Steven Long 2

Steven Paton

Oct 18, 2021

Steven Long 2

Steven Paton

Oct 18, 2021

The CX-Virtual Agent – Am I Really Not Talking to a Human?

In 1950, one of Britain’s most famous mathematicians, Alan M. Turing, proposed a test that could be used to evaluate artificial intelligence. In this test, a human judge observes a simple text conversation between two parties, knowing that one is a human and the other one is a machine but not knowing which one is which. Then, the judge guesses which party is the machine. If the guess is incorrect, the machine passes the Turing test.

Despite being more than 70 years old, the Turing test is still surprisingly applicable to certain types of artificial intelligence solutions, for example, virtual agent systems. By using advanced technologies in the development of such systems, their creators ultimately try to make them feel more human. Such technologies help the machines understand the conversation better, draw better conclusions, and ultimately serve the customer better.

Does the BMIT CX-VA virtual agent pass the Turing test? Give it a try and see for yourself if it would pass it for you.

The Power of Natural Language Processing

In the heart of every virtual assistant/virtual agent lies a natural language processing (NLP) engine. This is the element that lets the agent understand human input and behave appropriately based on that understanding.

NLP engines begin by pre-processing the text entered by a human. They tokenize it (break it down into smaller fragments), recognize parts of speech (verbs, nouns, adjectives, etc.), convert words to stems (root forms), and eliminate words with little meaning (such as articles). After that, the engine classifies the text, usually by assigning categories and/or tags.

Natural language processing involves several classification methods. For example, by performing sentiment analysis, the NLP engine recognizes the emotions within the text and can figure out whether the customer is excited, neutral, agitated, etc. Intent detection lets the NLP engine recognize what is the goal of the conversation, for example, whether the customer needs help or is just looking for some extra information. Topic classification assigns text to categories, for example, technical, financial, services, etc.

Thanks to these NLP capabilities, our virtual agent resolves the most common/repetitive queries by providing the right answers or routes the ticket to the most appropriate human agent. For example, using intent detection and sentiment analysis, the CX-VA can detect that the customer is aggravated and requires immediate help. Then, it can calm the customer down by saying that the ticket is recognized as critical and immediately routed to the dedicated staff/ department. At the same time, another customer that is just looking for current pricing information is served without the need to involve any humans.

One of the features that allows the CX-VA to stand out from the crowd is that it has access to over 100 languages (anke il-Malti!). This is enabled by its architecture being agnostic to the Natural Language Processing platform (NLP) it is set up to use. This allows the CX-VA to access any of the languages available via NLP engines from Google, Microsoft, IBM, or Facebook. We will choose the best platform for the language your business needs.

Together with the Human

The CX-VA will augment alongside and work with your human agents, allowing them to focus on real queries & VIP nurturing, dedicating the CX-VA to triage and handle the high volume, first-line repetitive queries without bias and emotional attachment. This way, the human factor can be used efficiently where it is most needed and the machine will eliminate the biggest causes of human agent frustration and burnout.

We designed the CX-VA so that it can learn more about the user. For example, if your customers are authenticated, the CX-VA will consider the entire conversation history, appearing to the customer as if it remembered the conversation that they had, for example, a year earlier. The CX-VA can also take a lot of other types of data into account, for example, the country that the customer is playing from.

Last but not least – the CX-VA is based on artificial intelligence. This means that it gets better with time. The more your customers use it, the better decisions it will make to help them. And to make it even better, you can lurk in the back and influence its choices to meet your specific/unique business branding.

Working Together – Working Better

While the CX-VA is very powerful on its own, it shines even more when you make it part of your customer support environment. For example, you can connect it to the CRM system and other solutions that you use to serve your customers better so that it gets as much contextual data about the customer as possible.

Using our REST API, you can make your agent work together with any system that supports REST. You can also greatly enhance the capabilities of the virtual agent. For example, you can use the API to allow the agent to perform multi-factor authentication (MFA) during the conversation with the customer.

Will You Be Talking to a Virtual Agent?

To see the CX-VA in action, just ask us for a free demo. We’ll be happy to show you its capabilities and help you learn how to make the best use of them in your business.

Oh, and we promise. We won’t be doing a Turing test on you 😉

Read next