How to Build a n8n Voice AI Agent That COLD CALLS Leads for You

Voice AI Agent for Upselling: A Step-by-Step Guide

Key Concepts:

Voice AI Agent: An automated system that makes phone calls to customers, powered by AI.
No-Code Automation: Building automated workflows without writing code, using platforms like Naden.
Retail AI: A voice platform used to create and manage voice AI agents.
Function Calling: Using dynamic variables or parameters to reach out to third-party APIs.
Webhooks: Automated HTTP requests triggered by events, used to receive data from Retail AI.
CRM Integration: Connecting the voice AI agent to a Customer Relationship Management system (e.g., Google Sheets, Airtable).

1. Introduction and Live Demo

The video demonstrates how to build a voice AI agent that automatically calls past customers to upsell them on new services. The agent is built using Naden (a no-code automation tool) and Retail AI (a voice platform). The video starts with a live demo where the AI agent, named Jarvis, calls the presenter and successfully books a free brake and tire inspection appointment. The call summary is then automatically updated in a Google Sheet and an email confirmation is sent to the customer.

2. Step-by-Step Build Process

The presenter walks through the process of building the voice AI agent from scratch using Naden and Retail AI.

2.1. Setting up the Naden Workflow:

Trigger: A manual trigger is used to initiate the workflow, but this can be replaced with a scheduled trigger for automated calling.
Google Sheets Integration: The workflow retrieves customer data (name, email, phone number, last visit date, details) from a Google Sheet. The "Get Rows" operation is used to fetch the data.
HTTP Request Node: An HTTP request node is used to communicate with the Retail AI API. The method is set to "POST," and the URL is obtained from the Retail AI platform.

2.2. Configuring Retail AI:

Agent Creation: A new voice agent is created in Retail AI, selecting the "Single Prompt Agent" option.
Prompt Design: The agent is given a persona (Jarvis, a virtual AI assistant for Apex Auto Garage) and instructions on how to interact with customers. Dynamic variables (name, email, last visited, details) are used to personalize the interaction.
- Example Prompt: "You are Jarvis, a virtual AI assistant for Apex Auto Garage based in San Francisco California. Caller information: {name}, {email}, {last_visited}, {details}. You are supposed to call returning customers that are previously visited the garage and you're supposed to offer these free brake and tire inspections that's available till July 20 31st."
Function Definition: Functions are added to the agent to enable specific actions:
- Transfer Call: Transfers the call to a human agent.
- End Call: Terminates the call.
- Check Availability: Integrates with cal.com to check appointment availability. The event ID from cal.com is used to specify the event type.
- Book Appointment: Books an appointment using cal.com.
LLM Selection: GPT-4.1 is selected as the large language model for the agent.
Voice Selection: A voice is selected for the agent (e.g., Brian).
Webhooks: A webhook URL from Naden is configured in Retail AI to receive call transcripts after each call.

2.3. Connecting Naden and Retail AI:

API Authentication: A generic credential type with custom authorization is used to authenticate with the Retail AI API. The API key is obtained from Retail AI settings and included in the authorization header.
Request Body: A JSON body is constructed with the necessary parameters for the Retail AI API, including the customer's phone number, name, email, last visit date, and the agent ID. Dynamic variables from the Google Sheet are used to populate the customer data.
Phone Number Configuration: A phone number is purchased from Retail AI (using Twilio or Telnyx) and associated with the agent. Identity verification is required to purchase a phone number.

2.4. Handling Call Transcripts and Updates:

Webhook Trigger: The Naden workflow is triggered by the webhook from Retail AI when a call ends.
Data Extraction: The workflow extracts the call transcript from the webhook data.
Summarization (Optional): The call transcript is summarized using a summarization tool.
Google Sheets Update: The call summary is updated in the Google Sheet in the "Upsell Call Summary" column.

3. Community and Monetization

The presenter emphasizes the growing popularity of voice AI agents and their potential for monetization. He mentions examples of community members who have successfully sold voice AI agents to clients. He promotes his community, which provides resources and support for building and selling AI automations, including:

A course on launching an AI agency.
Step-by-step tutorials on using Naden.
Daily calls for AI agency support and technical assistance.

4. Key Arguments and Perspectives

Voice AI agents are a valuable tool for businesses to upsell services to existing customers.
No-code platforms like Naden make it easy to build and deploy AI automations without programming experience.
Retail AI provides a user-friendly platform for creating and managing voice AI agents.
The combination of Naden and Retail AI enables robust integrations and complex workflows.
Monetizing AI automations is a viable business opportunity.

5. Conclusion

The video provides a comprehensive guide to building a voice AI agent for upselling using Naden and Retail AI. It covers the essential steps, from setting up the workflow to configuring the voice agent and integrating with external services. The presenter highlights the potential of voice AI agents for businesses and the opportunities for monetization in the AI automation space.