Define intention types

In the overview of intention design, we mentioned that intentions can be classified according to different domains and dimensions.

Based on the clarity of user intentions, they can be divided into clear intentions and vague intentions. Generally, vague intentions can be more effectively understood and fulfilled through conversation; while users with clear intentions can achieve their goals through simple operations, such as clicking buttons or icons.
Based on the interaction purpose between the user and the system, intentions can be divided into information inquiry and task execution. Information inquiry intentions are mainly related to users' viewing and searching behaviors, reflecting the user's intent to obtain information; task execution intentions are mainly related to users' operation and management behaviors, reflecting the user's intent for the system to perform specific tasks or operations.

We found that intention classification is closely related to user behavior. The above information helps us better understand user intentions, thus designing interface interaction patterns that better meet user expectations.

Based on the quadrant graph of intent types and user behavior, if we overlay the existing intervention methods of mainstream AI products, we find that intent type + user behavior has a certain relationship with AI intervention methods.

Do is suitable for embedded systems: Mainly interface-based operations, occasionally triggering AI shortcuts, better suited for clear intent and behavior management.
Chat is suitable for standalone formats: Mainly natural language, with almost no interface operations. More suitable for those whose intent is ambiguous and behavior-based search is more important.
Chat+Do is suitable for assistant style: Natural language and interface operations work in a balanced way. Strong versatility makes it better suited for the intersection of the above two scenarios.

This does not mean that every intent can correspond to only one mode of interaction intervention. In practical applications, product designers need to choose the most suitable form of AI intervention based on specific scenarios and needs.