and lose on speed. We now compete on judgment and win on speed. That's the underwriter business we wanted.
Chief Underwriting Officer, specialty commercial insurer

Reference call available through your xyner account team; the deployment has been referenced in PRA materials on AI in insurance underwriting.

Talk to a partner

Could the same outcome work in your environment?

Tell us your sector. A senior xyner partner will walk you through a tailored plan.