Phrony

Decisions, governed.

Insurance decisions sit between humans following checklists and rigid workflows that fail on edge cases. Phrony replaces both with policy-bound agents: one platform for claims, underwriting, and support — governed, explainable, and traceable from the first run.

Incoming claim
Policy termsFraud signalsClaim history
Claims Agent
Phrony gate
Approve
Reject
Escalate
Reasoning log

Claims Decisioning

Deploy an agent that receives an incoming claim, cross-references policy terms, checks fraud signals, reviews claim history, and produces an approve / reject / escalate decision - with the full reasoning documented. What used to take days of manual review runs in minutes, with every decision explainable and compliant.

Underwriting

An underwriting agent that reasons dynamically across applicant data, risk factors, and market conditions - without encoding every edge case into a rigid decision tree that breaks on anything unexpected. Phrony keeps the agent within your risk appetite and logs every factor it weighted.

Applicant data
Risk factors
Market conditions

Phrony — Risk appetite

Underwriting Agent
Decision + weighted factors
Reasoning log
Policyholder query

Phrony — allowed scope

Policy documentsKnowledge base
Support Agent
Human handoff
Answer
Reasoning log

Customer Support (Policy-Bound)

Deploy agents that handle complex policyholder queries by reasoning directly against policy documents and internal knowledge bases. They give accurate answers, escalate when uncertain, and never improvise outside the policies you've defined - because Phrony governs what they can and can't say.

Multi-Agent Claims Investigation

For complex claims, deploy a coordinated team: one agent assesses the claim, a second runs fraud detection, a third retrieves precedent cases. Phrony orchestrates them as a single governed session - with run trees that show exactly what each agent did and why.

Agent 1 — Claim assessment
Agent 2 — Fraud detection
Agent 3 — Precedent retrieval
Phrony — Governed session
Investigation outcome
session
agent 1
stepstep
agent 2step
agent 3
stepstep

Run tree — what each agent did, and why