Agent catalog

Status-aware domain agents for reinsurance work.

5 agents are labeled shipped. 12 remain tuning, preview, or development so public copy does not overstate maturity.

Agent activity

Product proof surface

Maturity labels travel with every handoff.

The catalog shows shipped, tuning, preview, and development agents without pretending every workflow is production-ready.

Intake agent shipped

Extraction agent shipped

Placement agent tuning

Contract agent preview

Agent handoff

Agent maturity board

shipped and preview separated

Input

Runtime state

source data enters the workflow

Output

Human review

reviewable state exits the workflow

Intake agent shipped

Source and state visible

Intakeshipped

Intake Agent

Routes inbound submissions from email, portal, or API to the correct extraction pipeline and LOB handler.

In:Raw submission (PDF, Excel, email), source channel

Out:Classified submission with LOB tag, priority score, assigned extractor

Async

Code

Intakeshipped

Email Summary Agent

Summarizes long broker email threads into structured submission context, preserving attachments and action items.

In:Email thread text + attachments

Out:Structured summary, extracted submission intent, missing-info checklist

Async

Code

Intaketuning

Follow-Up Generator

Drafts contextual follow-up emails when submissions are incomplete or underwriters request clarifications.

In:Missing-info report, original submission, tone preference

Out:Formatted follow-up email with specific questions and document requests

Sync

Code

Extractionshipped

Extraction Agent

Baseline document extractor for PDFs, Word, and Excel. Maps unstructured text to canonical schema fields.

In:Document bytes, target schema name

Out:Structured JSON with confidence scores per field

Async

Code

Extractiontuning

Reasoning Extraction Agent

Chain-of-thought extractor for complex contract language. Explains its reasoning alongside each extracted value.

In:Complex contract or treaty wording, schema

Out:Extracted fields + reasoning trace per field

Async

Code

Extractionshipped

LOB Extractors

Line-of-business-specific extractors (property, casualty, marine, aviation, cyber) with domain vocabularies.

In:LOB-tagged document

Out:LOB-specific structured data (limits, deductibles, premiums, exclusions)

Async

Code

Extractionpreview

Extraction A/B Testing

Routes documents between control and treatment extractors to measure quality improvement over time.

In:Document, routing strategy (random / deterministic / confidence-gated)

Out:Extraction result + group assignment + quality metrics

Async

Code

Assessmenttuning

Risk Assessment Agent

Evaluates submission risk profile against portfolio appetite, accumulation limits, and cat exposure.

In:Structured submission, portfolio context, peril data

Out:Risk score, appetite match, accumulation flags, recommended capacity

Async

Code

Assessmenttuning

Validation Agent

Runs rule-based validation against submission schemas, flagging inconsistencies and missing required fields.

In:Extracted submission JSON, validation rule set

Out:Validation report with errors, warnings, and fix suggestions

Sync

Code

Placement & Treatytuning

Placement Agent

Optimizes layer allocation across reinsurer panels using LP modeling and market-intelligence signals.

In:Risk structure, panel constraints, pricing targets

Out:Recommended layer matrix, reinsurer allocation, pricing benchmark

Async

Code

Placement & Treatytuning

Treaty Comparison Agent

Compares treaty terms across versions or counterparty proposals, highlighting deltas in wording and economics.

In:Two treaty documents or versions

Out:Side-by-side delta report with clause-level diffs and economic impact

Async

Code

Placement & Treatypreview

Contract Drafting Agent

Generates MRCv3-compliant draft contract clauses from risk assessments and placement structures.

In:Risk assessment, placement structure, negotiated terms

Out:Draft contract with recitals, clauses, schedules, execution block

Async

Code

Operationsshipped

Bordereaux Agent

Reconciles periodic bordereaux reports against expected premium and claims data. Detects anomalies and missing entries.

In:Bordereaux Excel/CSV, expected premium register, claims feed

Out:Reconciliation report, anomaly list, booking recommendations

Async

Code

Operationstuning

Claims Agent

Triages inbound claims notifications using LLM analysis, extracts key facts, flags potential coverage issues, fraud risk, and routes to the correct handler.

In:Claims notification (email, ACORD, PDF)

Out:Structured claim record, severity score, coverage flag, recommended reserve, handler assignment

Async

Code

Operationstuning

Reserving Agent

AI copilot for actuarial reserving that generates narrative explanations of reserve changes, adequacy assessments, and strengthening recommendations.

In:Previous and current reserve estimates, line of business data

Out:Narrative explanation, adequacy assessment, uncertainty bands, next actions

Async

Code

Compliancetuning

Compliance Agent

Checks submissions and documents against Lloyd's compliance rules, GDPR requirements, and internal policy standards.

In:Document or submission, compliance framework (Lloyd's / GDPR / internal)

Out:Compliance score, violation list, remediation actions, evidence bundle reference

Async

Code

Basepreview

Agent Framework (base.py)

Shared base class for all agents: lifecycle management, observability hooks, retry logic, and cost tracking.

In:Agent configuration, task payload

Out:AgentResult with data, cost, latency, and trace ID

Mixed

Code

Design-partner review

Review the agent roster with the ReRao team.

Discuss which agents match your workflow, which are evaluation-ready, and which are still roadmap-facing.

Start design-partner evaluation Login

Evaluation requested

Review-ready

Review pathBroker intake through the review packet

Evidence trailSource links, exceptions, and audit context

Evaluation scopeEvaluation language only, no live outcome claims