What is Conversation Design?
In Conversational AI, there is a discipline known as Conversation Design. But what is it exactly?
When designing Chatbots or Smart Assistants, a Business Analyst will take on several tasks. In addition to identifying intents and slots, she will need to create the conversation design. This is a known as a happy path when determining how a person will interact with a smart assistant.
Here’s a quick example of someone wanting to update their mailing address at a bank.
Smart Assistant: Hello, Welcome to Midwest Community Bank, how can I help you?
Customer: I’d like to change my mailing address on file.
Smart Assistant: I can certainly do that for you, first let me verify your account. Please enter your last four digits of your SSN.
Customer: 1234
Smart Assistant: Great thanks! Now, please enter your mailing address currently on file.
Customer: 123 Any St. Middleton, NJ 12345
Smart Assistant: Perfect, now please enter your security passphrase.
Customer: Bananas
Smart Assistant: Excellent! Hello Mary, thank you for validating your identity. Now, let’s update your mailing address.
Customer: 456 State St. Apt 2 Chicago, IL. 60001
Smart Assistant: OK, please let me validate that address in our database. I’ve found this address: 456 STATE ST. APT 2 CHICAGO, IL. 60001–1234. Is this correct?
Customer: Yes.
Smart Assistant: OK, I’ve gone ahead and made that change. You will receive a confirmation letter in the mail. Is there anything else I can help with today?
Customer: Nope, that’s it!
Smart Assistant: Great, thank you for being a Midwest Community Bank customer! Have a great rest of your day.
This is an efficient conversation that doesn’t have any diversions or irrelevant questions from the customer. When the designer is planning out this dialog, she needs to build in the branding of the bank along with an empathetic tone. These are some basic elements that will help improve the customer experience.
Also, as part of the overall conversation design, it’s important to plan out the various conformations that will be necessary. In our example, we confirmed the SSN, passphrase, current and new mailing address.
Even though only the happy path is shown here, it is vitally important to plan for diversions or “branches” of the conversation. What if the customer changed their mind and wants to talk to a human agent? Or, what if the customer has a different question about their account?
The full conversation design needs to take into account these possibilities. However, since it is not possible to anticipate every utterance, the NLP platform must have AI-driven algorithms to be able to map possible responses. Over time, the self-learned and trained dialog will improve and that’s the job of nextgen chatbots.
The key takeaway is that “Conversation Design” is the foundation of a successful smart assistant. It should fully support the business use case and demonstrate a complete happy path along with some possible diversions.
