Voice Technology and ChatBots

How to design conversations with voicebots that build trust from the ground up

It's human to be wrong.

We make mistakes every day: we forget our headphones at home, we miscalculate the bill, we nod and smile when we really didn't understand what they told us.

It has happened to all of us.

We don't stop trusting ourselves—or other people—just because we make a mistake. It is usually the sum of errors over time that ends up breaking our trust. Maybe that's why it's easy to understand when a person makes a mistake - but not so much when a robot makes it.

In this second session of Vozy Unplugged, we were accompanied by Leonardo Noriega, Conversational Lead at IBM, and Alfredo López, Conversational Lead at Vozy, to explore the practices and characteristics of a conversation that truly builds trust.

 

Fear and Trust: The Two Faces of Prejudice

Robots, although they have better technical abilities to handle and consult information than people, are not perfect. The margin of error is much smaller, but in the end, we build them ourselves. That makes them imperfect.

When we talk to a human agent, we expect several things from him:

  • We expect to receive better treatment than if I were a robot. It doesn't matter that it's just as feasible to be mistreated; since you have a greater capacity for expression, it's easier to identify with it.
  • We hope to have the opportunity to persuade you to help us. Sometimes we think that it is necessary to convince the agent to help us, we give him reasons or we threaten him to give us a solution. With a bot this is impossible, because it does not have the capacity to understand that context.
  • If you can't solve the problem for us, at least we can complain. We know that won't solve the problem, but we do it anyway.

These false perceptions don't take into account that it's possible to be mistreated by a human agent; and a bot can make mistakes. The opposite is also possible.

In addition, ironically, a person will often rely on a machine to consult any information, while the bot and the machine can talk to each other more effectively, faster.

Who to trust, then, if neither one is perfect?

Trust is a process built on good experiences: no matter who serves the customer, as long as the experience is satisfactory, trust will be strengthened.

 

Without empathy, there is no conversation that lasts over time.

The key to trust, then, is to design good user-centered experiences. It's putting yourself in their shoes and interpreting what their need is, what they're trying to solve, and based on that, putting yourself in their shoes and wondering what would happen if you couldn't solve it.

The key to trust is, in other words, empathy.

What would I like to be told? How would I like to be told? What alternatives or solutions would I find satisfactory? These questions need to be thought carefully about. The objective should always be to minimize the anxiety, frustration and anger that the user feels, and to achieve this, it is necessary to anticipate the triggers for those emotions.

One way to do this is to offer you alternatives when it's not possible to help you at that time. After all, it's not about being perfect; as Leonardo said,

We cannot pretend as human beings to call a bot and to solve our lives, to be capable of absolutely anything. Because we are not capable of that. And back to the point: we are the ones behind the bot. And if we can't solve absolutely everything ourselves, we can't expect a bot to solve everything for us either. We have to understand that there are situations that do and situations that do not, and in which no, how do I accompany you in this situation that you did not expect to seek a resolution on this path that does not leave you with a bitter, or at least bittersweet, taste?

 

Understanding this makes it possible to develop conversations that last over time and generate a bond of trust with the user.

 

Behind each Voicebot there are always... people.

As we mentioned in the previous paragraph, behind the design of a voicebot there is an entire team of people dedicated to building the best possible conversational flow for a specific user. But their work goes further: they teach machines how human beings speak to generate reliable, consistent and predictable experiences.

However,It really takes an entire team to design a conversation?

As Leonardo said: “We don't need to think about conversation, just like fish don't need to think about water.” That is to say: although anyone can design a conversation because it is something we do even in dreams, not everyone can design any conversation.

Precisely because we humans are fallible on our own, it is important to have the collaboration of a team of professionals with diverse perspectives. From the user experience, it is necessary to consider that we are not the user. We are designing a particular experience for someone else. If we stick with what “I” think is best, it is very likely to be wrong, to fall short, or not to consider all possible options. It is very important to analyze each variable and each situation before designing the conversation. What type of audience will we contact or will they contact us? What are your needs, your objectives?

If we don't know who we're talking to, we can hardly manage to channel a conversation with a clear objective and build trust.

 

Ambiguity is the worst enemy of trust

However, knowing your user is only half the way. To speak to him in the best way, it is necessary to understand that we are talking to human beings, and human beings are subject to their own contextual and environmental circumstances.

Maybe he'll get your call while he's driving and in a hurry. Or when you're in a mall, surrounded by noise. I could be about to enter a meeting. Our lives are accompanied by unforeseen events, noise and bad satellite signals, and all of them are variables to take into account when designing a conversation so as not to lose the user's attention.

Because the reality is that human beings deconcentrate very easily, and the only way to compensate for this is to eliminate ambiguity from the conversation.

This is possible only if we make sure of 3 things:

  • That the information we give to the user is only the most necessary and of quality;
  • That the way in which we deliver information is very synthetic, clear and with an obvious intention.
  • That the way in which we ask for information leaves no room for open answers.

If we don't take care of these points, the likelihood that the conversation with the voicebot will be successful is drastically reduced. As Alfredo said, “If there is ambiguity, we fail. That is why it is essential to evaluate it at every point. We have to learn to detect if that is happening at any point in the conversation to turn it around; what information can we remove, or show in another way?”

 

In the future it will be normal to talk to voicebots

As Alfredo said,”Trust is a process that is built over time on the basis of good experiences. Perhaps robots are already a part of our lives today, but not as much as they will be in the future, surely. This trust is going to be forged.”

As we adjust the technology and get to know our users better, we will be able to create ever better experiences for them. Many years ago, any management was done by telephone; it was unthinkable to enter personal data into an application or a website. And now? Now nobody even crazy wants to call a hotline before going through the application first and solving it comfortably.

A similar process is likely to occur with voicebots. Over time, trust in the effectiveness and professionalism of robots will become normal, and they will become a daily presence in our lives.

 

Conversational Good Practices

If you follow these good practices, the quality of your conversations will increase dramatically and it will be easier to create trusting relationships with your users.

  1. Carefully analyze the purpose of the conversation. It is crucial to be very clear about what we intend to achieve with that conversation, and to never lose sight of it in order to arrive as quickly and simply as possible. What are we trying to solve?
  2. Be clear about who you are going to talk to. Understanding the person you're trying to help is necessary to give them the best possible user experience. What are their needs, their fears, their weaknesses?
  3. Make turn-based taking obvious. It should be clear when the voicebot will speak, and when the user will speak.
  4. Take steps to repair the conversation. It is necessary to consider all possibilities to redirect the user to the conversation when they stray to help them achieve their objective.
  5. Choose and synthesize the information that will be given to the user. We must analyze very carefully what information we will provide, and carefully assess what should be left out and what should be inside. It's very common to want to give a lot of information to the user, but it's often not necessary because it doesn't add anything to the conversation.
  6. Use discursive markers. These types of elements not only indicate to the user that they have moved on to another section of the conversation, but they also give a lot of naturalness to the conversation (” I understand”, “first of all”, “we're almost done”). It almost feels like talking to a person.
  7. Avoid monologues. When trying to convey too much information to the user, it's easy to lose not only the user's attention, but the purpose of the conversation. Monologues aren't conversations, and they can lead to a poor user experience.
  8. Offer alternatives. A palliative solution or an escape route must always be given to the user. If you get caught in a dead end loop, you'll most likely have a very bad experience.
  9. Take care of your voice. It is important to ensure that the voicebot's voice is clear and has the right intonation to ensure that the user will be able to understand it.
  10. Iterate over the process. When a bot comes out, it has a life of its own. You have to go out and test it, put it on the ropes and always find its weaknesses from the user's perspective to improve it. Reviewing, analyzing and studying the interactions it has with users allows us to really learn how a bot relates to its users.




 

If you liked this article, you might be interested to find out why Although offering Anti-Customer Service is expensive, so many companies are willing to pay the price.

Compartir en LinkedInCompartir en WhatsAppCompartir en Facebook

We have more content of your interest

Technology and collections

ITS Infocom: 5 Keys to Accelerate Your Billing with Voice Assistants

Read more
Technology and Business

Menos espera, más soluciones: lo que pasa cuando aplicas IA conversacional en tu CX

Read more
Technology and Business

Vozy strengthens information security with its ISO 27001, ISO 27017, ISO 27018 certifications and SOC type I and II reports.

Read more
Customer service and sales

The Immediacy of AI + Human Empathy: A New Formula for Exceptional Customer Service.

Read more