AI Agent for Vehicle Insurance: Quote Generation, Claims Intake, and Policy Questions

AI Agents in Vehicle Insurance

Vehicle insurance involves complex workflows: generating quotes based on driver profiles and vehicle details, processing claims that range from minor fender-benders to total losses, answering policy questions about coverage limits and deductibles, and routing escalations to the right department. These interactions follow predictable patterns that an AI agent can handle efficiently.

The agent we build will generate personalized insurance quotes, walk customers through claims intake with proper incident classification, collect required documentation, and provide accurate answers about policy coverage.

Driver and Policy Data Models

from dataclasses import dataclass, field
from datetime import date
from enum import Enum
from typing import Optional

class CoverageType(str, Enum):
    LIABILITY = "liability"
    COLLISION = "collision"
    COMPREHENSIVE = "comprehensive"
    UNINSURED_MOTORIST = "uninsured_motorist"
    PIP = "personal_injury_protection"

class ClaimType(str, Enum):
    COLLISION = "collision"
    THEFT = "theft"
    WEATHER = "weather"
    VANDALISM = "vandalism"
    GLASS = "glass"
    ANIMAL = "animal_strike"
    HIT_AND_RUN = "hit_and_run"

@dataclass
class DriverProfile:
    driver_id: str
    name: str
    age: int
    years_licensed: int
    violations_3yr: int
    accidents_5yr: int
    credit_tier: str
    zip_code: str

@dataclass
class VehicleProfile:
    vin: str
    year: int
    make: str
    model: str
    trim: str
    annual_mileage: int
    ownership: str  # owned, financed, leased
    anti_theft: bool = False
    garage_kept: bool = False

@dataclass
class Policy:
    policy_number: str
    driver: DriverProfile
    vehicle: VehicleProfile
    coverages: dict[str, dict]
    premium_monthly: float
    deductible_collision: float
    deductible_comprehensive: float
    effective_date: str
    expiration_date: str

Quote Generation Tool

The quote engine scores risk factors and calculates premiums:

flowchart LR
    CALLER(["Policyholder or Lead"])
    subgraph TEL["Telephony"]
        SIP["Twilio SIP and PSTN"]
    end
    subgraph BRAIN["Insurance AI Agent"]
        STT["Streaming STT<br/>Deepgram or Whisper"]
        NLU{"Intent and<br/>Entity Extraction"}
        TOOLS["Tool Calls"]
        TTS["Streaming TTS<br/>ElevenLabs or Rime"]
    end
    subgraph DATA["Live Data Plane"]
        CRM[("CRM and Notes")]
        CAL[("Calendar and<br/>Schedule")]
        KB[("Knowledge Base<br/>and Policies")]
    end
    subgraph OUT["Outcomes"]
        O1(["Quote captured"])
        O2(["Claim opened in core"])
        O3(["Licensed agent handoff"])
    end
    CALLER --> SIP --> STT --> NLU
    NLU -->|Lookup| TOOLS
    TOOLS <--> CRM
    TOOLS <--> CAL
    TOOLS <--> KB
    NLU --> TTS --> SIP --> CALLER
    NLU -->|Resolved| O1
    NLU -->|Schedule| O2
    NLU -->|Escalate| O3
    style CALLER fill:#f1f5f9,stroke:#64748b,color:#0f172a
    style NLU fill:#4f46e5,stroke:#4338ca,color:#fff
    style O1 fill:#059669,stroke:#047857,color:#fff
    style O2 fill:#0ea5e9,stroke:#0369a1,color:#fff
    style O3 fill:#f59e0b,stroke:#d97706,color:#1f2937

from agents import function_tool

BASE_RATE = 120.00  # Monthly base

@function_tool
def generate_insurance_quote(
    driver_age: int,
    years_licensed: int,
    violations_3yr: int,
    accidents_5yr: int,
    vehicle_year: int,
    vehicle_make: str,
    vehicle_model: str,
    annual_mileage: int,
    zip_code: str,
    coverage_level: str = "standard",
) -> str:
    """Generate a vehicle insurance quote based on driver and vehicle details."""
    rate = BASE_RATE

    # Age factor
    if driver_age < 25:
        rate *= 1.45
    elif driver_age > 65:
        rate *= 1.15

    # Experience factor
    if years_licensed < 3:
        rate *= 1.30

    # Violations and accidents
    rate *= 1.0 + (violations_3yr * 0.15)
    rate *= 1.0 + (accidents_5yr * 0.25)

    # Vehicle age factor
    vehicle_age = 2026 - vehicle_year
    if vehicle_age <= 2:
        rate *= 1.20
    elif vehicle_age > 10:
        rate *= 0.85

    # Mileage factor
    if annual_mileage > 15000:
        rate *= 1.10
    elif annual_mileage < 7500:
        rate *= 0.90

    # Coverage levels
    coverage_configs = {
        "basic": {
            "multiplier": 0.70,
            "liability": "50/100/50",
            "collision_deductible": 1000,
            "comprehensive_deductible": 1000,
            "uninsured": False,
            "pip": False,
        },
        "standard": {
            "multiplier": 1.00,
            "liability": "100/300/100",
            "collision_deductible": 500,
            "comprehensive_deductible": 500,
            "uninsured": True,
            "pip": False,
        },
        "premium": {
            "multiplier": 1.35,
            "liability": "250/500/250",
            "collision_deductible": 250,
            "comprehensive_deductible": 250,
            "uninsured": True,
            "pip": True,
        },
    }

    config = coverage_configs.get(coverage_level, coverage_configs["standard"])
    monthly = round(rate * config["multiplier"], 2)
    annual = round(monthly * 12, 2)

    lines = [
        f"=== Insurance Quote ===",
        f"Driver: Age {driver_age}, {years_licensed} years experience",
        f"Vehicle: {vehicle_year} {vehicle_make} {vehicle_model}",
        f"Coverage: {coverage_level.title()}\n",
        f"Liability: {config['liability']}",
        f"Collision Deductible: ${config['collision_deductible']}",
        f"Comprehensive Deductible: ${config['comprehensive_deductible']}",
        f"Uninsured Motorist: {'Included' if config['uninsured'] else 'Not Included'}",
        f"Personal Injury Protection: {'Included' if config['pip'] else 'Not Included'}\n",
        f"Monthly Premium: ${monthly:.2f}",
        f"Annual Premium: ${annual:.2f}",
        f"\nQuote valid for 30 days.",
    ]
    return "\n".join(lines)

Claims Intake Tool

The claims tool classifies the incident and collects the required information:

@function_tool
def file_insurance_claim(
    policy_number: str,
    incident_date: str,
    incident_description: str,
    incident_location: str,
    police_report_number: Optional[str] = None,
    other_party_involved: bool = False,
    injuries_reported: bool = False,
) -> str:
    """File a new insurance claim with incident details and classification."""
    description_lower = incident_description.lower()

    # Auto-classify claim type
    if any(w in description_lower for w in ["hit", "crash", "rear-end", "collision"]):
        claim_type = ClaimType.COLLISION
    elif any(w in description_lower for w in ["stolen", "theft", "broke into"]):
        claim_type = ClaimType.THEFT
    elif any(w in description_lower for w in ["hail", "flood", "storm", "tree"]):
        claim_type = ClaimType.WEATHER
    elif any(w in description_lower for w in ["deer", "animal", "bird"]):
        claim_type = ClaimType.ANIMAL
    elif any(w in description_lower for w in ["windshield", "glass", "window"]):
        claim_type = ClaimType.GLASS
    elif any(w in description_lower for w in ["vandal", "keyed", "graffiti"]):
        claim_type = ClaimType.VANDALISM
    else:
        claim_type = ClaimType.COLLISION

    claim_number = f"CLM-{policy_number[-4:]}-{incident_date.replace('-', '')}"
    priority = "HIGH" if injuries_reported else "STANDARD"

    required_docs = ["Photos of damage", "Driver's license"]
    if police_report_number:
        required_docs.append("Police report copy")
    if other_party_involved:
        required_docs.append("Other driver's insurance info")
        required_docs.append("Other driver's contact details")
    if claim_type == ClaimType.THEFT:
        required_docs.append("Police report (required for theft)")
    if injuries_reported:
        required_docs.append("Medical records / bills")

    lines = [
        f"=== Claim Filed ===",
        f"Claim Number: {claim_number}",
        f"Policy: {policy_number}",
        f"Type: {claim_type.value.replace('_', ' ').title()}",
        f"Priority: {priority}",
        f"Date of Incident: {incident_date}",
        f"Location: {incident_location}\n",
        f"Required Documents:",
    ]
    for doc in required_docs:
        lines.append(f"  - {doc}")

    if injuries_reported:
        lines.append("\nIMPORTANT: This claim involves injuries and has been escalated to a senior adjuster.")

    lines.append(f"\nA claims adjuster will contact you within 24 hours.")
    return "\n".join(lines)

Policy Lookup Tool

POLICIES = {
    "POL-AA-12345": Policy(
        policy_number="POL-AA-12345",
        driver=DriverProfile("D-001", "Maria Gonzalez", 34, 16, 0, 0, "excellent", "90210"),
        vehicle=VehicleProfile("1HGCG5655WA027834", 2024, "Honda", "Accord", "Sport",
                               12000, "financed", True, True),
        coverages={
            "liability": {"limit": "100/300/100"},
            "collision": {"deductible": 500},
            "comprehensive": {"deductible": 250},
            "uninsured_motorist": {"limit": "100/300"},
        },
        premium_monthly=142.50,
        deductible_collision=500.0,
        deductible_comprehensive=250.0,
        effective_date="2026-01-15",
        expiration_date="2026-07-15",
    ),
}

@function_tool
def lookup_policy(policy_number: str) -> str:
    """Look up policy details including coverages, deductibles, and premium."""
    policy = POLICIES.get(policy_number.upper())
    if not policy:
        return f"Policy {policy_number} not found. Please verify the number."

    lines = [
        f"=== Policy Details ===",
        f"Policy: {policy.policy_number}",
        f"Insured: {policy.driver.name}",
        f"Vehicle: {policy.vehicle.year} {policy.vehicle.make} "
        f"{policy.vehicle.model} {policy.vehicle.trim}",
        f"VIN: {policy.vehicle.vin}",
        f"Period: {policy.effective_date} to {policy.expiration_date}\n",
        f"Coverages:",
    ]
    for cov, details in policy.coverages.items():
        name = cov.replace("_", " ").title()
        info = " | ".join(f"{k}: {v}" for k, v in details.items())
        lines.append(f"  {name}: {info}")

    lines.append(f"\nMonthly Premium: ${policy.premium_monthly:.2f}")
    lines.append(f"Annual Premium: ${policy.premium_monthly * 12:.2f}")
    return "\n".join(lines)

Assembling the Insurance Agent

from agents import Agent, Runner

insurance_agent = Agent(
    name="Insurance Advisor",
    instructions="""You are a vehicle insurance assistant. Help customers:
    1. Generate insurance quotes based on their driver profile and vehicle
    2. File claims with proper classification and document requirements
    3. Look up policy details and explain coverages
    Always explain coverage terms in plain language. For claims involving injuries,
    emphasize the importance of seeking medical attention first.""",
    tools=[generate_insurance_quote, file_insurance_claim, lookup_policy],
)

result = Runner.run_sync(
    insurance_agent,
    "I hit a deer on Highway 101 last night. My policy is POL-AA-12345. No injuries but significant front-end damage."
)
print(result.final_output)

FAQ

How do I make the quote engine more accurate?

Production insurance rating engines use actuarial tables, territory codes (based on zip code loss history), vehicle symbol ratings (from ISO), and proprietary loss models. Integrate with rating APIs from providers like Guidewire or Duck Creek. The agent collects inputs and calls the rating engine rather than computing premiums directly.

Can the agent handle policy changes like adding a driver or changing vehicles?

Yes. Add endorsement tools that modify an existing policy. Each endorsement type (add driver, change vehicle, adjust coverage) requires specific inputs. The tool should calculate the pro-rated premium adjustment and present it to the customer for approval before applying the change.

How should I handle sensitive personal information during claims intake?

Never store sensitive data like Social Security numbers in conversation logs. Use the agent to collect non-sensitive claim details, then redirect the customer to a secure portal for document uploads and sensitive information. Implement PII detection in the agent's response pipeline to redact any accidentally shared sensitive data.

---

#VehicleInsurance #ClaimsProcessing #QuoteGeneration #InsurTech #Python #AgenticAI #LearnAI #AIEngineering