Generative interfaces that make the next step clear
2 min read
Generated content can change from one request to the next. Navigation, ownership, and the meaning of the primary action still need to be understandable. A user should know which parts are a proposal and which parts have already changed.
In an editable workspace such as KIMEO, generation is followed by deliberate changes on a canvas. That distinction offers a useful starting point for any interface that mixes AI suggestions with precise user control.
Give changing content a stable frame
Keep navigation, the current task, and the primary action in predictable places. Generated content can vary without making the entire workspace rearrange itself.
A person should be able to recognize what has changed since the last interaction. Group new content and preserve useful context.
Make the proposed action inspectable
Show a preview before an action changes a document or sends information elsewhere. Let the user edit the underlying inputs and understand the scope of the change.
Choose controls that fit the work. A table is useful for comparing rows; a form is useful for collecting structured decisions; a conversation is useful for clarification.
Design a complete set of states
Include loading, partial results, missing evidence, and recovery in the design. These states should offer a useful next step in plain language.
Test the experience using long answers and empty results as well as the polished example. Flexible content needs deliberate boundaries.
Put this into practice
Explore the related work and delivery scope.
- Related project: KIMEO
Inspect the published product workflow and delivery context.
- Explore the relevant service
See the engagement scope and start a specific project conversation.
Give generated content a usable product frame
Share example outputs and the actions a user must be able to inspect, change, or approve.
Discuss an AI interface
