Using AI Agents for Initial Customer Onboarding Calls in Insurance

Independent agencies often rely on their AMS for everything, but its client intake tools are basic. The standard process involves a producer or CSR spending 20-30 minutes on the phone asking a new client a list of scripted questions. They type notes directly into the AMS or jot them down to be transcribed later. This process is slow, prone to data entry errors, and inconsistent from one agent to another.

Some agencies try using off-the-shelf chatbot builders to automate this. These tools fail because they cannot deeply integrate with core insurance systems. For example, a generic chatbot can't query Vertafore to see if a phone number already belongs to an existing client, leading to duplicate records. It also won't understand insurance-specific terms like 'additional insured' or 'subrogation,' forcing the client into frustrating conversational loops.

Consider an agency that wants to onboard a new small business client for a BOP policy. The agent must gather details on revenue, payroll, property, liability, and vehicle information. A generic form tool can't handle the conditional logic required. If the client mentions they have a delivery vehicle, a new set of questions about auto coverage should trigger. Off-the-shelf tools either can't do this or require complex, brittle branching logic that is difficult to maintain.

The structural problem is that these tools are stateless and disconnected from your system of record, the AMS. They are designed for simple lead capture, not for collecting the structured, interdependent data required to quote and bind an insurance policy. This forces your licensed agents to spend their time acting as human data-entry clerks instead of risk advisors.

The first step would be a process audit. Syntora would map out your current onboarding workflow, from the first point of contact to the data being entered in your AMS. We would review your question scripts for personal and commercial lines to understand the exact data fields, conditional logic, and documents you need to collect. You would receive a scope document detailing the proposed conversational flow and the specific integration points with your AMS.

The core of the system would be an AI agent powered by the Claude API, wrapped in a FastAPI service. This architecture is chosen for its ability to handle natural language and follow complex, stateful conversational paths. The FastAPI service would connect to your AMS platform (e.g., the Applied Epic API) to check for existing clients and push new data directly into the correct fields. The entire system would run on AWS Lambda, ensuring it costs nothing when idle and scales instantly, typically for under $50 per month.

When a new client engages, the agent would handle the initial conversation, collect all necessary information, and schedule a follow-up with a licensed agent. The final deliverable would be a complete, structured summary of the conversation and all collected data points posted as a note in the client's new AMS record. Your agent would start their first call with a complete file, ready to provide advice instead of asking basic questions. You receive the full source code and documentation.