Overview

Answering Machine Detection (AMD) is a feature that allows Exotel’s system to distinguish whether an outbound call was answered by a Human or a Machine (such as a voicemail or IVR system).

By analyzing audio patterns in real-time, AMD enables businesses to make intelligent decisions on how to handle the call workflow—such as connecting to an agent immediately if a human answers, or dropping the call/leaving a message if a machine answers.

Note: This feature is currently in Beta. To enable this for your account, please contact hello@exotel.com.

Supported APIs

This feature is currently supported on the following APIs:

  1. Connect 2 Numbers API: (For the second leg of the call)

  2. Get Call Details API: (For post-call analysis)

How It Works

Exotel’s AMD engine analyzes the audio stream immediately after the call is answered. It uses specific acoustic markers to classify the recipient.

Key Logic Parameters:

  • Initial Silence: Detects if the recipient is silent for too long (indicative of a machine waiting for a recording trigger).

  • Greeting Duration: Detects if the greeting is unusually long.

  • Word Count & Length: Analyzes the number of words spoken in the greeting. High word counts often indicate a voicemail greeting (e.g., "Thank you for calling, we are currently unavailable...").

  • Post-Greeting Silence: Detects the natural pause humans make after saying "Hello?"


Classification Outcomes


Based on the logic above, the system classifies the call outcome into one of three statuses:

Status

Description

HUMAN

The system detected a short greeting followed by a pause, indicating a live person.

MACHINE

The system detected a long greeting, high word count, or specific silence patterns indicating voicemail/IVR.

NOT_SURE

The system could not confidently determine the entity (e.g., due to background noise or ambiguous audio).


Accuracy Note: Our AMD logic is benchmarked at approximately 80% accuracy. Please factor this margin into your workflow logic.

API Documentation & Implementation

For full API specifications, response parameters, and integration details, please refer to our developer documentation:

If you have questions regarding the logic or wish to provide feedback on the Beta performance, please reach out to hello@exotel.com.