In recent years, the world of customer service has seen the emergence of a new type of call center agent: the callbot, also known as voicebot, voice agent, or virtual agent. But what exactly is a callbot? How does it work? And what benefits can this conversational AI-based technology bring to the contact center? All the answers in this article!

At the end of 2019, a BVA group study revealed that the telephone remains the preferred communication channel for the French. When calling a company, 65% of people prefer to use the phone call. Yet customer service departments often lack the resources to handle calls, which are the most complex flow to handle (need for immediate availability, major load variations, etc.).

At a time when artificial intelligence makes it possible to automate 40% of the calls received in customer relations centers (McKinsey Future of Work report, 2018), many companies have decided to rely on an innovative technology to automate these calls: voice agents.

Callbot: definition

A callbot is a voice assistant capable of conversing with a caller on the phone call, to understand his problem and solve it autonomously, 24 hours a day, without queue time.

You certainly know Siri or Alexa, which equip your smartphone or your connected speakers. A callbot is like Siri picking up the phone when you call a customer service.

Callbot, voicebot: what's the difference?

As the technology is recent, you may have heard about callbots under other names: voicebots, voice agents, phonebots, conversational agents, virtual agents... and there is still a lot of abuse of language!

In short, all these technologies are based on conversational Artificial Intelligence (AI). A voicebot is a vocal assistant capable of dialoguing with a person thanks to its understanding of human language. Siri, Alexa or Ok Google are voicebots. Voicebots can be installed on different channels: connected objects, GPS, connected speakerphone, smartphone, computer, etc.

The callbot is therefore a voicebot set up over the phone channel. This channel is associated with many specificities: while our interactions with smartphones or connected speakers are often very brief, limited to one or two sentences in a row, the exchange with a callbot is like a real phone call: it can sometimes last three, four, five minutes, or even more!

The callbot is therefore a specialized voicebot: it is capable of deploying a much richer and more advanced scenario, and of executing more complex tasks, but within a limited perimeter.

The benefits of callbots

For a customer service department, managing large call flows represents a significant financial cost. At the same time, many contact centers are experiencing recruitment difficulties and want to improve the quality of work life of their employees to limit turnover.

This is what drives automation to the forefront of customer relations: automating routine calls allows companies to reduce costs while reducing the repetitive tasks performed by their agents.

Not all call flows have the same value to the business. Callbots are designed to automate low value-added calls for which the company wants to remain available over the phone call channel. This frees up hundreds of working hours for agents, who can then devote themselves to higher value-added tasks.

Calldesk has deployed and put into production several dozen callbot projects. In terms of concrete benefits measured at the contact center level, here are a few key figures:

• 80% self-care by phone call: this increases customer autonomy and allows callers' problems to be solved without mobilizing the agents.
• 40% reduction in average call handling time (AHT): automating repetitive team tasks, such as data entry, during and after the call, increases agent productivity.
• A 10% increase in customer satisfaction: the callbot improves the customer experience by removing irritants from their journey, as it is available immediately, 24 hours a day, even during call spikes.

How does a callbot work?

A callbot is based on the most innovative technologies in conversational AI. To conduct a conversation lasting several minutes, a callbot combines different technological bricks capable of performing many complex tasks in a few milliseconds.

Connected to the contact center's phone system, and often hosted in the cloud, the callbot retrieves calls made to a predefined phone call number. It then engages in a discussion scenario with the caller, as programmed upstream by the business teams.

When the caller speaks, the system detects his or her voice thanks to the VAD (Voice Activity Detection) brick. This sound is then transformed into text thanks to an ASR (automatic speech recognition) system. These two bricks make up the STT (speech to text) engine, and make it possible to transcribe a voice into written form.

This text (transcript) is then analysed by the artificial intelligence engine, called NLU (natural language understanding), which understands the objective of the caller's request (intention) from the words it uses. Just like a human does, after all!

Thanks to this intention, the dialog manager uses the scenario and the business rules to choose the most relevant answer to give according to the object of the request. An answer is chosen, and will be transformed into a voice thanks to the TTS (Text to Spech) brick, which allows a message to be read thanks to a synthetic voice. And that's it: allowing a machine to speak with a human, it's as simple as that!

The challenge of today's callbot solutions is to perform all these operations in a very short time - one to two seconds at most. Technical performance has a very important influence on the caller's experience, and therefore on their satisfaction.

At calldesk, for example, this has led us to adopt a flow architecture, which allows these different operations to be performed simultaneously rather than sequentially. This allows our callbots to respond as quickly as a human would, and to ensure real fluidity in conversations.

What place for the callbot in tomorrow's contact center?

The potential of artificial intelligence for customer relations is extremely important. It is estimated today that 40% of the calls received in contact centers are fully automatable, using existing technology.

But the development of artificial intelligence is not at the expense of the company's employees. Quite the contrary, call center agents will have a central role to play in tomorrow's customer relations.

If the callbot can easily handle simple and repetitive tasks, the whole issue of customer retention is to mobilize employees at critical moments in the customer relationship to enable them to deliver an authentic and differentiating experience.

Tomorrow, we can therefore imagine a contact center where voice agents will manage all the repetitive tasks related to calls (understanding the need, identifying the customer, etc.), and will be able to deliver solutions that promote customer autonomy, while transferring only calls with high added value to highly qualified advisors. The development of the callbots is therefore accompanied by an increase in the skills of the teams.

Finally, a callbot is a talking chatbot?

In principle, yes.

However, a callbot is fundamentally different from a vocalized chatbot. Both technologies are certainly based on conversational artificial intelligence, but the voice has many specificities that must be natively integrated into the technology.

This is what enables the callbot to deliver the best experience to callers. For example, we don't care that a chatbot can take four seconds to answer us: we are used to this temporality in written discussions. However, in oral discussions, it is imperative that the callbot be able to answer in less than two seconds, otherwise the conversation will be interrupted by awkward pauses.

How much does a callbot cost?

Setting up a callbot is an ambitious project, reserved for companies receiving high volumes of calls. Setting up a callbot is profitable from 50,000 calls per year.

Setting up the bot can be done in a few days, especially in the form of a POC callbot, which costs between 10 000 and 30 000 €.

Secondly, most solutions have a pay-per-use pricing model, which depends on the number of calls handled or the volume of call minutes.

It is estimated that the cost of handling a call by a callbot is five to ten times less than by a human agent.

In conclusion, callbot appears to be a promising technology for achieving significant productivity gains in the contact center, but for the time its usage appears limited to companies receiving large volumes of calls. It is a good bet that the more mature the technology becomes, the more its cost will decrease. Tomorrow, callbots will be able to equip not only large companies, but also SMBs!