right  Talk To Us!

How To: Detect Live Speakers from Voicemail with 99% Accuracy

Have you been scratching your head recently over how to improve the efficiency of your contact centre while still complying with regulations? Aculab has been hard at work, developing a solution for this exact problem: AI-driven Answering Machine Detection (AI-AMD) allows contact centres around the world to save time and money with ease.

The problem with traditional AMD

Answering Machine Detection, or AMD - is a staple technology of the contact centre space. The premise is simple. Software will analyse the opening few seconds of a call to determine whether it was answered by a live speaker or their voicemail. While the technology seems simple enough, its performance has generally been less than stellar, often falling foul of regulations.

‘Silent calls’, where a live speaker picks up the phone but hears no response, can occur for a multitude of reasons. One is when the AMD fails to hear the initial ‘hello’ and thus continues to wait, silently, for it to occur. Another is when the AMD mis-detects a live speaker as voicemail, leading the contact centre system to wait, again silently, for the end of the outgoing message so it can leave one in response. A more accurate AMD which makes a reliable classification shortly after the initial greeting is required.

Ofcom effectively banned AMD back in 2008 because it was a major cause of silent calls.

The Ofcom (UK) requirements stipulated that silent calls are not acceptable in any form. This, along with similar issues in the US and other countries, has in turn spurred innovation in the AMD space - with modern approaches utilising machine learning technology to significantly increase accuracy. The very best of the latest generation AMD systems are accurate enough to comply with the 3% call abandonment rate guideline provided by Ofcom.

So, if you chose to ditch Answering Machine Detection, or to skew it heavily in favour of live speakers at the expense of agent efficiency (because of the 3% rate), now might be the time to reconsider. The world of technology moves fast - and well over a decade of innovation has taken place since 2008.

What is AI-AMD?

Aculab’s AI-AMD takes the traditional AMD system and enhances it with a trained neural network. This network, trained on tens of thousands of audio recordings, carefully examines outbound calls for speech or voicemail audio patterns, as well as using conventional cadence based analysis.

This approach can yield in excess of 99% accuracy.

The end result

So, what does Aculab AI-AMD do for the contact centre?

  • Reduces the percentage of silent calls, improving customer relations and allowing compliance with regulations.
  • Improves agent efficiency by reducing the percentage of voicemail services they’re connected with in error.
  • Saves agents’ time and enhances their productivity.
  • Allows the maximisation of bulk calling, while staying compliant with legislation.

What sets Aculab’s offering apart from the rest?

Aculab’s AI-AMD system has class leading accuracy. It can be hosted either on-premise or in Aculab Cloud, our versatile CPaaS solution. Our in-house development team also offers bespoke integration support and development assistance.

AI-AMD and Aculab Cloud provide powerful tools for creating your own custom contact centre solution, and facilitate easy integration with other Aculab products with high level APIs.

Interested in exploring AI-AMD? Sign up for a free Aculab Cloud developer account, contact an expert, or check out our video on AI-AMD today.