Automate FNOL Intake and Cut Response Times by 95%
Insurance agencies automate First Notice of Loss intake by using AI to parse claim reports. The AI scores claim severity and routes the case to the correct adjuster with a summary.
The system's complexity depends on intake channels (email, web form, PDF) and your Agency Management System integration. A single email inbox feeding into Applied Epic is a standard build. Parsing scanned faxes and integrating with a custom legacy system requires more discovery.
We built an FNOL triage system for a 6-adjuster agency that processed claims from three email inboxes. We cut their average first-response time from 4 hours to 12 minutes within a 4-week build cycle.
What Problem Does This Solve?
Many agencies try to use their Agency Management System's built-in rules. Vertafore's basic email rules can file an email into a client's record, but they cannot read the content. This means an adjuster still has to manually open the email, read the FNOL report, determine if it is a minor fender bender or a total loss, and then decide who gets the case.
A typical agency with 10 employees gets 40 FNOL emails a day. The office manager spends their first two hours every morning reading each one. They have to decide if a claim is for a commercial or personal line, which adjuster handles that client, and how urgent it is. A complex commercial property claim gets delayed because it arrived after a batch of 15 simple auto glass claims, making a high-value client wait 3 hours for a response.
The core issue is that keyword-based rules and manual sorting cannot understand context. An email containing the word "fire" could be a catastrophic commercial property loss or a minor kitchen flare-up in a rental unit. Without an AI model that understands the narrative of the claim, every report requires human review, creating a permanent bottleneck that scales with claim volume.
How Does It Work?
We start by connecting to your FNOL intake channels, typically an Office 365 or Google Workspace inbox, using their respective APIs. We pull the last 3 months of claim reports, totaling 1,000 to 5,000 documents, to create a dataset for prompt engineering. We use Python's `imaplib` for email parsing and `pypdf` for extracting text from PDF attachments.
The core logic is a FastAPI service that receives the raw claim text. This text is sent to the Claude API with a carefully engineered prompt to extract key entities: policy number, claimant name, incident type, and a narrative summary. The API call returns a structured JSON object in under 800ms. We then use a second prompt to assign a severity score from 1-10 based on the extracted information.
The FastAPI service runs on AWS Lambda, costing less than $30/month for up to 10,000 claims processed. Once a claim is scored, a webhook pushes the data directly into your AMS. We have pre-built connectors for Applied Epic, Vertafore, and HawkSoft. Routing logic, defined in a simple Python dictionary, assigns the claim to the right adjuster's queue based on severity score and line of business.
Every AI decision is logged to a Supabase database table with its confidence score. For any claim scoring above a 7/10 severity, the system requires a human review. A notification is sent to a manager's Slack channel with a link to the record, ensuring high-stakes claims get immediate human oversight. The entire process from email receipt to adjuster notification takes under 90 seconds.
What Are the Key Benefits?
From 4 Hours to 12 Minutes
Our system for a 6-adjuster agency cut average first-response time by 95 percent. Claims are parsed, scored, and assigned in under two minutes.
Fixed Build, Predictable Hosting
A one-time project cost with monthly hosting on AWS Lambda typically under $30. No per-claim fees or per-user licenses that penalize growth.
You Own the System and Code
You get the full Python codebase in your own GitHub repository. There is no vendor lock-in; your system can be modified by any developer.
Self-Monitoring with Real-Time Alerts
We use structlog for structured logging and send alerts to Slack via webhook if the Claude API fails or parsing errors exceed 1 percent.
Connects Directly to Your AMS
Native API or webhook integrations for Applied Epic, Vertafore, and HawkSoft. Adjusters work in their existing system, no new software to learn.
What Does the Process Look Like?
Week 1: System Access and Data Collection
You provide read-only access to your FNOL intake channels and your Agency Management System. We pull historical data to build the training set.
Weeks 2-3: Core System Development
We build the FastAPI service, refine the Claude API prompts, and configure the routing logic. You receive a demo of the system processing live claims.
Week 4: Integration and Deployment
We connect the system to your AMS and deploy it to AWS Lambda. Your adjusters begin receiving AI-triaged claims in a test environment.
Weeks 5-8: Monitoring and Handoff
We monitor system performance and AI accuracy for 30 days post-launch. You receive a runbook detailing the architecture and maintenance procedures.
Frequently Asked Questions
- How much does a custom FNOL system cost?
- Pricing depends on the number of intake sources and the complexity of your AMS integration. A system for a single email inbox connecting to Applied Epic is a 4-week build. Supporting multiple channels or legacy systems extends the timeline. We provide a fixed-price quote after our initial discovery call at cal.com/syntora/discover.
- What happens if the AI misinterprets a claim?
- The system logs every decision with a confidence score. If the score is low, or if the claim is high-severity, it is flagged for mandatory human review. For critical errors, the system forwards the original FNOL email to a designated manager's inbox with an error alert, ensuring no claim is ever lost.
- How is this different from using an off-the-shelf Insurtech platform?
- Insurtech platforms offer a suite of tools with a monthly per-user subscription. Syntora builds just the system you need, and you own the code. There are no recurring seat licenses. This is for agencies that need a production-grade system for a specific, high-value process without buying an entire platform with features they will not use.
- How is sensitive claimant data handled?
- Claim data is processed in-memory and is not stored long-term outside your AMS. We use AWS Lambda for compute, which provides a secure, isolated environment. All API connections, including to the Claude API and your AMS, use TLS 1.2+ encryption. We sign a DPA and adhere to your data handling policies.
- Do I need technical staff to run this?
- No. The system is designed to run with zero manual intervention. We set up monitoring that alerts us directly if there are any issues during the initial support period. After handoff, the provided runbook explains how to handle common scenarios. Any competent IT consultant can manage the system with the documentation we provide.
- Why do you use the Claude API instead of another model?
- We use Anthropic's Claude 3 family via API because it has a large context window and strong instruction-following capabilities. This allows us to get high-quality structured data from messy, unstructured emails and PDFs reliably. We continuously test new models, but Claude currently provides the best performance for this specific task.
Related Solutions
Ready to Automate Your Small Business Operations?
Book a call to discuss how we can implement ai automation for your small business business.
Book a Call