Configure the DTMF Node in an AI Agent
Updated
The DTMF Response - Voice node enables an AI Agent to collect caller input during voice conversations using keypad entries, spoken responses, or both.
DTMF (Dual-Tone Multi-Frequency) signals are generated when a caller presses keys on a telephone keypad. Use this node when callers need to select menu options, enter numeric values, or provide information that is more reliably captured through keypad input than speech.
Depending on the configuration, callers can:
- Press 1 to check an order status.
- Press 2 to speak with support.
- Enter a 10-digit mobile number.
- Enter an order ID followed by the # key.
- Provide an OTP or verification code.
The node captures the response, validates the input, and routes the conversation to the appropriate next step.
Note: The DTMF node uses the voice configured for the AI Agent. Prompts are not played using the voice provider's default voice.
Prerequisites
Before adding a DTMF node, ensure that:
- Enable Voice was selected when the application was created.
- A language is configured in the Dialogue Tree settings.
- A valid voice configuration is available for the AI Agent.
- The dialogue flow includes paths for valid responses, invalid responses, timeouts, and fallback handling.
When to Use the DTMF Node
Menu-Based Navigation
Use keypad selections to route callers to different conversation paths.
Example prompt
Press 1 for order status. Press 2 to speak with customer support.
The AI Agent routes the caller based on the selected digit.
Fixed-Length Numeric Input
Collect a predefined number of digits, such as:
- Mobile numbers
- Account numbers
- Verification codes
- OTPs
Example prompt
Please enter your 10-digit registered mobile number.
The AI Agent waits until the configured number of digits is entered before proceeding.
Variable-Length Input
Collect values whose length can vary.
Common examples include:
- Order IDs
- Customer IDs
- Reference numbers
Example prompt
Please enter your order ID followed by the hash key.
The AI Agent captures input until the configured end key is pressed.
Sensitive Data Collection
Use the node to securely collect sensitive information when the required data protection configuration is available.
Examples include:
- Card numbers
- CVV values
- Card PINs
- OTPs
- Identification numbers
- Dates of birth
Add a DTMF Node
Add the DTMF node at the point in the conversation where caller input is required.
Steps
- Open Tasks.
- Click the arrow next to + Add Task and select Create Dialogue Tree.
- Enter a task name and description.
- Click Save.
- In the Dialogue Tree Builder, click + and select DTMF Response - Voice.
Configure the DTMF Node
Basic Configuration
Enter the following details:
- Name: Enter a node name.
- Languages: Select supported languages.
Response Type
Select how callers can provide responses.
Response Type | Description |
DTMF | Accepts keypad input only. |
Speech | Accepts spoken input only. |
DTMF and Speech | Allows callers to respond using either keypad input or speech. |
Recommendations
- Use DTMF for menu selections and numeric input.
- Use DTMF and Speech when callers should be able to choose between speaking and pressing keys.
Note: Speech input is supported when using the Sprinklr Voice Connect provider.
Number of Input Digits
Specify how many digits the caller must enter.
Use Case | Digits |
Menu selection | 1 |
Four-digit OTP | 4 |
Mobile number | 10 |
After the specified number of digits is received, the AI Agent processes the input and continues the conversation.
Configure Valid Digits
When Number of Input Digits is set to 1, configure the accepted keypad values.
Example
Prompt:
Press 1 for Returns. Press 2 for Support.
Valid digits:
- 1
- 2
Create an output path for each configured digit.
Configure Input Variables
For inputs longer than one digit, use an input variable instead of valid digits.
The entered value is stored and can be used later in the flow.
Example
orderIdmobileNumbercustomerIdotpValue
The stored value can be used to:
- Call an API.
- Retrieve order details.
- Validate user input.
- Confirm information back to the caller.
Variable Naming Guidelines
Variable names:
- Must not start with a number.
- Must not start with a special character.
- Must not contain spaces.
Valid examples:
orderIdmobileNumbercustomerIdotpValue
Speech Input Settings
Configure these settings when Speech or DTMF and Speech is selected.
Setting | Description |
Speech Model | Defines how spoken input is processed. |
Speech Variable | Stores the caller's spoken response. |
Confidence Tolerance | Defines the acceptable confidence threshold for speech recognition. |
Profanity Filter | Filters inappropriate language from transcripts and responses. |
Configure the Prompt
Use Operator Says to define the message played before input is collected.
The prompt should clearly explain:
- What information is required.
- Whether the caller should speak or press a key.
- Available options.
- Whether an end key is required.
Example Prompts
Press 1 to check your order status. Press 2 to speak with support.
Please enter your order ID followed by the hash key.
Please enter your 10-digit registered mobile number.
Additional Settings
Enable Recording
Record caller interactions at this node for:
- Quality monitoring
- Auditing
- Troubleshooting
Sensitive Data
Enable this option when collecting protected customer information.
PCI Data
Use when collecting payment-related information such as:
- Credit card numbers
- Debit card numbers
- CVV
- Card PIN
- OTP
PCI data is securely processed through the configured Cardholder Data Environment (CDE).
PII Data
Use when collecting personally identifiable information such as:
- Date of birth
- Government identification numbers
- Social Security numbers
PII values are masked in the user interface while remaining available to backend systems according to configured data-handling policies.
Wait for End of Prompt
Enable this option to prevent callers from entering input while the prompt is still playing.
Input is accepted only after the prompt finishes.
Timeout
Specify how long the AI Agent should wait for a response.
If no response is received within the configured duration, the timeout path is triggered.
End Key
Specify the key that indicates input completion for variable-length values.
Retry Count
Specify how many times the AI Agent should repeat the prompt when no valid response is received.
For example, a value of 2 causes the prompt to play once initially and repeat up to two additional times.
Configure Next Paths
Configure routing for:
- Valid keypad input
- Valid multi-digit input
- Valid speech input
- Invalid input
- Timeout
- Default fallback path
Example Menu
Prompt:
Press 1 for order status. Press 2 for customer support.
Configure the following paths:
Input | Action |
1 | Retrieve order status |
2 | Transfer to customer support |
Invalid Input | Play invalid input message |
Timeout | Play timeout message |
Default Path | End call or transfer to support |
After configuration, connect each output path to the appropriate node, such as a bot response, API call, agent handoff, or conversation end node.
Example: Order Status Menu
Field | Example Value |
Response Type | DTMF |
Number of Input Digits | 1 |
Valid Digits | 1, 2 |
Operator Says | Press 1 to check order status. Press 2 to speak with support. |
Timeout | 10 seconds |
Retry Count | 2 |
Default Path | Enabled |
Output Paths
- 1 → Order Status
- 2 → Customer Support
- Invalid Input → Retry message
- Timeout → Timeout message
- Default Path → End call or transfer to support