Best practices and downfalls for conversation design for chatbots and voice assistants
The art of conversation evolved early in human evolution. As a species, we’ve been perfecting our verbal and nonverbal communication skills over hundreds of thousands of years, and it still is something that takes us years to learn.
We’re at one of the most pivotal periods since the beginning of human dialogue. With artificial intelligence (AI) powering chatbots and voice assistants, we’re now teaching human brains how to communicate with artificial brains. When designing conversations, you need to understand the needs of both humans and machines and ensure the dialogue is helpful, natural and persuasive.
When writing dialogue for a chatbot or voice assistant, there are key things to follow. Here are the best practices for conversation design for chatbots and voice assistants.
5 Strategies for Engaging Conversation Design
1. Start with a great introduction.
Introductions should be kept short and serve as a way to let the user know what you can help them with (don’t over-explain, but set expectations). Avoid using phrases like “I’m still learning” and avoid being arrogant with phrases like “smart virtual assistant.”
You will also want to create different introductions for returning users who will already be familiar with the capabilities. Never try to pass off a bot as a human agent.
Example: Hi, I’m Stella, the Sullivan Clothing virtual assistant. I can help you track your order, start a return or answer product questions. So, what can I help you with?
2. Write compelling and clear prompts.
Prompts let a person know that it’s their turn to speak and give input. It’s good practice to use question marks with prompts as it’s a clear indicator that it’s the user’s turn. Always end the message with a prompt (don’t add additional information after) and avoid combined questions and instead ask multiple simple questions.
Example: Let’s track your order. Are you looking for the order placed on September 13?
3. Use Implicit and Explicit Confirmations to ensure the conversation is moving in the right direction.
Use implicit and explicit confirmations to give users the chance to correct the chatbot in the case that wrong intent was triggered. Explicit intents ask the user for confirmation. Whereas an implicit intent keeps the conversation moving forward, but the bot does restate the user’s request. This gives the user an opportunity to correct the bot.
Explicit Intent Example: You would like me to initiate a return, is that correct?
Implicit Intent Example: I’ll start your return now. What’s your order number?
4. Use acknowledgments to have more natural conversation.
To help your users feel like you’re listening to them, use acknowledgments like “Alright,” “Sure,” or “Got it.” It’s a way to build trust and make the conversation natural. You’ll want to make sure to add variety and not use the same acknowledgment again and again.
Example: Got it. So you’d like to return the sweater because it’s too small?
5. Focus on the user, not the bot.
You want to be as user-centric in your writing as possible. Limit talking with “I” and “me,” and focus more on “you.”
Example: What would you like to know? vs. I’m an expert in fall fashion trends.
Conversation design is a science.
People communicate differently, so too do chatbots. And good conversation design is critical to the success of a chatbot and user engagement. When designing dialogue for chatbots and voice assistants, it’s important to understand both the needs of humans and bots. Write dialogue that is natural, similar to how a conversation would be had amongst friends, and less robotic. Understand the technical limitations, what information a bot needs from a user to complete a task, and always have human escalation, whether it’s direct handoff or call / email back.