How AI Voice Assistants Made An Extra $20k/mo For Our Client | Vapi x Zapier Tutorial

Introduction

Introduction

In today’s video, we are diving deep into a tutorial on how to build an AI Voice Assistant that has helped one of our clients generate a surplus of $ 20,000 a month. Studies show that businesses miss 27% of calls they receive. This was a major problem our client faced. The voice assistant we’re building today is tailored for a property management business to handle all their inbound calls efficiently.

Review of Current Inbound Call Handling

Previously, during business hours, inbound calls were either handled immediately by an available Customer Service Representative (CSR) or went straight to voicemail if no one was available. After hours, all calls went directly to voicemail. The issue here is customers hate leaving voicemails, and businesses even more so hate checking them. Thus, we’re replacing the voicemail step with our AI Voice Assistant.

Voice Assistant Call Flow Structure

The inbound call flow structure of the voice assistant includes the following:

1. Greeting Message: Initial greeting and extraction of call reason.
2. Three Paths Based on Call Reason:
   • New Lease: Collect name, email, and property address of interest, and send leasing application link if requested.
   • Maintenance and Repairs: Collect property address and unit number, notify caller, and submit the maintenance request.
   • Monthly Bill Pay: Guide caller to access the payment portal and collect email to send the link if requested.

Building The Voice Assistant in Vapi

Initial Steps

1. We used Vapi to create a new voice assistant titled "Lenny Property Management."
2. The initial message was set: "Hello, thanks for calling Lenny Property Management. How can I help you today?"

ChatGPT for Knowledge Base

1. Using ChatGPT to create a business overview for the property management company.
2. Constructed a prompt based on Persona, Rules, and Steps to follow, ensuring the assistant understands and responds appropriately.

Vapi Configuration

1. Provider: Anthropic’s Cloud 3.5 Sonnent.
2. Transcriber: Deepgram, Model: Nova 2.
3. Voice: Polly (Eleven Labs preferred but Polly used here)
4. Advanced Settings: Enabled audio recording, server URL set for Zapier webhook, and input purposeful structured notes for call analysis.

Knowledge Base Upload

1. Created a comprehensive knowledge base in markdown format encompassing business overview, managed properties, amenities, and business information.
2. Uploaded the knowledge base as a PDF to Vapi.

Setting Up Automations in Zapier

1. Webhook Creation: Added webhook to Vapi’s server URL.
2. Testing: Created test calls to review captured end-of-call reports to build the automation structure.
3. Path Creation in Zapier:
   • New Lease Inquiry: Collect name, email, and property address, update call dashboard, and send application link.
   • Maintenance Requests: Collect property address and unit number, update call dashboard, and email maintenance team.
   • Monthly Bill Pay: Guide caller access payment portal, collect email to send payment link, and update the call dashboard.
4. Final Testing: Ensured each path functioned correctly by making test calls and checking emails received and call dashboard updates.

Conclusion

The AI Voice Assistant we built successfully managed inbound calls for a property management business, reducing missed opportunities and increasing monthly revenue by $ 20,000. If you’re interested in implementing an AI Voice Assistant in your business, consider scheduling a discovery call with us.

Keywords

• AI Voice Assistant
• Property Management
• Vapi
• Zapier
• Call Flow
• Automation

FAQ

Q: How does the AI Voice Assistant handle different call reasons?

A: The assistant extracts the call reason and follows one of three paths: New Lease, Maintenance and Repairs, or Monthly Bill Pay. Each path has specific steps for handling the caller’s needs.

Q: Which tools are used to create the AI Voice Assistant?

A: Vapi is used to build the voice assistant, while Zapier handles the post-call automations.

Q: What models and transcribers are preferred in this setup?

A: The setup uses Anthropic’s Cloud 3.5 Sonnent as the provider, Deepgram Nova 2 for transcribing, and typically Polly for the voice.

Q: How is the call dashboard updated?

A: After each call, details such as date, time, phone number, call summary, transcript, and call recording are automatically updated in a Google Sheet via Zapier.

Q: Can this AI Voice Assistant be customized for other businesses?

A: Yes, the assistant can be tailored to suit different businesses by adjusting call flow steps and incorporating relevant knowledge bases.