How can we help? 👋

Voice AI: DTMF Support

Voice AI understands natural language very well, but there are cases where using a keypad makes more sense. Typing an account number, PIN, or code reference is more convenient and efficient, and secure because the customer doesn't have to speak the numbers aloud. DTMF allows you to accept numeric keypad input from your customers during conversation with voice AI bots. The customer enters the numbers from their phone, and then your bot receives those digits and processes them as a customer message in the transcript.


📎 How it Works

  • Available in Speech-to-Text-to-Speech mode only.
  • When a customer presses keys (by phone or in the preview tool), the digits are passed to the bot as if the customer had said them.
  • The bot waits for a short pause (around 3–4 seconds) between presses before treating the sequence as complete. Customers can enter as many digits as they like, as long as they don't pause longer than that gap.
  • If a customer speaks and presses keys at the same time, the bot finishes listening to speech before handling the digits.

🛠️ How to Use it

  1. DTMF is already enabled for Voice AI bots in Speech-to-Text-to-Speech mode — there's nothing to switch on.
  1. In your bot's prompt, tell it what to do when the customer presses digits (e.g. "If the customer enters their account number, look it up using..."). The bot will be told something like "the customer pressed 1379" — your prompt decides what happens next.
  1. Test it using the dial pad in the Voice AI browser testing tool.
💬

Keypad entries show up in transcripts as a customer message noting which numbers were pressed.

💡

DTMF is not currently supported in Speech-to-Speech mode.

 

❓Frequently Asked Questions

When should I use DTMF over voice input?

DTMF is ideal for more structured data. Things like account numbers, PINs, reference codes, and phone numbers. Where spoken numbers could be misinterpreted by the bot.

How does keypad input appear in transcripts?

Digits pressed by the customer are captured and displayed as a customer message in the interaction transcript.

Did this answer your question?
😞
😐
🤩