Generate AI voice receptionist agents for local businesses using VAPI
Automate the creation of personalized AI phone receptionists for local businesses by scraping Google Maps, analyzing websites, and deploying voice agents to VAPI.
Who is this for?
- Agencies offering AI voice solutions to local businesses
- Consultants helping SMBs modernize their phone systems
- Developers building lead generation tools for voice AI services
- Entrepreneurs launching AI receptionist services at scale
What this workflow does
This workflow automates the entire process of creating customized AI voice agents:
- Collects business criteria through a form (city, keywords, quantity)
- Scrapes Google Maps for matching local businesses using Apify
- Fetches and analyzes each business website
- Generates tailored voice agent prompts using Claude AI
- Automatically provisions voice assistants via VAPI API
- Logs all created agents to Google Sheets for tracking
The AI adapts prompts based on business type (salon, restaurant, dentist, spa) with appropriate tone, services, and booking workflows.
Setup requirements
- Apify account with Google Maps Scraper actor access
- Anthropic API key for prompt generation
- OpenRouter API key for website analysis
- VAPI account with API access
- Google Sheets connected via OAuth
How to set up
- Import the workflow template
- Add your Apify credentials to the scraping node
- Configure Anthropic and OpenRouter API keys
- Replace
YOUR_VAPI_API_KEY in the HTTP Request node header
- Connect your Google Sheets account
- Create a Google Sheet with columns: Business Name, Category, Address, Phone, Agent ID, Agent URL
- Update the Sheet URL in both Google Sheets nodes
- Activate the workflow and submit the form
Customization options
- Business templates: Edit the prompt in "Generate Agent Messages" to add new business categories
- Voice settings: Modify ElevenLabs voice parameters (stability, similarity boost)
- LLM model: Switch between GPT-4, Claude, or other models via OpenRouter
- Output format: Customize the results page HTML in the final Form node