Secrets of Great Conversation Design: Setting Your Virtual Agent Up for Success

Discover how to tweak a virtual agent so it delivers exceptional experiences

Sponsored Post
Secrets of Great Conversation Design: Setting Your Virtual Agent Up for Success
Contact CentreTalkdeskInsights

Last Edited: September 30, 2022

Sandra Radlovački

Sandra Radlovački

Every interaction can be a base for conversation. When in need, it’s common knowledge that people like to talk to other human beings to find the answer to their issue quickly. After all, who understands humans better than other humans?

But the technical ease of growing Gen-Z consumer bases is challenging our assumptions. More customers want their answers faster, and advancements in artificial intelligence techniques are helping companies deliver on these expectations.

Companies are still struggling to make a virtual agent that mimics humans and meets all customers’ needs. While companies try to follow general principles of conversation design in this process, there is still room for improvement.

Indeed, general principles do serve as the foundation for a virtual agent. But how does one take virtual agents to the next level?

CX Today spoke with Dawn Harpster, Senior Conversation Architect at Talkdesk, to learn more about the features and nuances that make virtual agents excel in interaction with humans.

May I Speak to a Human Please?

Virtual agents have come a long way from basic bots that give generic answers to prompts. With the help of AI, Natural Language Processing, and Natural Language Understanding, virtual agents can now handle complex self-service requests. Or they should be.

Companies often notice their virtual agent is causing more customer frustration rather than providing a seamless experience. Why is this the case?

Harpster cites the word ‘containment’ as the trigger for customer frustration with some virtual agents. As she explains, companies aim to ‘contain’ the user within the virtual agent to prove its efficacy, but it ends up trapping users with no exit. She says:

“You need to enable the user to self-serve, but you also want them to be able to get to a human agent because a lot of times a customer needs something that is not on the menu.”

“People are often left guessing as to what they need to say or do to get to a human agent.”

Another frustration point in dealing with virtual agents is information overload. Customers can feel overwhelmed with too many choices in the menu and they might not be sure which option to choose to solve their issue. Harpster continues:

“If your virtual assistant is supposed to be transactional, just keep it that way.

“Give customers what they need to get their business done and let them go about their day. Often sales and marketing teams like to put extra messages into a virtual assistant but it can hinder the user from finding the right answer.”

Saving Empathy for Human Agents

In a Talkdesk ebook authored by Harpster, entitled Designing customer conversations: Best practices, she mentions the phrase ‘uncanny valley’ when talking about empathy in virtual agents.

While empathy inherently is what humans look for in conversations, virtual agents expressing empathy can come off as weird and can make people feel uncomfortable. Harpster explains what uncanny valley is and why organisations should steer clear of it:

“The uncanny valley is the point where the human like-capabilities of a robot stop being delightful and they start being creepy.

“Somewhere in the interaction an invisible line is crossed, and the human becomes completely repulsed by the robot and doesn’t wish to interact with it anymore.”

Robots that resemble and behave like humans a little too much might look borderline terrifying. The advantage of virtual agents is that customers only hear their voice, or, in other cases, read their messages.

This lowers the likelihood of the uncanny valley, but according to Harpster, it can still happen.

“It happens when companies try to trick users into thinking they’re talking to a live human when they’re not. And when people figure it out, they become annoyed and don’t want to interact with a virtual agent anymore.”

Another piece of advice for not scaring off customers is revealing only the data necessary for the current transaction. Since organisations have access to lots of data when operating a virtual agent, it’s best to refer only to data relevant to the task at hand.

Otherwise, customers might feel creeped out and get the impression that organisations are invading their privacy.

Handling Errors

As with meetings or other face to face conversations, people can get interrupted or distracted during calls with virtual agents. Life happens – dog barks unexpectedly, a doorbell rings or the kids scream. All these noises can affect how people follow prompts and virtual agents take in information. Harpster says:

“It’s important to set your virtual assistant up so that people can recover from those types of errors easily.

“Always be aware of the unexpected. We all want to design the happy, trouble-free path but it’s important to get the person back on the path when they wander off.”

In cases when virtual agents are asking for specific information and the user is unsure of where to find it, it’s best to provide instructions before asking the question. This way users can prepare for what is coming, follow instructions, and answer the question correctly.

Partner with Talkdesk

Talkdesk helps organisations drive better customer engagement through natural dialogue at any time of the day.

By building a powerful virtual agent following these steps, you can ensure that customers’ needs are met even when humans are out of the office.

Discover the DO’s and DONT’S of conversation design by reading Designing customer conversations: Best practices.


Augmented AgentConversational AIVirtual AgentVirtual Assistant

Brands mentioned in this article.


Share This Post