Build an AI Powered Phone Agent - with Retell, Google Calendar, and RAG
This Workflow simulates an AI-powered phone agent with two main functions:
1. **Appointment Booking** - It can schedule appointments directly into Google Calendar.
2. **RAG-based Information Retrieval** - It provides answers using a Retrieval-Augmented Generation (RAG) system. For example, it can respond to questions such as store opening hours, return policies, or product details.
The guide also explains **how to purchase a dedicated phone number** (with a +1 prefix) and link it to the AI agent. This setup is cost-effective, as it uses a **FREE $10 credit** to operate without additional charges in the beginning.
---
### **Advantages**
- **24/7 Availability** - The AI agent can answer calls and assist customers at any time.
- **Automation** - It reduces the workload on human staff by handling repetitive tasks like appointment scheduling and FAQ responses.
- **Easy Integration** - Built with n8n, it's flexible and customizable for various platforms and tools.
- **Low-cost Setup** - Using the free credit, businesses can get started without an upfront investment.
---
### **Use Cases**
- **E-commerce** - Answer common product questions or order inquiries.
- **Retail Stores** - Provide store hours, address info, and return policies.
- **Restaurants** - Make reservations or share menu information.
- **Service Providers** - Book appointments or consultations.
- **Any Local Business** - Offer phone support without needing a live operator.
---
### **How It Works**
This Workflow simulates an AI-powered phone agent with two primary functions:
1. **Appointment Booking**
- The workflow captures call events (e.g., `call_ended` or `call_analyzed`) and extracts key details (transcript, caller info, duration, etc.).
- Using OpenAI, it summarizes the conversation and parses structured data (e.g., names, contact info, dates).
- For scheduling, it converts user-provided dates into Google Calendar-compatible formats and creates events automatically.
2. **RAG-Based Information Retrieval**
- When a query is received (e.g., store hours, product details), the workflow retrieves relevant information from a Qdrant vector store.
- An AI agent processes the query using the retrieved data and responds via a webhook, ensuring accurate, context-aware answers.
---
### **Set Up Steps**
1. **Prepare Qdrant Vector Store**
- Create/refresh a Qdrant collection (via HTTP requests).
- Upload and vectorize documents (e.g., from Google Drive) using OpenAI embeddings.
2. **Configure RetellAI Agent**
- Sign up for RetellAI, create an agent, and set the webhook URLs (`n8n_call` for call events, `n8n_rag_function` for RAG queries).
- Purchase a Twilio phone number and link it to the agent.
3. **n8n Workflow Setup**
- Connect OpenAI, Qdrant, Google Calendar, and Telegram nodes with credentials.
- Customize prompts for summarization, date parsing, and RAG responses.
- Test the workflow to ensure data flows from call events to processing to actions (e.g., calendar bookings, Telegram alerts).
4. **Deploy**
- Trigger the workflow via RetellAI webhooks during calls.
- Monitor outputs (e.g., call summaries in Telegram, calendar events).
**Note**: Replace placeholders (e.g., `QDRANT_URL`, `COLLECTION`, `CHAT_ID`) with actual values.
---
### **Need help customizing?**
[Contact me](mailto:info@n3w.it) for consulting and support or add me on [Linkedin](https://www.linkedin.com/in/davideboizza/).
n8nFree