The Actions on Google platform lets you build actions that allows users to get things done on a variety of experiences and devices. For example, a shoe store can provide actions to order items, check on an order delivery, learn about new products, or sign up for promotions and these actions can appear on surfaces such as Google Home, Android phones, and Google Allo.
Users will be able to discover and invoke actions in a variety of ways. Currently, the main way is through the Google Assistant, their own personal Google that lets them find, organize, and get things done in their world. You can extend the functionality of the Assistant by building apps that contain actions, the entry points into your app.
An action is an entry point into your app that defines two components, an intent, an object that describes the action, and fulfillment, the web service that defines the logic to fulfill the intent.
When users request an action in the Assistant:
- The Assistant asks Actions on Google to invoke the best app to fulfill the intent.
- Actions on Google sends a request to the app's fulfillment and receives a response that it hands off to the Assistant.
- The Assistant renders the response in its UI and displays it to the user. The conversation between the user and the app begins.
- The Assistant sends subsequent user input directly to the app and the app responds directly to the Assistant. This conversation continues until the app gathers the user input it needs to fulfill the intent.
Here's an example of how this works with an imaginary retail shop app named Sekai.
A user talks to the Assistant and requests an action, "I'm looking for sneakers from Sekai". The Assistant asks Actions on Google to find the best app to handle the action and receives a response back from Sekai.
The Assistant displays the response to the user and the conversation begins with the user. The Assistant waits for user input to occur and when this happens, the Assistant sends this request over to Sekai's fulfillment.
During the conversation, the Assistant brokers the conversation between the user and Sekai, and the two exchange information so Sekai can gather the information it needs to fulfill the intent.
Finally, the conversation ends when the user's intent to purchase some shoes is fulfilled or until the user wants to end the conversation.