7 Secrets of Great Conversation Design for Chatbots
Seen by some as an artificial intelligence annoyance, and others as the future of customer service, the chatbot has become an essential tool for those wanting to have conversations at scale.
As you know, we at Lever are passionate about learning transfer, and we have our own Coach M chatbot. Coach M helps us in our mission of making learning a key contributor to sustainable business growth. Plugging into any learning innovation, using Coach M couldn’t be easier. Through the pursuit of excellence in chatbot design, we have learned some secrets of great conversational designs for chatbots, and I’d love to share some of these with you.
1. Get to know your chatbot personally – what, why, when, who and how
As odd as it sounds, you need to think about your chatbot personally. Ask all of these questions as you begin: What is the aim of your chatbot? Why would anyone need your chatbot? When would they use it? Who would use it? And how would they encounter it? Is your chatbot mobile-accessible?
Your chatbot could be designed to serve as a concierge to support a program, or it could share post-learning activities. It could be a FAQ – a question and answer chatbot for those needing to access information or answers to common questions, or like ours a coaching chatbot. Really think about the functions of your chatbot as you begin its design.
2. Understand the deep conversation flow
Our Coach M was modeled as a result of more than 20,000 human interactions that my team has had with participants regarding learning interventions. You must take the intelligence that is gathered from human-to-human conversation flows and program them into your chatbot.
Think about the types of conversations you are trying to model. It may be a very short interaction. It may be on the longer side. Write down your conversations into different stages – the greeting, the expression of the problem, the follow-up action, the set of solutions or stages of solution, and where to go for help.
3. Be friendly but not too friendly
There is an art form to creating a chatbot, and much of it comes from how the chatbot talks.
Initially, you need to ask how you want the chatbot to say hello, and how it is going to end a conversation. If the chatbot has an audio component, will it speak with an accent that models your own?
We learned the hard way that the more friendly your chatbot seems, and attempts to model human conversation, the more casual the conversation will become. Sometimes those casual conversations are more difficult for the chatbot to comprehend because the terminology is too loose or indirect. The chatbot may find it challenging to understand what the user is trying to convey.
You and your staff are likely to have an intense period of learning as you start to create a voice for your chatbot.
If you would like any assistance with the conversational design of your chatbot, reach out to us if we can’t help you ourselves, we can put you in touch with someone who can!
4. Read aloud your chatbot transcripts
This is not a bad idea for any written work, but it is vital in terms of your chatbot. When you have a chatbot in a beta stage, make it a habit to read your transcripts aloud. Do they sound like human conversations? If not, then you can tweak them so that they do.
Think about how they can flow better. Is it improper punctuation? Is it word choice? Is there an acknowledgment that is missing? The more you read them aloud, the more they come to life.
When you read them aloud, it is recommended that you read it into a recording device and then play back the recording to yourself to hear what the conversation sounds like.
5. Be transparent and accurate
Make it clear upfront that users are talking to a chatbot. Have the chatbot explain its own purpose, and explain how it is going to interact with the user.
“Don’t worry about what you say or how you say it. I am a chatbot and I don’t have emotions, so I can’t judge you.” That helps create the proper expectation in the mind of the user and adds trust in the purpose of the chatbot.
If the chatbot has a human backup team, share that information. If, then, your human is stepping in to respond to a redirected request, make certain the user knows they are talking to a human. Make it clear that the human service is set to solve that individual query.
Being transparent and accurate builds trust and creates an atmosphere for a more meaningful conversation.
6. Back up your chatbot – and not the way you think!
Digital backup is a huge topic. No one wants to lose data or functionality. Backups provide the security that key information will not be lost due to power outage or system malfunction.
But here, we mean the backup to the service the chatbot provides. Make certain there is a rescue option that allows the user to speak to a person when the user feels it is essential to do so.
If you are in a business that requires a call center, set proper expectations for how long it may take for a human to get involved.
The human interaction may be typed or texted; it need not be verbal.
One of the reasons for your chatbot may be to avoid the requirement of having a human hired to handle the conversation. In our research, we have found that less than 5 percent of chatbot users require or request a human interaction.
7. Be human in your design
Avoid stilted conversation. Depending on the complexity of the conversation, it is best to keep the chatbot experience as light as it can be, while also being direct. The more that the user can ‘relate’ to the chatbot in a person-to-person way, the more value the chatbot will be to the user and the more frequent the use of that chatbot will become.
In conclusion
I hope you find these tips helpful for whatever stage you are at in the development of your chatbot. Coach M can be bought in a self-service format, supported or fully serviced where we do everything for you. We would be very happy to talk to you about what suits your needs and for any chatbot support you require within your learning models.
We look forward to hearing from you.