Making Voice Automation Human

"Hi, we are calling about an accident you've had in the last year."

‍

Sound familiar? We aren't surprised. Most people in the UK have experienced annoying, robotic-sounding calls that attempt to persuade them to hand over important information. In 2021 alone, more than 1 billion nuisance calls occurred to people in the UK.

‍

Voice automation is often viewed with suspicion, thanks to this behaviour. Despite this, we are seeing ever-increasing numbers of people installing and interacting with voice automation at home, in the form of Google Assistant, Amazon Alexa, and several competitors, thus proving that voice automation is welcomed warmly at the right time and place. And businesses are coming to this same realisation, but the challenge is to make it sound great and work effectively for the customer.

‍

Increasingly, companies are seeing the importance of implementing voice automation into their business workflow. Unlike Alexa and Assistant, which are fantastic for general information, businesses need solutions tailored to their sector and brand.

‍

So how do you choose the right automation in the first place? It starts with understanding the business’s pain points for their customers and teams. At that point can we design and build what the voicebot needs to be able to do and say just like training a customer service agent. We then use the voicebot where it going to be most powerful and add value to the customer experience and of course saving vital time and overhead.

‍

How do you give the bot a natural voice of its own?

When it comes to creating the voice, there are several options.

‍

The simplest and quickest options are pre-existing Text-To-Speech (TTS), which comes in many flavours. For example, we can choose from UK, US, or Australian English and pick from numerous male or female voices. The greatest strength of this option is that it is quick to deploy and can respond with any words that we need in the flow of conversation and is used widely for pilots and trials or very broad conversational responses.

‍

An alternative option is Clear Cut Text. Here we take a voice artist into the studio for an hour or more and record all the required words for the scripts we have established for the business needs. In addition to the critical dialogue, we record a variety of inflexions, tones and natural conversational sounds, such as "ah!" The Clear Cut Text route allows the brand to select a voice artist that suits them to be the voice of their voice automation, but also allows a very human sounding response set. We know the journey and can anticipate all the usual hesitancy and allowances that make a conversation flow and this very natural response sound is how we make the voicebot human.

‍

Alternatively, we can create a customised TTS by taking the voice artist into the booth for several hours. We record all the language sounds that are needed to make a TTS, but with the client’s own voice artist. A customised TTS allows for future business changes, where the responses may need to be very broad. While this may need some extra time and effort and is a little more expensive, it is a valuable asset to businesses that value the ability to retain all branding for years to come, including the sound of any voice applications from phone voice mail to of course the voicebot.

‍

A natural conversation is a human conversation?

After the voice is created, the conversation flow is where we start to make our voicebot sound human. Our aim is not to fool the caller into thinking they are speaking to a human, but instead to relax into the conversation and let them forget they are talking to a computer.

‍

Of course, each caller would be ready with all their information in an ideal world, and the conversation would follow an expected path. Yet every customer agent or team leader knows that there will be interruptive conversations that need resolving, not having information to hand, or going back in a journey and any of the flows making the conversation, natural.  

‍

We factor these unhappy paths in too, ensuring that your voicebot is ready with the tools to switch conversation where needed all the while having the right pauses, tones, and phrases to help reassure callers that despite talking to a bot, their enquiry gets dealt with reliably. We can escalate to a human agent of course, but we find that this is very much the exception rather than the rule now that technology is mature enough.

‍

Nuance and understanding.

Humans react best to situations where there appears to be an element of consideration in the process. As mentioned the technology of understanding speech is very mature and to give the nuance to conversation, we instantly know the next step of the process, by understanding that we know we have been asked an informed question and what is the context of the question. By doing so, we obtain a more natural, humanistic conversation.

‍

For example, our voicebots can handle multiple elements of a request and then prompt for the outstanding information necessary to proceed. So, If a caller states that they wish to book an appointment for 3 pm with Doctor Smith, the bot will crucially understand that informed request and simply reply to ask which day they are referring. Once that information arrives, the entire request can be passed to any calendar or system and a confirmation email or text sent to the customer.

‍

This ability to prefill the data as the customer speaks and fill outstanding points with tailored responses leaves the customer satisfied with their journey while shaving precious hours off the overall call time and of course reduce call centre costs on calls that can be self served.

‍

We know we cannot stop customers from insisting on speaking to a real human being, but we have shown time and again that voicebots typically cut down contact by 80-90%. It is clear that the power and maturity of voicebot technology is here to stay and should be part of any regular comms package.

‍

‍