Conversational AI: Dialogue Traits of Sophisticated IVAs

Verint Team April 28, 2021

Verint has developed comprehensive expertise building, implementing, and delivering high-quality IVA experiences for customer service leaders such as Novo Nordisk and Alight Solutions. So, if you’re looking into incorporating or elevating an IVA within your customer service strategy, here are the traits your IVA needs to encompass to deliver an exceptional customer experience and gain a competitive advantage.

Human-in-the-loop Intent Understanding and Continuous Improvement

Verint IVA applies a combination of AI and Machine Learning techniques for intent identification and understanding that human trainers approve. If you’re just getting started, this approach allows you to quickly and accurately train the IVA on large volumes of data—such as chat logs and call-centre transcripts—by matching what people want to know with the many variations of the language they use to ask for it.

It allows Verint IVA to go beyond understanding the meaning of words to understand the customer’s true intent. Verint IVA also uses this approach to improve understanding and intelligence—and remove ambiguity—as it continuously extends capabilities with the human trainer approving and managing recommendations.

Dynamic, Complex Conversation: Multi-turn and Multi-Slot Design

Verint IVA delivers a rich and satisfying customer experience. A conversation with one can feel as authentic as a conversation with an actual person. These capabilities require a level of sophistication, including multi-turn intent determination and multi-slot filling, enabling the IVA to gather and keep information in the conversation.

Conversations, whether between two people or between a person and a computer, are complex interactions. When thinking about what defines a meaningful and dynamic dialogue, understanding the words is really the most basic component.

Verint IVA uniquely offers an authentic automated conversational experience that delivers higher capabilities and deflection by emulating the characteristics of interpersonal communication through a variety of approaches with multi-turn and multi-slot dialogue design:

Concept Memory: Conditions, Context, and Clarifying Questions

When speaking with people, there are often two types of information found in a particular sentence: general concepts and specific pieces of data. The IVA remembers the idea or concept the user was conveying in the previous turn and decides if it’s applied to the next turn. Verint IVA often uses this method to clarify the user needs and provide an accurate outcome.

For example, when the user asks, “what is your warranty policy?” the IVA will probably need to ask a few clarifying questions to see what type of product for the warranty. That is, unless the customer and the IVA have been talking about a specific product and the IVA has a contextual point of reference to add meaning to the question. This method also enables the IVA to compare the current user input with any links provided in the previous turn to assume intent and act accordingly, usually navigating to a specific area or directly responding.

Mixed-Initiative and Goal-Based Conversations

Sometimes we have conversations to obtain information, while in other instances, we are having a chat to achieve a specific goal. And then other times, it’s both. For example, a customer wants to find a good place to vacation and then book a flight, or perhaps they need to shop for a new insurance policy.

These scenarios start with a high-level objective, but the customer may not know the detailed processes or specific information required to complete their task. In a conversation with an agent or customer service representative, all of the necessary information is extracted, clarifications are made, special conditions are evaluated and applied, and, eventually, the customer is guided to their goal.

While online forms allow many of these tasks to be completed, the customer must adapt to the system’s design. Verint IVA handles the goal of finding and booking travel the same way as the human representative, even jumping out of the flow to answer ancillary questions, then jumping back in when the user is ready. Verint IVA requires no adaptation: everyone understands how a conversation works—even the computer.

Relational Cognition: A Breakthrough in NLP and Intent Understanding Beyond Slot-Filling

Verint’s latest innovation in conversational capabilities enables the IVA to go even further in mimicking human intelligence with the ability to represent and reason about relationships; connecting relations between concepts and information that otherwise would not be connected in a typical system.

With relational cognition, Verint IVA goes beyond slot-filling to apply anaphoric reasoning (first order or priority logic) and connect layers of understanding in relation to the context and other surrounding ideas within the ecosystem of understanding. This enables the IVA with a semantic representation of the conversation state and stronger inference for determining next best actions.

Simply put, it holds on to every piece of information to query in later turns of the conversation to provide more robust, manageable, targeted, and hyper-relevant experiences that drive human-like connections and understanding in language. Relational Cognition provides accuracy and robustness in knowledge and capabilities, which is incredibly important in compliant-rich environments that want to achieve sophisticated automation while truly supporting customers in a high-touch manner.

For example, if a patient is having a conversation with an IVA about being on vacation in the sun and then later shared a side effect they’re having while taking a medication, Verint IVA can correlate if there is a potential relation between sun exposure and the medication, and then determine the next best action.

Leading the Way

These new and advanced capabilities signal a new era of conversational AI, and Verint is at the forefront of these innovations. Through our work with some of the world’s most respected and trusted brands, we’ve seen that conversational AI is only going to become more of a necessity for nearly every enterprise, across most industries.

We’ve also seen the rapid adoption of digital and voice self-service accelerate as a result of the pandemic, and we’ve responded with IVA technology and solutions that don’t just use AI for simple use cases—they are now empowering organisations to implement AI throughout their entire company.

At the same time, these companies—as well as their customers and employees—will expect personalised, fast, accurate information in multiple languages and across channels. That’s not a prediction—it’s the reality we at Verint are seeing play out in real time and have the solution with which to rise to the occasion.