Rate this page:

How to Use AI to Detect Vocemail

There is a number of different scenarios when you would like to know if you've reached voicemail or live person during the call, and if voicemail was reached it's a good thing to know when you can play a pre-recorded voice message to leave it in the inbox. Keeping in mind that Voximplant is a perfect choice for building power dialer or predictive dialer app it makes a lot of sense to offer this functionality to our developers. We've substantially updated our voicemail detection using the machine learning approach. The AI-based voicemail is highly precise in detecting pre-recorded audio messages which allows to handle reaching a voicemail or answering machine.

To have an access to this functionality, you should use the AI module in scenarios (see the further examples).

Power dialer

Let's start with a simpler scenario – agent calls customers using some call list and while he reaches voicemail he sends some pre-recorded mp3 message to the inbox. In terms of VoxEngine scenario this code will look like as follows:

Power dialer

Power dialer

Predictive Dialer

In case of predictive dialing, we don't rely on a live person to detect voicemail, since it's an automated process and we don't want to spend call center agents' time. Since voicemail systems behave differently it's not possible to offer 100% warranty in case of automated voicemail detection, but we made our best and implemented number of adjustable subsystems that help with the process. We will use the same detectVoicemail function to analyze audio stream in both pre-connected and connected states of a call. The handlers themselves will be the same as in the Power dialer example:

Predictive Dialer

Predictive Dialer

So when voicemail is reached, the corresponding variable will be set to true, otherwise, it will be set to false. A developer can make a decision depending on the data while forwarding a call to a free operator using ACD and CallList modules for predictive dialing.