Last week I attended the first meetup from the Voice Assistant Technology NL group. It was a great meetup, where I got some new inspiration about communication with voice assistants. The focus of both talks was based on communication between your users and the assistant.
The communication between your users and the assistant needs to be how humans communicate with each other. When starting the development of your app for the Google Assistant, you may tend to answer all questions like a robot. This will not be a smooth conversation and is not preferred by Google. For voice assistant, Google already made a conversation design document on how interactions should be handled by the assistant.
Answer like a human
When start developing an app for the Google Assistant, you might just return a robot text. Think about how real conversations with human take place, to make sure your app will return answers like a real human instead of a robot. This makes the conversation between your user and the assistant more comfortable.
There may be some cases in your assistant app, where you ask the user to give up an option. For example, if the assistant want to give the user an advice to buy a product as gift for a friend.
There are multiple ways to allow the assistant to ask for this. One of them is just an open question, something like “What category may the gift for your friend be?”. There is a problem that the user will say a category which is not defined, so it may be better to give some options to the user to make the user aware of limited product groups.
A better way to ask the user for a category, may be something like: “What category may the gift for your friend be? Some examples of categories that are available, are travel, sport or education.”.
Test the unhappy flow
If the Google Assistant is not able to answer the question of your user, it might return a weird answer like “I don’t understand, can you try again?”, but this doens’t give any feedback to your user which part of the question wasn’t clear.
When a request fails, try to be clear to the user what went wrong. For example, if the user is asking for a gift idea, and the assistant is asking in which category the gift may be, it could be that the user answers that question with a category that is not existing. If the assistant will answer “I don’t understand, can you try again?”, it’s not clear to the user what went wrong and the user will try again and makes the same mistake.
If the assistant answers the question with “Sorry, that category is not available. I’ve got gifts in the category sports, education and travel. Can you try again?”, it’s more clear to the user what went wrong with the previous answer.