This AI Voice Agent Answers All Your Calls (+ free blueprints)

Key Concepts

Voice AI agent, Vapi, N8N, Twilio, System Prompt, Tools (Invoice, Scheduling), Webhooks, Go High Level, Helper Agent, JSON, Stringify, Google Calendar integration, AI Automation Agency.

1. Introduction: Voice AI Receptionist

The video introduces a voice AI agent receptionist built using Vapi that can answer incoming calls, answer client questions, book/cancel/update appointments in real-time, and send invoices. The blueprints are available for free download.

• Example: The video demonstrates the AI assistant, named Elliot, answering a call, explaining AI services, sending an invoice, and scheduling a consultation.

2. Testing the AI Assistant

A demonstration showcases Elliot's capabilities:

• Answering questions about AI services.
• Collecting name and email for invoice generation.
• Scheduling a consultation for the next day at 11:00 AM.
• Sending a calendar invite.

Quote: "Hey you know welcome to ABC company my name is Elliot how can I help you today?" - Example of a first message from the AI agent.

3. Important Considerations

• The AI is not perfect and can be clunky, potentially due to the system prompt.
• AI calling requires consent in most jurisdictions.
• The technology is in its infancy with room for growth.
• Pros: Infinitely scalable, no constant training, cheap.
• Cons: Potential for errors, may lower customer trust.

4. Building the AI Agent: Vapi Setup

The process involves setting up Vapi with a Twilio phone number and creating an assistant.

• Step 1: Sign up for a free Vapi account.
• Step 2: Create an assistant (e.g., "Elliot").
• Step 3: Create two tools: "Smart Invoice" and "Smart Scheduler."
• Step 4: Get a phone number. It's recommended to use Twilio for a free phone number.
  • Sign up for a free Twilio account at twilio.com.
  • Copy the phone number, Account SID, and Auth Token from Twilio.
  • Import the Twilio phone number into Vapi.
• Step 5: Generate a system prompt and a first message for the assistant.

5. System Prompt and First Message

• The first message is what the AI says when answering the phone.
• The system prompt provides context and information for the AI to succeed.
• The system prompt includes:
  • Responsibilities (appointment management, answering FAQs, invoice handling, forwarding calls).
  • Instructions on using tools.
  • Categorization of user requests (create, delete, update, search appointments).
  • Conversation style (tone, communication).
  • Example conversations.
• A PDF file containing frequently asked questions is used to provide the AI with information about the company.

6. Vapi Settings

• Model: The video suggests using the "4o mini" or "4.1 mini" model.
• Tokens: A context window of 1000 tokens is recommended.
• Temperature: A temperature of 0.7 is suggested.

7. Connecting Vapi to N8N: Tools

The core of the video focuses on connecting Vapi to N8N to extend the AI agent's functionality. This is done using "Tools" within Vapi.

• Two tools are created: "Smart Invoice" and "Smart Scheduler."
• Tools are defined as "functions."

8. Smart Invoice Tool

• Purpose: Sends invoices to customers.
• Parameters: Name and email of the user (defined in Vapi and passed to N8N).
• Server Settings: The N8N webhook URL is placed in the server settings of the tool.
• N8N Workflow:
  • A webhook node receives data from Vapi.
  • An HTTP request node generates the invoice using Go High Level.
  • A "Respond to Webhook" node sends a success or failure message back to Vapi.
• Go High Level: Used as an invoice generator.

9. Smart Scheduler Tool

• Purpose: Books, checks availability, cancels, and updates appointments.
• Data Sent to N8N: Category (type of action requested), name, email, day, time, original day/time (if updating).
• N8N Workflow:
  • A webhook node receives data from Vapi.
  • A "Helper Agent" node processes and formats the data.
  • A "Calendar Agent" node interacts with Google Calendar.
• Helper Agent:
  • Converts JSON data to text using "stringify."
  • Corrects email formatting.
  • Standardizes time zones and date formats.
• Calendar Agent:
  • Uses available tools (book, search, cancel, update).
  • Defines the order of operations (e.g., search before creating).
  • Integrates with Google Calendar nodes.
  • Sends a response back to Vapi.

10. Helper Agent Details

• The helper agent takes unstructured data and formats it properly for the calendar agent.
• It corrects common errors, such as misinterpreting the email address.
• It standardizes time formats to work with N8N and Google Calendar.

11. Calendar Agent Details

• The calendar agent uses a system message to define available tools (book, search, cancel, update) and rules for using them.
• It defines the order of operations, such as searching for conflicts before creating an appointment.
• It integrates with Google Calendar nodes to take the appropriate action.

12. Real-Time Interaction

The video emphasizes that all these tools work in real-time while the voice agent is on the phone with the customer. This creates a human-like experience.

13. Conclusion

The video concludes by highlighting the value of the system and encouraging viewers to build something similar. The blueprints are provided for free. The speaker also promotes his school community for those interested in AI automation and agency creation.

Quote: "Hey your appointment has been successfully completed on this day" - Example of a response sent back to Vapi after booking an appointment.

14. School Community

The speaker promotes his school community, which focuses on:

• Automating up to 80% of business operations.
• Creating an AI automation agency.
• Providing a thriving community with weekly calls.