Automate Initial Claims Intake and Documentation for Your Agency

Regional insurers rely on an AMS like Applied Epic or Vertafore as their system of record. These platforms are excellent for managing policies and client relationships but are not designed for unstructured data intake. The First Notice of Loss (FNOL) process remains a major source of manual work because claims arrive as emails and PDF attachments, not as clean, structured data.

For example, an adjuster receives an email with the subject "Water Damage Claim - Policy #789123" containing a PDF from a restoration contractor. The adjuster must open the email, download the PDF, find the policyholder's name, search for them in Applied Epic, create a new claim, then manually copy and paste the date of loss, cause of loss, and contact information from the PDF into dozens of separate fields in the AMS. This can take 15 minutes per claim. For an agency handling 25 new claims a day, this is over 6 hours of purely administrative work.

The structural problem is that an AMS is a database with a user interface. It expects data in a specific format. It has no native capability to read a sentence like "The water heater burst on Tuesday morning" and extract the cause and date of loss. Off-the-shelf document parsing tools exist, but they are generic and struggle with the specific vocabulary and formats of insurance forms, leading to high error rates and requiring just as much manual review.

The engagement would start with a discovery phase to audit your current claims intake process. Syntora would review the 3-5 most common FNOL document types you receive, whether they are standard ACORD forms, emails, or third-party reports. We would map every piece of required information back to its corresponding field in your Vertafore or HawkSoft AMS, creating a clear data specification before any code is written.

The technical system would be built around an event-driven architecture using AWS Lambda. A serverless function would be triggered by a new email in your dedicated claims inbox. This function would pass the email body and any attachments to the Claude API, which is highly effective at extracting structured data from unstructured text. We've used this same pattern to parse complex financial documents, and it applies directly to FNOL reports. A FastAPI application would then validate the extracted data against Pydantic schemas and use your AMS's API to create a new claim record.

The delivered system fully automates the creation of a structured claim file from an unstructured email or PDF. The system would also perform initial severity scoring (e.g., a 1-5 rating) based on keywords and route the new claim to the appropriate adjuster's queue. An adjuster's first touchpoint becomes reviewing a pre-filled claim record, not performing 15 minutes of data entry. The entire process, from email arrival to a new claim in your AMS, would take less than 30 seconds and could handle a volume of over 1,000 claims per day for less than $50 per month in AWS hosting costs.