Built for carriers, MGAs, and insurtech platforms
Agents that underwrite risk without creating it.
Carriers and MGAs are deploying agents across claims adjudication, underwriting, and fraud investigation. The NAIC AI Model Bulletin has now been adopted in 24 states — requiring documented AI governance, bias testing, and explainability for every adverse decision. Record builds that evidence trail as a byproduct of running agents.
Agents you can deploy today.
Every agent ships with Cedar policies pre-configured for insurance compliance requirements. Deploy in minutes, not quarters.
Claims Adjudication Agent
Processes first-party P&C claims end-to-end — validates coverage, orders inspections, calculates settlements. Routes claims above configurable thresholds to human adjusters via HITL before payment. Every decision logged with structured reason codes satisfying NAIC documentation requirements and state Unfair Trade Practice statutes.
Underwriting Decision Agent
Pulls risk data from ISO, LexisNexis, and credit bureaus under Cedar policies enforcing which external sources are permitted per line of business. Flags out-of-appetite risks for underwriter review. Produces bias assessment data and explainability documentation for NYDFS Circular Letter No. 7 examination.
Fraud Investigation Agent
Cross-references claim data against ISO ClaimSearch, SIU databases, and behavioral patterns. Flags suspicious claims for investigator review — cannot block payment or cancel a policy without explicit human confirmation. HITL gate creates the documented investigator decision chain required by state fraud statutes.
Regulatory Filing Agent
Compiles state annual statements, rate filings, and loss development data. Routes filings through responsible actuary and compliance officer sign-off before SERFF submission — actuary certification is enforced structurally, never bypassed.
Policy Renewal Agent
Analyzes loss history, exposure changes, and market conditions to generate renewal pricing recommendations. Routes renewals outside pricing authority to underwriting management. Produces Colorado ECDIS-compliant documentation of external data used in every pricing decision.
Every autonomous insurance decision is now a regulatory documentation problem. The NAIC AI Model Bulletin (adopted in 24 states as of March 2026) requires carriers to document AI systems, their inputs, and decision-making processes — and provide reason codes for adverse outcomes. Colorado's ECDIS regulation (expanded October 15, 2025 to include auto and health insurers) prohibits unfair discrimination through external consumer data and AI. NYDFS Circular Letter No. 7 (2024) requires bias testing and assessment before any AI system goes into production. An agent that traverses 15 tool calls to reach a claim denial has no mechanism to produce this documentation unless governance infrastructure was built in from day one.
Governance built for insurance.
NAIC reason codes generated automatically
The NAIC AI Model Bulletin (24 states as of March 2026) requires documented AI decision-making processes and reason codes for adverse outcomes. Record's Context Graph captures every tool call, data source, and gate decision as structured events. Reason codes for claim denials and underwriting declinations are generated as a byproduct of governance — not a separate system built after your first regulatory examination.
State-level regulatory variation in Cedar policy
Colorado ECDIS (expanded October 2025), NYDFS Circular Letter No. 7 bias testing requirements, and state Unfair Trade Practice statutes vary significantly by state and line of business. Cedar policies encode state-specific rules per agent invocation — agents in California apply different data access rights than those in Texas. One policy engine, enforced at the kernel.
HITL gates that survive litigation discovery
Every claim approval above threshold requires adjuster confirmation with the approver's identity, timestamp, and decision context captured. When a claim goes to litigation or a denial is challenged under bad-faith allegations, you have a complete timestamped record of human review — not just a note in the claim file.
Third-party data governance by line of business
Cedar policies control which external data sources — ISO, LexisNexis, MVR, credit — can be queried for which policy types. An auto agent cannot access life underwriting data sources, and vice versa. Prevents the proxy variable discrimination that Colorado ECDIS and NYDFS are actively examining carriers for.
Three gates. Every action. Zero exceptions.
Every agent action passes through all three enforcement layers simultaneously — not just one. Here's what that means for insurance.
Cedar policy enforces line-of-business data isolation, dollar-threshold HITL gates, and state-specific compliance rules per agent invocation. Fraud flag actions always route to human investigators. Autonomous claim denial above any threshold is structurally prevented — satisfying NAIC's documented human review requirement across all 24 adopting states.
Content filters prevent policyholder PII from crossing claim files. All LLM interactions with claim data archived for NAIC examination and state insurance department review. Rate limiting prevents bulk policyholder data extraction. Decision event logging provides the explainability data NYDFS requires.
eBPF sandbox ensures policyholder PII and actuarial model outputs cannot be exfiltrated outside the approved data environment — satisfying state insurance code data security requirements and the technical safeguard provisions of the NAIC model laws.
Ready to govern agents in Insurance?
See how Record works for your team in a 30-minute demo.