Custom Voice AI for Insurance Intake: A Better Provider
The best voice AI provider for insurance brokers is a custom-built intake system using Anthropic's Claude API. This approach avoids per-seat fees and integrates directly with your existing agency management system (AMS).
We built a voice intake system for a 15-person commercial lines broker. They were spending over two hours per day manually entering call notes into Applied Epic. The system now processes 60+ daily calls, transcribes them with 98% accuracy, and drafts a complete policy summary in under 90 seconds.
A custom build is for agencies that need more than just a call transcript. It is for brokers who need structured data like VINs, policy numbers, and coverage limits extracted from a call and automatically entered into the correct fields in their AMS. The system handles the entire workflow from recording to data entry.
What Problem Does This Solve?
Many brokers try to solve this with generic transcription tools like Otter.ai. These services provide a wall of text, but the broker still has to manually read the transcript, find the relevant details, and copy them into the AMS. This saves almost no time and fails to solve the core data entry problem.
Next, they look at call center platforms like Aircall or Talkdesk. These are powerful phone systems but their AI features are built for sales coaching and sentiment analysis, not insurance intake. They might tell you a customer sounded upset, but they cannot reliably extract a driver's license number or prior carrier information. These platforms also come with high per-user monthly fees that are difficult for a 10-broker agency to justify.
A typical failure scenario involves a personal lines broker on a 10-minute call for a new auto policy. They take handwritten notes, then spend another 15 minutes typing the VIN, address, and coverage limits into Vertafore. With 5 new business calls a day, that's over an hour of data entry per broker. This manual process is slow, prone to typos, and takes away from time that could be spent quoting and binding policies.
How Does It Work?
When a call is completed, the audio file is automatically uploaded to an AWS S3 bucket. This triggers an AWS Lambda function that calls the Deepgram API for transcription. We use Deepgram's model trained specifically on phone call audio, which provides speaker labels and 98% accuracy. For a typical 5-minute intake call, the full transcript is ready in under 30 seconds.
The transcript is then passed to a FastAPI service that calls the Claude 3 Sonnet API. We use a carefully crafted prompt that instructs the model to act as an insurance intake specialist. It extracts specific entities like policyholder names, addresses, vehicle information, and requested liability limits into a structured JSON object. The process of extraction and summarization takes less than 15 seconds.
This structured JSON output is then mapped to the fields in your agency's AMS. Using httpx, we make an API call to post the data directly into your system, whether it is Applied Epic, Vertafore, or HawkSoft. A complete copy of the transcript, summary, and extracted data is archived in a Supabase database for audit purposes. This costs less than $20 per month for over 5,000 calls.
For monitoring, the entire process is instrumented with structlog. If the AMS API is down or Claude returns an unparsable response, a Pydantic validation error triggers a Slack alert. We use Amazon SQS as a simple queue to automatically retry failed jobs up to 3 times, ensuring that transient network issues do not result in lost data.
What Are the Key Benefits?
From Call End to AMS Entry in 60 Seconds
The entire pipeline, from transcription to structured data entry in your AMS, completes in under a minute. Brokers review and approve, they don't re-type.
Pay for Usage, Not for Seats
Your cost is based on API calls, typically under $0.50 per intake call. No expensive per-user monthly software licenses.
You Own the System and the Code
We deliver the complete source code to your GitHub repository. There is no vendor lock-in and you control the entire data pipeline.
Know Instantly When an Intake Fails
Get real-time Slack alerts if an API fails or data cannot be parsed. The system includes automatic retries for transient network errors.
Works With Your Existing AMS
We build direct API integrations for platforms like Applied Epic, Vertafore, and HawkSoft. No need to change your core agency software.
What Does the Process Look Like?
Scoping & AMS Access (Week 1)
You provide read-only API access to your AMS and 10-20 sample call recordings. We map the required data fields and finalize the extraction logic.
Core System Build (Week 2)
We build the core FastAPI service, transcription pipeline, and Claude API integration. You receive a link to the private GitHub repository to see progress.
Integration & Testing (Week 3)
We connect the system to your phone system and AMS. You test with live calls in a staging environment and receive daily progress summaries.
Launch & Monitoring (Week 4+)
We deploy the system to production. For the first 30 days, we monitor every transaction and provide a weekly performance report before handing over the runbook.
Frequently Asked Questions
- How much does a custom voice intake system cost?
- A typical build takes 3-4 weeks. The cost depends on the complexity of your AMS integration and the number of distinct data points to extract. A simple auto policy intake is faster to build than a complex commercial liability workflow. We provide a fixed-price quote after our initial discovery call.
- What happens if the transcription or AI summary is wrong?
- The system flags summaries with low confidence scores for human review. Brokers can edit the summary and structured data directly in the AMS before finalizing. The original audio and transcript are always linked, so it takes 15 seconds to verify a detail. This feedback can also be used to refine the extraction prompts over time.
- How is this different from using a call center SaaS like Talkdesk?
- Talkdesk and other CCaaS platforms are phone systems with add-on AI features, but they are not built for insurance-specific data extraction. They provide sentiment analysis, but they won't pull a VIN into the correct field in Applied Epic. Syntora builds the last-mile integration that makes the voice data useful for your specific workflow.
- Is this compliant with privacy regulations?
- Yes. The entire system is deployed in your own AWS account, giving you full control over the data. We use BAA-compliant services where necessary and ensure data is encrypted in transit and at rest. Unlike multi-tenant SaaS platforms, your customer data is never commingled or used to train third-party models.
- What phone systems does this work with?
- We can integrate with any VoIP system that allows call recording and can export audio files, for example via webhook or to a cloud storage bucket. We have previously integrated with RingCentral, Dialpad, and Twilio. Programmatic access to call recordings is the key requirement, which we verify during discovery.
- What are the ongoing maintenance costs?
- After the one-time build fee, your only costs are for cloud services and APIs. For an agency processing 50 calls a day, this is typically under $200 per month for AWS, Deepgram, and Claude API usage. We offer an optional flat monthly retainer for monitoring, updates, and prompt tuning if you do not have an internal technical resource.
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