Here’s how Apple describes it in the patent:
User input can be received and in response to receiving the user input, the user input can be displayed as a first message in the GUI. A contextual state of the electronic device corresponding to the displayed user input can be stored. The process can cause an action to be performed in accordance with a user intent derived from the user input. A response based on the action can be displayed as a second message in the GUI.
In case you find that a little hard to understand (I know I did), this means the digital assistant (Siri), will perform an action in response to your input in a text conversation, then respond to the text.
The most obvious use for this is environments where talking to Siri aloud is impractical — a very loud environment where Siri can’t hear you, like a sporting event; or a very quiet one where you can’t speak, such as a library (though definitely not a movie theater, you heathen).
This might also be helpful for hearing-impaired users or those with speech impediments who might have trouble interacting with the audio-focused assistant.