AI Design Guidelines

As you design Alexa experiences with artificial intelligence (AI), keep the following guidelines in mind.

AI is the backbone of Alexa

People use Alexa to help them schedule their day; check their homework; tell them a fact, joke, story; or simply play a game or listen to music. Alexa has become highly effective with these tasks and is quickly becoming embedded into the fabric of our everyday life. Alexa’s AI can now help reduce development time) and increase customer engagement. It can also help create more ambient, natural, and flowing conversations. The AI handles an infinite number of possibilities, enabling Alexa to manage a topic-switch as it occurs.

Less is more

When you design your dialogs with AI in Alexa Conversations, you no longer need to provide all possible dialog variations. You can let Alexa’s AI tap into the knowledge of the world and extrapolate more diverse conversations for your skill. All you need to do is provide a set of sample dialogs to give some direction to Alexa’s AI. For more details, see  about dialog writing for Alexa Conversations, see Dialog Design for Alexa Conversations.

 

By doing less and activating the tools that Alexa’s AI provides, you can enable your skill to engage more naturally with your customers. You can enable interactions between skills without making users remember skill names (with Latent Goal Discovery) or without having users have to say the invocation name.

Automate your dialog creation

Alexa’s AI tools require your help to ensure accuracy and the right context. You can use the dialog evaluation tool to automate your dialog creation. You provide feedback and choose the right prediction. It is critical to be diligent and accurate with your responses because your responses have a direct impact on how your final Alexa experience works.

Explainability

While Alexa’s AI can follow the unchartered path of a conversation, it can be helpful for you to assist users to understand Alexa's responses. Your users might want to know how Alexa came to a particular conclusion or answer. Your users should be able to ask questions regarding the decisions that Alexa’s AI made in your skill. This level of explainability helps your users continue to build trust with Alexa. Sometimes you might find it difficult to explain some decisions. It is a good practice to explore other ways of explaining Alexa's responses, for example, by showing prediction-confident levels in a gameified way.