Minerva AI
The Commercial Decision Engine Behind Carpe's Underwriting and Pricing Products
Minerva is Carpe's purpose-built commercial AI engine, fusing more than a decade of proprietary SMB data with agentic AI reasoning so carriers get the shortest path from data to decision at quote, at renewal, and across the policy lifecycle.
Two Gears, One Engine
Minerva runs in two operating modes that share the same data, models, and reasoning fabric, so carriers get sub-second answers at the point of quote and deeper multi-minute analysis on the risks that warrant a closer look without standing up two different systems.
Minerva Instant
Sub-second commercial intelligence at the point of quote, built on a continuously maintained datastore of 50M+ small business profiles covering approximately 90% of all insurable small businesses in the U.S., returning 30+ predictive indexes and scores along with business characteristics, hours of operation, and reason codes ready to embed directly into rating, triage, and straight-through processing.
Minerva Thinking
One to three minutes of deep analysis for complex middle-market risks and edge cases, combining the datastore with live online search and AI-powered analysis. Match rates run up to 25% higher than Instant alone.
Minerva at Every Stage of the Policy Lifecycle
Minerva does not stop working once a quote is bound. The same engine follows each policy from first submission through loss-ratio feedback, refreshing the picture every time the carrier needs it.
Step 01
Quote
Sub-second commercial intelligence and Carpe Risk Score flow into the rating engine on first submission, sharpening binding decisions the moment a quote arrives.
Step 02
Bind
Thinking-mode analysis runs on flagged risks before bind, pulling deeper operational and online signals into the file without slowing the workflow.
Step 03
Mid-Term
Minerva watches the bound book for material changes in business activity, web presence, and operational signal that would shift appetite or pricing.
Step 04
Renewal
Each renewal returns a refreshed picture of the risk and the score, with carrier-defined logic deciding which accounts get the deeper Thinking-mode review.
Step 05
Loss-Ratio Feedback
Outcomes feed back into Minerva so the models keep learning from how each carrier's book actually performs against the predictions Minerva produced.
How Minerva Thinks
Four capabilities anchor every Minerva response, each one tuned to the realities of commercial underwriting and pricing (rather than retrofitted from a consumer model).
Proprietary Commercial Data Foundation
More than a decade of refined SMB business data sits underneath Minerva, with AI-driven data quality that lifted precision and recall by fifteen percent across the commercial graph.
Agentic AI Reasoning
Multi-step reasoning chains emulate how a senior underwriter or pricing analyst evaluates a risk, with every decision path traceable back to the data that produced it.
Real-Time Risk Score Integration
Carpe Risk Score is baked into every Minerva response, whether that response arrives in sub-second Instant mode or a deeper Thinking-mode analysis, so the rating engine is ready on day one.
Carrier-Defined Logic and Thresholds
Carriers tune Minerva to their own appetite, segmentation, and pricing strategy without writing code, so the same engine fits a high-volume SMB carrier and a complex middle-market book.
What Minerva Produces
Every Minerva response, whether sub-second or multi-minute, is built from the same set of scores, indexes, and operational attributes. All of it comes with reason codes and public-source evidence, so underwriting decisions stay consistent, explainable, and defensible.
Click or tap any tile for the full description.
Scores
Carpe Risk Score
A composite 1-5 score that brings together indexes, anomaly detection, loss propensity, classification, trends, and business characteristics into a single decision point for automation and underwriting action.
Loss Propensity Score
A pre-trained model predicting likelihood of a loss, built on historical policy and claims data. Carriers using this score have reduced adverse selection by up to 40%.
Anomaly Score
Measures operational complexity relative to industry peers. Simple, in-appetite businesses score high; complex or atypical businesses score low, signaling the need for manual review.
Indexes
Customer Rating
Aggregate signal from customer reviews and ratings across the open web.
Reputation
Broader brand and operational reputation from public sources.
Health and Sanitation
Inspection records, complaints, and observed conditions where available.
Maintenance and Condition
Physical state of the operation from observable evidence.
Visibility
Web presence, business activity signal, and operational footprint.
Attributes
Business Characteristics
200+ boolean indicators of actual operations: services offered, delivery, alcohol, pools, late-night hours, loss-control observations, and more.
Business Trends
12-month change detection across operations, hours, profiles, and customer reviews, with directional indicators and carrier action recommendations.
Location Insights
Proximity and density scores capturing address-level and zip-code-level risk from surrounding businesses and conditions.
NAICS/SIC Classification Validation
Class-code verification using evidence from actual business operations, catching misclassification before it becomes premium leakage.
AI Business Summaries
Multi-paragraph, AI-generated descriptions of business operations, risk factors, and underwriting considerations (Thinking mode).
Tenant and Occupancy Data
For commercial property and lessors risk: active tenants, occupancy profiles, aggregated hours, late-night exposure, and industry breakdown at the address.
What Minerva Delivers in Production
Results carriers are seeing in production today across BOP, General Liability, Workers Compensation, and commercial property lines.
reduction in adverse selection
increase in straight-through processing
points of pricing model lift
loss ratio improvement
expense ratio reduction
return on investment
Individual results vary based on book composition and implementation depth.
See Minerva on Your Risks
The fastest way to evaluate Minerva is on a sample of your own commercial book, with both modes and Carpe Risk Score returned alongside your current decisioning so the lift is visible side by side.