### Answer Great that you've started with FAQs! Here are specific ways to take your hotel AI voice agent to the next level and reduce unnecessary transfers: Advanced Capabilities for Hotel Voice Agents: Automated Reservations & Modifications [(1) ]() - Handle room bookings with real-time availability from your hotel database - Process reservation modifications and cancellations - Confirm and modify existing bookings with ease Guest Service Automation - Process room service requests, spa reservations, and additional services - Handle purchases and upgrades directly through the agent - Provide instant answers to common guest questions Smart Call Routing - Direct calls by urgency to the right department - Escalate to human staff only when truly needed - Use automated call routing to ensure guests reach VIP assistance or special services when appropriate Reducing Unnecessary Transfers: Use Conversation Flow to create structured pathways that guide conversations logically [(2) ]() . This feature helps mitigate AI hallucinations by defining specific interactions at each node . Node-Based Structure - Break conversations into nodes that define specific interactions or responses - Set conditional transitions between nodes to ensure appropriate responses based on previous inputs Enhanced Control - Map out potential conversation paths in advance to reduce ambiguity and improve accuracy - Define precisely what words are spoken by the agent to maintain brand voice consistency Post-Call Optimization: Implement Post-Call Analysis to continuously improve your agent [(3) ]() : - Track sentiment and ta

sk completion status - Extract custom metrics (Boolean, text, number, selector) - Identify patterns where transfers occur to refine your flows Training Your Agent: Use fine-tuning with call transcripts from your hotel's real conversations [(2) ]() . Train the AI on thousands of actual guest interactions including appointment scheduling, service inquiries, and billing questions . Apply prompt engineering to guide specific behaviors without full retraining . Craft prompts that instruct the AI how to structure responses and handle different scenarios . The key is designing workflows where the agent can handle 80% of common requests autonomously, only escalating complex or sensitive issues that truly require human intervention.

​ To learn more, see the sources I considered: [[1] Future of Hospitality with AI Voice Agents | Retell AI | Retell AI](https://www.retellai.com/blog/future-of-hospitality-with-ai-voice-agents) [[2] Training & Customizing AI Voice Agents | Retell AI](https://www.retellai.com/blog/training-and-customizing-voice-agents-with-retell-ai) [[3] Post Call Analysis Overview](https://docs.retellai.com/features/post-call-analysis-overview) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.

@AI Evy how do you perfectly build those

### Answer Based on the context from your previous question about building advanced hotel voice agents, here's how to build those capabilities: ## Building Conversation Flow Agents Configure Global Settings [(1) ]() Click on the empty canvas and access global settings to configure: 1. Voice Settings - Select and customize the voice, adjusting temperature, speed, volume, and model 2. Language Selection - Pick the language(s) your agent will understand and speak 3. Select a Language Model - Choose your LLM (GPT-4.1 is recommended for optimal balance of response quality, latency, and cost-effectiveness) 4. Write Global Prompt - Specify the agent's persona, identity, and guardrails that will be available in every node 5. Configure Knowledge Base - Supply context to your agent via documents, URLs, and texts 6. Configure Speech Settings including responsiveness, interruption sensitivity, backchanneling, and boosted keywords Design Conversation Pathways [(2) ]() Use the Conversation Flow feature with a node-based structure : - Each node represents a specific point where the agent can respond or ask a question - Define transition logic that determines how conversations progress from one node to another based on user input or predefined conditions - Design workflows that allow agents to transition to specific nodes based on triggers or commands Training Your Agent The training process involves two steps: 1. Fine-Tuning the LLM with Call Transcripts - Train on thousands of real conversations to improve understanding of customer queries and conversational flow 2. Prompt Engineering for Specific Behaviors - Craft specific prompts to guide the AI's behavior in

different contexts, telling the AI how to structure responses and what tone to use Implement Post-Call Analysis [(3) ]() Set up post-call analysis to extract: - Boolean (True/False) determinations - Text (String) detailed information - Number (Numerical value) quantitative measurements - Selector (Enum) categorization from a fixed list This allows you to continuously improve your agent by tracking what works and what doesn't .

​ To learn more, see the sources I considered: [[1] Step 1: Configure global settings](https://docs.retellai.com/build/conversation-flow/global-setting) [[2] Training & Customizing AI Voice Agents | Retell AI](https://www.retellai.com/blog/training-and-customizing-voice-agents-with-retell-ai) [[3] Post Call Analysis Overview](https://docs.retellai.com/features/post-call-analysis-overview) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.

### Answer To connect your Retell AI agent to Choice Advantage PMS and enable room bookings, you'll need to use Custom Functions to integrate with your PMS API [(1) ]() . How Function Calling Works: Function calling enables your agent to interact with external systems like your PMS . The process works as follows : 1. Agent Decision - Based on the conversation context, the LLM determines when a function is needed 2. Parameter Extraction - The agent extracts required information from the conversation 3. Function Execution - Retell calls your function with the extracted parameters 4. Response Handling - The function result is processed and incorporated into the conversation Implementation Steps: 1. Create a Custom Function In your agent's Functions section, add a Custom Function that connects to Choice Advantage's API . You'll need to: - Define the function name (e.g., "book_room") - Set up parameters like check-in date, check-out date, room type, guest name - Configure your Choice Advantage API endpoint URL - Add authentication headers for your PMS 2. Alternative: Use Code Tool for Simpler Integrations [(2) ]() If you need lightweight logic or simple API calls, you can use the Code Tool to execute JavaScript directly :

javascript
// Example: Make API call to PMS
const response = await fetch("<https://api.example.com/submit%22,> {
 method: "POST",
 headers: { "Content-Type": "application/json" },
 body: JSON.stringify({ name: dv.customer_name })
}) ;

3. Pass Dynamic Variables [(3) ]() When creating calls via API, inject guest information as dynamic variables :

{
  "agent_id": "YOUR_RETELL_AGENT_ID",
  "to": "{{phone}}",
  "from": "+1XXXXXXXXXX",
  "variables": {
    "first_name": "{{first_name}}",
    "check_in_date": "{{check_in_date}}",
    "room_type": "{{room_type}}"
  }
}

4. Extract Booking Details [(4) ]() Use the Extract Dynamic Variables tool to capture guest information during the conversation : - Guest name (Text) - Phone number (Text) - Check-in/check-out dates (Text) - Room preferences (Enum) Important Notes: - Custom Functions support GET, POST, PUT, PATCH, and DELETE requests with custom headers and query parameters [(5) ]() - You can add custom headers for API authentication - The Code Tool is limited to 5,000 characters and doesn't support external packages [(2) ]() - For secure integrations requiring authentication or write access to production systems, use Custom Functions hosted on your backend rather than Code Tool You'll need to work with Choice Advantage's API documentation to determine the exact endpoints and authentication method required for room bookings.

​ To learn more, see the sources I considered: [[1] Function Calling Overview](https://docs.retellai.com/build/single-multi-prompt/function-calling) [[2] Code Tool](https://docs.retellai.com/build/single-multi-prompt/code-tool) [[3] GHL + Make + Retell](https://community.retellai.com/t/ghl-make-retell/806) [[4] Extract Dynamic Variables](https://docs.retellai.com/build/single-multi-prompt/extract-dv) [[5] Platform Changelogs | Retell AI](https://www.retellai.com/changelog) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.