📲Inbound Phone Calls
Learn how to set up AI voice agents to handle inbound phone calls
How to Create a New Voice AI Agent for Inbound Phone Calls
Get a Phone Number
Purchase a phone number from Stammer AI.
Prices vary depending on which country the phone numbers reside in.
Example: USA phone numbers start at $1.75 per month.
Another option is to connect your Twilio account with existing phone numbers.
Connect a Voice Agent
Set up the AI agent settings by selecting a AI model, language, voice, first sentence and adding a prompt with instructions on how the voice agent should act during the conversation.
Set up the call settings by adding the optional background audio, setting a max call duration, enable the option to record the phone call, setting the timezone and which phone call to transfer to.
Adding a Knowledge Base (optional)
AI voice agents work best when all relevant information has been added to the prompt. In cases where you have a lot of additional information, you can add data into the knowledge base by uploading files and scraping websites.
Connecting Custom Tools
Create connections to 3rd party tools like Shopify, CRM, Scheduling tools and more.
Review the Analytics
With each phone call you can summarize the conversation, extract lead data, and send a full transcript of the call to any email address or webhook.
AI Configurator
Model
This setting controls which AI model powers your voice agent. Each model has different pricing and performance levels — for example, the Primary model is GPT-4.1 ($0.2/min) and offers the most powerful model for conversational quality.
Primary model: Ideal for most business calls.
Think of this as choosing the “engine” behind your AI voice.
Language
Determines which languages the agent can understand and respond in.
The “Multi” option allows the agent to automatically detect and respond in multiple languages during a call.
You can also choose a specific language (e.g., English, Spanish, French) if your business only serves one region.
This ensures your AI agent communicates clearly with your audience.
Voice
Here, you select the AI voice persona that callers will hear. Example: Paul (Male) Each voice has a unique tone, accent, and delivery style. You can preview different voices to find one that best fits your brand — friendly, professional, energetic, etc.
Temperature
This slider adjusts the creativity and variability of the agent’s responses.
Lower (0–0.3): More consistent and factual — best for customer support or compliance-heavy industries.
Medium (0.5): Balanced tone — natural, conversational, but still reliable.
Higher (0.7+): More flexible or creative — good for casual, personality-driven calls.
For most business calls, 0.5 is the sweet spot.
First Sentence
This is the initial greeting that plays when someone calls in. Example:
“Hey there, I'm an AI agent for Stammer AI, how can I help you out today?”
You can edit this to reflect your business’s tone and name. It sets the tone for the conversation and gives the caller immediate clarity that they’re speaking with your company’s virtual assistant.
Prompt
The Prompt is the “brain” of your voice agent — it defines how the agent behaves, what it says, and how it handles calls.
It usually includes two main sections:
Customization Instructions
This is a guide for you (the user or agency) on how to personalize the prompt before going live. It includes steps like:
Replacing placeholders (e.g.,
[COMPANY NAME],[PRODUCT/SERVICE DESCRIPTION])Customizing benefits and scripts to your business
Adding legal disclaimers or policies
Removing instructional text before finalizing
This ensures your AI sounds customized to your brand, not generic.
b. Generic Voice Agent Script
This is the core logic your agent follows during calls. For example:
Greeting the caller
Collecting key information (name, phone, reason for call)
Routing or scheduling calls
Answering basic sales or service questions
Handling objections or ending the call professionally
It acts like a template for inbound sales or support calls that you can modify per client or industry.
Call Settings
Background Audio
Adds subtle ambient noise behind your AI’s voice — like a Coffee Shop, Summer Outdoor, or Static Noise. This makes the AI sound more natural and human by simulating a real environment.
Example: “Coffee Shop” adds soft chatter and background hum to make it feel like a live conversation. Use this when you want your agent to sound more personable or casual.
Max Duration (Minutes)
Sets the maximum length for each call. Once this time limit is reached, the AI will automatically end the conversation politely.
Example: If set to 10 minutes, the agent will wrap up any ongoing conversation at that point. This helps manage costs and prevent long, unnecessary calls.
Record
Enables or disables call recording. When switched on, every conversation is saved and can be reviewed later for:
Quality assurance
Lead verification
Training and compliance
Note: Always make sure call recording complies with your local laws or regulations — some regions require consent from all parties.
Ignore Button Press
Toggles whether the AI should ignore DTMF tones (the sounds made when callers press phone keypad buttons).
When off, the agent can detect button inputs — useful for menu navigation or call routing.
When on, it ignores button tones entirely — better for simple voice-only conversations.
Timezone
Specifies the time zone your AI agent operates in.
Used for time-sensitive actions like appointment scheduling, callbacks, or timestamped reports.
Example: -5:00 America/New York aligns call logs and schedules with Eastern Time.
Call Transfer Settings
This section defines how the AI hands off calls to a real person or another line.
Enter the phone number where calls should be transferred when certain conditions are met — for example:
When a caller requests to speak to a human.
When the AI identifies a qualified lead.
When an issue requires escalation.
The AI will connect the caller directly to this number in real time.
Example format: +1234567890 (include country code).
Knowledge Base for Voice Agents (Optional)
The Knowledge Base is where your AI voice agent learns about your company, products, and services in more detail. While the main prompt gives your agent general behavior and tone, the knowledge base provides deeper context — allowing it to answer more specific or technical questions during calls.
When callers ask detailed questions (like “What’s your refund policy?” or “What are your business hours?”), the AI pulls accurate answers directly from your knowledge base data.
This is especially useful for:
Companies with detailed product catalogs
Agencies serving multiple clients
Businesses with FAQs, manuals, or documentation
How It Works
The AI automatically references your uploaded or entered data during conversations. It doesn’t read the data word-for-word — instead, it learns the context and retrieves the most relevant parts when needed.
Knowledge Base Tabs
You can add data through four different tabs:
1. Text
Manually paste or type any text you want your agent to learn from. This can include:
FAQ lists
Product descriptions
Company overviews
Policy details
Support procedures
Example:
“We offer free returns within 30 days of purchase. Customers can initiate a return by visiting our website or calling our support line.”
Once entered, click Train Chat Agent to include that text in your AI’s memory.
2. Files
Upload documents that contain your business information.
Supported formats typically include .pdf, .docx, .txt, or .csv.
Ideal for:
Service guides
Internal documentation
Compliance policies
Sales scripts
After uploading, the system automatically processes and indexes the file so your agent can use it in real time.
3. Websites
Input URLs for your business’s web pages. The AI will read and understand the page content to use during calls.
Use this for:
Product pages
Pricing or policy pages
Blog posts with helpful information
This ensures your AI always has up-to-date details pulled directly from your online presence.
Training Your Agent
After adding your text, files, or URLs:
Review the Trained Characters counter to see how much data has been processed.
Click Train Chat Agent to finalize and apply the new knowledge.
Your voice agent will now be able to reference this data in live calls.
Best Practices
Keep responses concise — prioritize clear, factual info over long documents.
Regularly update your knowledge base when company info changes.
Include FAQs and real customer inquiries for better training data.
Avoid confidential or sensitive information, since the AI uses this data during live conversations.
Custom Tools for Voice Agents
The Custom Tools section lets you extend your AI voice agent’s capabilities beyond conversation. These tools allow the agent to perform real-world actions — like scheduling an appointment, or fetching data from an external system — all during a live call.
Essentially, this is where your AI connects to APIs and triggers business workflows.
Purpose
Use Custom Tools to make your agent more than just a talker. Instead of saying, “I’ll send that confirmation to your email,” it can actually trigger the API call to send it automatically.
Examples of what Custom Tools can do:
Send appointment confirmation texts or emails
Pull real-time order or booking details
Create new contacts in your CRM
Submit lead info to Google Sheets or GoHighLevel
Trigger webhooks to start automations
How It Works
Each Custom Tool acts as a bridge between your voice agent and an API endpoint. You define what the tool does, what information it needs, and how it communicates with your external service.
During a call, if the AI detects a trigger phrase like “schedule me in for tomorrow” or “send me the quote,” it can automatically call that API tool.
Adding a New Tool
Click “+ Add New Tool” to open the configuration panel. Here’s what each field means:
Name
Give your tool a clear and descriptive name. Example:
“Send Confirmation Text” or “Create New Lead”
This name is used to identify the tool inside the platform and may be referenced in your prompts.
Description
Briefly describe what the tool does. Example:
“Sends an SMS confirmation message to the customer after booking an appointment.”
Helps you and your team remember what each integration is for.
Speech
This is what the voice agent will say to the caller while the tool is running. Example:
“Please wait while I process your request…” You can customize this to fit your brand’s tone: “Got it, I’m sending that over for you now.”
API Endpoint
Enter the full API URL that the tool should call. Example:
https://api.example.com/send-confirmation-text
Choose the appropriate HTTP method (GET, POST, PUT, DELETE). Most integrations use POST for sending data or GET for retrieving data.
Timeout (ms)
Defines how long the system should wait for a response before stopping the request. Example:
10000means the system will wait up to 10 seconds before timing out.
Adjust this depending on how fast your API responds.
Headers
Add any necessary request headers, such as:
Authorization: Bearer YOUR_API_KEYContent-Type: application/json
Click + Add Header to include as many as you need.
Query Parameters
If your API requires query parameters in the URL (e.g., ?user_id=12345), you can add them here using + Add Parameter.
Input Schema
Defines what data the agent needs to send to the API. This must be in valid JSON format.
Example Schema:
{
"example": {
"tracking_id": "123456",
"name": "John Doe"
},
"type": "object",
"properties": {
"tracking_id": "number",
"name": "string"
},
"required": ["tracking_id"]
}This schema tells the AI what information to collect from the user (in this case, a tracking ID and a name) before calling the API.
When to Use Custom Tools
Use them anytime your AI needs to:
Trigger an automation
Fetch or update external data
Complete an action in another app or database
They’re especially powerful for:
CRM and scheduling integrations
E-commerce order tracking
SMS or email notifications
Internal business tools
Best Practices
Always test your API endpoint before going live.
Use descriptive names and speech responses for clarity.
Ensure API responses return valid JSON.
Secure your API keys in headers, not in URLs.
Keep timeouts reasonable (5–10 seconds max).
Analytics for Voice Agents
The Analytics section gives you insights into how your AI voice agent is performing after each call. It handles everything from automatic call summaries to lead data collection and integrations with your CRM or email system.
This is where all post-call automation and reporting are configured.
1. AI Summary Prompt
This setting defines how the AI processes and summarizes each call once it ends. The summary can be used for review, analytics, or lead follow-up.
Summary Prompt
Type an instruction telling the AI how to summarize conversations. Example:
“Summarize the call in a few sentences.”
You can customize this to generate more detailed summaries depending on your use case:
Sales: “Summarize the caller’s main interest, budget, and next steps.”
Support: “Summarize the issue reported, troubleshooting steps, and resolution.”
Scheduling: “Summarize the booking details, requested date, and any follow-up actions.”
The AI then automatically generates a written summary based on the full transcript of the call.
2. Lead Collection Fields
This section lets you define what caller information your AI should collect and store. You can create custom fields for names, phone numbers, emails, or any other details needed for your workflow.
Example Fields:
name → String
phone_number → String
You can also click + Add Field to include more data points such as:
email
company_name
service_requested
preferred_date
Note: Make sure your AI prompt includes instructions telling the agent when and how to collect these fields. For example: “Before ending the call, confirm the caller’s name and phone number.”
These collected values are then included in call summaries, webhook payloads, and notification emails.
3. Webhook Integration
Webhooks allow Stammer to automatically send call data to external systems — like CRMs, databases, or automation tools (e.g., Zapier, GoHighLevel, Airtable, or custom APIs).
Enable Webhook Integration
Toggle this on to activate it. Once enabled, the system will push structured call data — including the summary and collected fields — to the webhook endpoint you’ve configured.
Use cases:
Add new leads to your CRM
Trigger follow-up emails or SMS via automations
Log call data for analytics dashboards
4. Email Notifications
If you prefer to receive updates by email, you can turn on this feature to get a summary of each call directly in your inbox.
Enable Email Notifications
When toggled on, Stammer sends automatic emails containing:
The AI-generated call summary
Caller name and contact details
Any additional lead fields collected
Perfect for small teams who want real-time visibility without logging into the dashboard.
Best Practices
Keep summaries short but detailed enough for quick review.
Add all relevant lead fields to streamline follow-up automation.
If using both Webhook and Email, test each to ensure data passes correctly.
Regularly review summaries to refine your AI prompts for better accuracy.
Last updated