Conversation design is the language designers use to create and ameliorate dialogue flows and provide clear, natural, and effective communication between users and virtual assistants.
But since you are here, you probably already knew that, right?
In this article, we aim to show you the basic elements that make up the conversation design process of AI-powered virtual assistants. By the end of the article, you will be able to talk about conversation design in detail and manage to sound like an expert.
A persona is a role or personality designated for virtual assistants to make their interaction with users as “humanlike” as possible. Creating a persona is one of the most critical stages in virtual assistant design. By personification, virtual assistants are embellished and tailored for the brands they represent.
Channels are basically the mediums where conversations happen. Some examples include, but are not limited to, WhatsApp, Instagram, Facebook Messenger, or SMS. Each channel's user profile and experience may be different. For this reason, the dialogue trees of virtual assistants differ according to the channels. It is a conversation designer’s job to know which channel your virtual assistant should work on, and to determine the most suitable channel for it and its target audience.
Anything that makes sense that the customer can say is an utterance.
For example: "I want to order 2 coffees" is an utterance.
An intent -short for intention- is what the customer wants to retrieve as information or service. It is a concept where we group the same ideas together.
For example, when we look at “I want to order 2 coffees”, we can see there is an intention for ordering a drink. Therefore, we label these kinds of sentences as “Order” in AI and start training AI accordingly by putting the same kind of sentences together.
“I want coke,” “We would like to buy 3 lattes” or “Please prepare an Americano to go” will all be joined under the intent “Order,” therefore, making sense to the designers.
Entities represent details that can be grouped within intents. When customers want to say something, they sometimes use the same sentences with very small differences, differences that can be grouped together. For example: “I want to order a latte”, “I want to order a cappuccino”, “I want to order a macchiato” etc. As you can see they can use the very same sentence with only a one-word difference. We humans understand a latte and a cappuccino are the same thing; they are coffee! But AI does not know that yet. So, in order to teach it, we group these words under entity labels and use them in intents.
In conversation design, a flow represents a structure that determines how a conversation can take place between the user and AI, taking into account possible responses the users can give, predetermined scenarios, and rules of a chatbot.
In its simplest definition, a threshold is the tipping point at which the chatbot decides whether it can answer a question or not. It usually has a value of 0 to 100 and the sentences with a match above the number determined by the designers (usually over 50) are answered by the chatbot as they are deemed to have passed the threshold. The more similar intents you have, the higher you can keep this rate, and the higher the probability of your answer being correct will be. Likewise, if there are not enough intents, sentences cannot go over the threshold and your virtual assistant cannot answer questions because it cannot understand what is being said.
What we mean by training is to teach the information collected from the customers in order to increase the capacity of artificial intelligence so that it can answer on a wider spectrum in the future. Maybe your virtual assistant can answer “Thanks a million!” with a “No problem!” but it may not understand when the customer says “I owe you one.” This is when the designers step in and start the training process and teach AI what “I owe you one” means. Basically, we can say that virtual assistants are rising above the shoulders of the giants; giants being, in our case, designers and different expressions of customers.
This stage has a very important place in the conversation design process because the more your virtual assistant knows and learns, the more it will understand you. This is something that builds the trust of the end-users of your brand and your virtual assistant.
Agents are contact center customer representatives who take the stage when customer questions are not registered in the chatbot and when a human touch is necessary. While not all chatbots have them, some brands prefer to have live agents waiting on the line when customers seek help from a “non-machine mind”.
A happy path, as the name suggests, is the best possible path a customer can take while talking to a virtual assistant. This term is used for cases where designers have assumed and worked on all possibilities and the customer acts according to the expectation. An example of a happy path can be a customer who is ordering a latte via a coffee shop’s virtual assistant. With the customer’s expectation fulfilled, and the designers’ predictions correct, everybody is happy in the end.
Another instance of a happy path would be if you have read this article and now have a better understanding of the world of conversation design. Although we only scratched the surface, you can be sure that you have learned quite a lot! Feel free to check MindBehind Glossary which consists of many more terms that can help you immensely along your journey.
Discover MindBehind’s advanced conversational solutions to boost your sales and amplify your support services. Using our easy-to-navigate interface, start building your own sophisticated virtual assistant with zero coding skills required.
When was MindBehind founded? What’s our story? How do we define our vision? Visit us on our About Us page to find all the answers to your questions about MindBehind!
Finding solutions to your problems with our Support Center page is easy! Visit right now for all the answers to FAQs, how-to-guides and troubleshooting instructions.