Good Chatbot Design Practices for Building Trust

• Culture Matters
• Design Practices

{ 💬 } – Chatbot design: focusing on what matters

A couple of months ago, a client reached out to me with an unusual request.  Normally, I approach chatbot design from the ground up, first going through a detailed process of creating a personality, assessing and ranking the client’s business objectives, and then writing the dialogue and coding the technical bits (such as third-party API calls).  

This particular client, however, had already built his chatbot, including writing all of the dialogue.  And what he wanted was for me to make it “less boring.”  But when I analyzed his chatbot, I realized that the biggest problem was not a “boring” impression of the chatbot’s dialogue but a distinct lack of trust-building.  The reason that the questions seemed “endless” and boring is that I had no idea who I was talking to, what they wanted, or why I was being asked so many questions.

Chatbots Are Always a “Face-to-Face” Conversation

Chatbot significantly differs from most other forms of digital design in that there are no “anonymous” spaces.  Although savvy Internet users know that every mouse movement and click is tracked via heat maps and cookies, that identities are being logged, and that personal profiles are being built (and often sold to third parties), there is still a perception that is possible to browse a website or use an app “anonymously.”

Not so with chatbots.  From the moment an interaction begins with a chatbot, the user will feel as though the chatbot is “talking” directly to them.  In meatspace terms, it is the difference between browsing in a clothing store versus having a shop assistant come up to you and engage in a conversation.  

Due to the way human beings take in social cues, this direct “face-to-face” conversation with a chatbot can often feel very intimidating.  From the user’s perspective, a “total stranger” has approached them and started asking questions.  If the questions are too direct or too personal, the entire experience will quickly feel off-putting, causing the user to want to end the conversation by navigating away to another website, webpage, or channel.  

In some cases, such as using a chatbot to order physical goods (like a pizza) from a familiar brick-and-mortar location (a local pizza restaurant) can help reduce this social anxiety.  Even if the user feels as though the chatbot is being too intrusive or too direct, the fact that the user can reach out to a human being (a staff member at the pizza restaurant, in this case) will often be enough to reduce the user’s anxiety.  Or, if a brand has enough recognition and awareness, this, too, can be enough for the user to perceive the chatbot as trustworthy.

In the vast majority of cases, however, chatbots are representing digital-only businesses with no previously existing basis of trust.  Chatbots, therefore, can come across as a total stranger stopping you on the street, wanting you to answer a few survey questions or sign a petition.  Yes, a few people will comply, but the majority will not.  And some people will reply but give false information because they do not trust the person asking the questions or how their information will be used.

Identification

Most employees in brick-and-mortar businesses wear a uniform, and many wear a nametag as well.  This is a social cue that the person in question is, indeed, an employee.  And whether it’s by reading the nametag or the person in question introducing themselves (“Hi, my name is Steve.  How can I help you today?”), it is imperative that the customer feels like they “know” to whom they are speaking.

The same is true with chatbot design.  Every chatbot should introduce themselves and explain both who they are and whom they represent.  A brief but informative “about the company” option should be included in every introductory sequence.  In many cases, the chatbot will have an actual name.

Building Rapport

The second element to building trust between a chatbot and a user is by building rapport.  Many businesses want their chatbot to “dive right in” and get straight to the business objective, whether that’s collecting contact information or making a sale.  But without first building rapport, very few users will trust the chatbot enough to hand over their contact information, much less highly sensitive information like birth dates, Social Security numbers, and credit card numbers.

Establishing rapport with a chatbot is done the same way as humans do to build rapport – finding common ground.  The user came to the chatbot looking for something, so why not build upon that to establish rapport?  For instance, I build a chatbot that was selling mortgages in the state of Florida.  But before asking the user to hand over their personal information, the chatbot invited the user to play a fun trivia game about unusual Florida facts.  This helped build rapport by establishing that both the chatbot and user were “from Florida” and had this “fact” in common.

No Surprises

Thirdly, explain exactly what you are going to ask before you ask it – as well as what you will do with that information.  Before asking for someone’s telephone number, for example, explain why you want it – as well as what you will do with it once you get it.  Are you going to sell that information to a third party?  Are you going to have a human staff member call that person?  If so, when?  And why do you need their telephone number as opposed to an email address?  And what benefits, if any, will accrue to the user if they give you an accurate telephone number?

Furthermore, if you do need to ask the user a lot of questions, explain that ahead of time.  If your chatbot is going to be asking questions on different topics, announce each topic beforehand.  Explain the procedure and ensure that the user knows what to expect, and they will be more likely to trust your chatbot by both answering all questions as well as providing truthful/accurate responses.

A Graceful Exit

Lastly, trust-building involves giving the user a polite “way out” to end the interaction with the chatbot.  Instead of forcing the user to close the window or navigate to a different website/channel to end the interaction, always give the user a way to end the conversation gracefully.  This can be in the form of an “exit” button in a menu or setting up an intent/keyword response that triggers when the user says “goodbye”, “quit”, or “exit.”  Just knowing that the user CAN leave will, perhaps counterintuitively, give them the freedom to explore and spend more time interacting with your chatbot.

Conclusion

Approach chatbot design exactly as you would in training a human staff member for a “customer-facing” position.  Start off every conversation with a greeting and an introduction.  Built rapport by finding common ground.  And if you do need to ask the user for something, be sure to explain both why you need to ask for it as well as what you will do with that information.  By following these basic design principles, users will greatly increase their trust in the chatbot (and, by extension, your business), lengthening overall interaction times as well as improving the accuracy of their responses.