The Art And Science Behind Conversational AI


It’s no secret that today’s customers expect highly-personalized, seamless communication across channels when they connect with companies. Artificial intelligence (AI) is rapidly becoming a ‘must-have’ for brands looking to have a competitive edge in the battle for customer loyalty.

Using conversational AI to field basic customer queries is often an accessible starting point for brands. By 2030, the conversational AI industry is expected to be worth $41.39 billion. As conversational AI technology advances, more brands will look to enhance their contact center operations with features like a virtual agent.

To ensure that a virtual agent can effectively meet customer expectations and help deliver a great customer experience (CX), there are some considerations that brands need to keep in mind when designing AI conversations

Keep it simple

Why is a customer reaching out to a contact center? To resolve an issue. Getting the customer from point A to point B, as quickly and efficiently as possible ensures the most seamless, satisfying, and trouble-free path to issue resolution.

Humans can realistically only remember three to five points at one time, so it’s critical to keep the conversational flow with a virtual agent simple.

Writing voice prompts that mirror the way customers actually speak, will provide a more engaging and effective experience. An overcomplicated, flowery language that may seem elevated on paper will likely fall flat, with customers in the conversational flow.

Companies are also wise, to avoid overburdening a custome,r with too many marketing or sales messages delivered through a virtual agent. These extraneous elements, at best, will slow down the journey to issue resolution and, at worst, may confuse or frustrate the customer. 

The virtual head nod

Experts who study the science of human-to-human interactions note the importance of tone, expressions, and mannerisms in productive conversations. Absent these visual and tonal cues in AI interactions, good conversational AI design will provide discourse markers. Words like OK, thanks, or got it are necessary elements to mark the progress of interaction in a virtual conversation.

Discourse markers indicate that the conversation is moving along and that the customer has been understood. In the virtual world of conversational AI, discourse markers are like the head nods often used in face-to-face conversation to indicate understanding.

These virtual head nods can also be used to signal a change in the topic within a conversational flow. Or as a reporting tool to indicate the completion of a key performance indicator (KPI). 

Managing expectations 

The system delays or dead space can occur – especially when the virtual agent needs to retrieve backend information from an application programming interface (API) or customer relationship management (CRM) system to advance the conversation.

Building in prompts that clearly communicate the potential for wait time while information is located is the best way to manage customer expectations proactively. A simple statement like ‘Please wait while I look that up’ will provide assurance to the customer that the conversation has not been disconnected or abandoned when the virtual assistant goes quiet. A sound effect or even music can provide the same assurance. 

Life happens 

There are a variety of factors that might lead a customer to “erroring out” of the conversational flow – a barking dog, a ringing doorbell, screaming kids, multitasking, or just plain being distracted. Unpredictable life moments sometimes happen during customer service encounters.

In human-to-human conversation, it’s easy for customers to communicate that they’re confused and need to reset focus. Conversational resets are more challenging in human-to-AI exchanges, but there are error recovery techniques to help get the conversation back on track.

Repetition built into pivotal moments in the flow is an error recovery strategy that solves many of the most common causes of conversational derailment. For example, if a user has missed a question, the virtual assistant can recognize a delayed response, a lack of response, or a response not fitting for the question – triggering a repetition of the missed prompt. Adding additional details with the repetition, like suggesting potential responses for that question, can further support the customer, get the conversation back on track, and move the interaction along.

Even with the best-intentioned and best-timed error recovery tactics, there will be customers who just can’t recover from a conversational wobble or don’t want to continue with a virtual agent. In these instances, it is important to have a defined point of no return that will trigger an escalated connection with a human agent. 

Game-changing technology for contact centers 

With conversational AI technology, contact centers can apply deeper intelligence and automation across communication channels. This can result in more meaningful personalizations for customers and an effective issue resolution. Virtual agents are rapidly replacing traditional interactive voice response (IVR) and basic chatbots to create more human-like customer interactions, even when the human agents are not yet involved in the conversation. This technology allows human agents to better manage and address more complex customer queries by liberating them from basic or repetitive tasks. Conversational AI can be a game changer for the modern contact center when designed and applied well.

Conversational AI design is both an art and a science. It’s more than just writing a script for a virtual agent. Those designing this type of customer engagement need to bridge the gap between the machine’s technical functionality and the human customer’s unpredictable nature. Good conversation designers will always be an advocate for the customer.  They will maintain realistic expectations about the role of conversational AI technology in the larger customer journey. 

If you liked reading this, you might like our other stories
Finding The Right CX Talent
All You Need to Know: Google Duplex

Previous articleLessons From The Auto Industry For Everyone
Next articleTools to Boost Your Customer Experience Management
Dawn Harpster is the senior conversation architect at Talkdesk. In this role, she helps contact centre leaders harness the power of conversational AI by designing intelligent conversation flows for virtual agents. She believes that conversational AI technology can simplify customer problems and help deliver outstanding CX. Dawn studied broadcasting and applied psychology at university, and has worked professionally as an engineer, a writer, and a television producer. Her unusual convergence of skills – understanding human behaviour, taking things apart, and writing – makes her uniquely suited for conversational AI architecture. In fact, Dawn is likely among fewer than 100 professionals worldwide currently working in this niche area of conversational AI architecture.