Overview
How might we design contextual ai suggestions so people can trust and act on AI output?
When to use
- Ideal for complex applications, design tools, and productivity platforms where contextual AI suggestions improve feature discovery and workflow efficiency.
When to skip
- Context too ambiguous to suggest safely.
- Suggestions on every keystroke causing jitter.
- Actions that need explicit user data consent first.
Rules
Same three suggestions regardless of selection.
Suggestion chip covering the text being edited.
No keyboard path to accept or ignore.
Running expensive AI on every hover.
Evidence
| Product | Implementation |
|---|---|
| Figma | AI actions on selected frames and text layers. |
| Cursor | Composer suggestions based on open files and selection. |
| Notion | Ask AI on highlighted blocks and databases. |
| Linear | Issue summarize and triage prompts in context. |