Automate Accounting Client Intake with Voice AI
Syntora is an AI automation company that builds custom voice AI solutions for accounting firms. We specialize in systems that transcribe and summarize client voicemails for automated data entry.
The scope is determined by your phone system's API and your practice management software. A firm using RingCentral and Karbon can have a direct integration. A firm with an on-premise phone system or legacy software may require an intermediate step, like processing audio files from a shared inbox.
We built a voice intake system for a 15-person CPA firm that received over 50 client voicemails a day. Their manual process took 4-5 minutes per message. Our system, deployed in 3 weeks, processes each voicemail and creates a structured task in their workflow software in 8 seconds.
What Problem Does This Solve?
Most accounting firms rely on their VoIP system's built-in voicemail-to-email feature. Services like RingCentral or Dialpad provide transcriptions, but they are often inaccurate with financial terms, confusing 'accrual' with 'a cruel'. This forces staff to listen to the original audio anyway, defeating the purpose. The output is just a block of text, not structured data that can be used for automation.
A common next step is a dedicated transcription tool like Otter.ai. While more accurate, it still only produces a text document. An administrator must read the transcript, identify the client, understand the request, open the practice management software like TaxDome, find the client record, create a new task, and copy-paste the relevant details. For a 90-second voicemail, this is a 5-minute manual workflow.
This manual process is not only slow but also error-prone. Transposing a tax year or misunderstanding a client's request creates rework for accountants. At 30 voicemails per day, this consumes over two hours of administrative time that could be spent on client-facing work. The core issue is that these tools don't connect the audio source to the system of record.
How Does It Work?
We start by connecting to your phone system's API to receive new voicemail audio files in real time. The audio is pushed to an AWS S3 bucket, which triggers an AWS Lambda function. This serverless architecture means you only pay for the exact time it takes to process a voicemail, which is typically a few seconds.
Our Python-based Lambda function uses the Claude 3 Sonnet API to perform transcription and structured data extraction in a single call. We engineer a prompt that specifically instructs the model to identify the client's name, the topic (e.g., quarterly taxes, 1099 question), and any mentioned dates or form numbers. The AI returns a clean JSON object, like `{"client_name": "John Smith", "topic": "Form 941 inquiry"}`, in about 8 seconds for a 60-second voicemail.
This structured JSON data is then passed to a second function that uses httpx for an asynchronous API call to your practice management software. We write code to interact directly with the Karbon, TaxDome, or Canopy API. The function creates a new task, populates it with the extracted data, assigns it to the correct team member based on routing rules, and attaches the full transcript. The entire automated workflow completes in under 15 seconds.
For monitoring, we use FastAPI and Supabase to log every transaction and its outcome. We implement `structlog` for structured logging and configure CloudWatch alerts. If an API call to your practice software fails after 3 retry attempts, a notification is sent to a designated Slack channel with a link to the audio file for manual handling. Typical hosting costs for processing up to 2,000 voicemails a month are under $25.
What Are the Key Benefits?
Client Voicemails in Your System in 15 Seconds
Go from a new voicemail notification to a structured, assigned task in your practice management software in under 15 seconds, not 5 minutes of manual work.
Pay Once for the Build, Not Per Minute
A single, fixed-price project replaces per-minute transcription fees. Monthly hosting on AWS is often less than a single user's software license.
You Get the Full Source Code
The complete Python codebase is delivered to your GitHub repository. You are never locked into our service and can have any developer extend it.
Know Instantly When an Intake Fails
Built-in monitoring sends a Slack alert if a voicemail fails to process or can't connect to your CRM, including the audio file for manual review.
Connects Directly to Your Practice Software
We build native API integrations for systems like Karbon, TaxDome, and Canopy. No more copying and pasting between your phone system and work queue.
What Does the Process Look Like?
Scoping & API Access (Week 1)
You provide read-only API credentials for your VoIP phone system and practice management software. We map out the exact data fields you need extracted.
Core Logic & Model Prompting (Week 2)
We build the core Python application for transcription and data extraction. You receive a demo showing sample voicemails processed into structured JSON data.
Integration & Deployment (Week 3)
We deploy the system on AWS Lambda and connect the API endpoints. You get a private Slack channel to see live processed voicemails populate your system.
Monitoring & Handoff (Week 4)
We monitor the live system for one week to resolve any issues. You receive the full source code, API documentation, and a runbook for maintenance.
Frequently Asked Questions
- What factors determine the cost and timeline?
- The primary factors are the number of systems to integrate and the quality of your phone system's API. A direct integration with a modern VoIP platform like RingCentral is straightforward. Connecting to an on-premise phone system or one without a public API requires more work. Most projects are completed in 3-4 weeks. We provide a fixed-price quote after the initial discovery call.
- What happens if a client's audio is unclear or the AI misinterprets it?
- The system is designed to fail gracefully. If the AI model's confidence score for a key piece of data (like a client name) is below a set threshold, it flags the item for human review. The system creates a generic task in your software with the full transcript and audio file attached, and sends an alert. This prevents bad data from being entered automatically while ensuring no message is lost.
- How is this better than using a virtual assistant (VA) service?
- A VA introduces human error, security risks with client data, and delays. They work set hours and cost a fixed monthly fee regardless of call volume. This automated system runs 24/7, processes calls in seconds, and has a variable cost tied directly to usage (typically cents per call). It eliminates the need for a person to listen to sensitive client financial details just for data entry.
- Does this only work for voicemails, or can it handle live calls?
- This system is designed for asynchronous intake from voicemails. Building a real-time agent for live calls is a different architecture involving WebSocket connections and much lower latency requirements. We typically build the voicemail system first, as it solves the highest-volume administrative problem. A live call summarization tool can be a future project, built on the same core components.
- What accounting-specific terms can it recognize?
- We fine-tune the prompts for the AI model to specifically listen for and correctly transcribe terms like 'accrual,' 'Form 941,' 'quarterly estimates,' and '1099-NEC.' During the discovery phase, we ask for a list of 10-15 terms or client names that are commonly misunderstood by standard transcription services to ensure the system is accurate for your practice.
- Is the client data processed securely?
- Client data is processed in-memory on AWS Lambda and is not stored long-term by Syntora. The original audio file and resulting data are passed directly to your practice management software via encrypted API calls. We only log metadata (like a timestamp and success/fail status) in a Supabase database for monitoring. The system architecture is designed to be HIPAA compliant.
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