Suggested Prompts
Provide pre-written prompts or starter actions that help users begin quickly.
Core Concept
Suggested Prompts are pre-written, clickable prompts or actions that help users start using AI quickly. They act like a shortcut into the experience, especially when the user does not know what to type first. Your document defines them as “smart and context-aware preset actions” that jumpstart engagement, reduce hesitation, and teach system capability without a tutorial.
In practice, this pattern works by showing a few relevant starting points such as:
- “Summarize this”
- “Rewrite more professionally”
- “Compare with last quarter”
- “Draft a client update”
This matches broader human-AI guidance: example prompts and empty-state suggestions reduce blank-screen friction and help users form a better mental model of what the system can do.
Problem it Solves
This pattern solves the blank input problem. Many users open an AI tool and immediately wonder:
- What can this do?
- What should I ask?
- How do I phrase it?
Your doc explains this clearly: blank screens feel intimidating not because users have no ideas, but because they do not know how to begin. That confusion can kill adoption before the first action happens.
It also solves a system problem: poor first prompts often lead to poor first outputs, which makes the product feel weak even if the model is strong. Public AI UX guidance consistently treats onboarding prompts, examples, and starter suggestions as a way to improve first-run success.
When to Use
Use Suggested Prompts when:
- the interface starts with an open text box,
- the product has many possible use cases,
- users may be new to AI,
- prompt quality strongly affects output quality,
- you want to guide discovery without forcing a rigid workflow.
This pattern is especially strong in chat-first products, copilots, search experiences, support assistants, writing tools, and productivity tools. Your document also notes that prompts should be low-commitment, one-click, and embedded in the right context.
When Not to Use
Do not rely too much on Suggested Prompts when:
- the workflow is already structured and the next step is obvious,
- expert users prefer direct control,
- the task is repetitive and templates or forms work better,
- too many suggestions would clutter the UI,
- suggestions could bias users into narrow or low-value behavior.
In those cases, Structured Input or Prompt Scaffolding is usually a better pattern than open suggestion chips.
User Value
Functional benefits
- helps users start faster
- reduces friction to first action
- improves first-prompt quality
- teaches product capability through examples
- speeds up common tasks with one-click shortcuts
Psychological benefits
- reduces anxiety
- lowers cognitive load
- increases confidence
- makes AI feel approachable instead of intimidating
- gives users a feeling of momentum and control early in the experience
UX Principles Applied
This pattern connects most strongly to these principles:
- User Value First — it helps users reach value immediately.
- Clarity of Capability and Limits — good prompts show what the system is good at.
- Familiarity and Learnability — examples teach by showing.
- Constructive Defaults — the system gives users a useful starting point.
- Trust Calibration — realistic examples help users understand the AI without overpromising.
Design Guidelines — Do & Don’t
Do
- Show 3–6 prompts at a time; your doc specifically warns against overwhelming the interface.
- Make prompts specific and outcome-focused, not generic.
- Tie prompts to the current context, file, page, or workflow.
- Use prompts to demonstrate common user goals.
- Keep them one-click and low-commitment.
- Refresh or personalize prompts over time when possible. Your document’s “future play” section points toward adaptive, context-aware, team-aware prompts.
Don’t
- Don’t show too many suggestions at once.
- Don’t use vague prompts like “Ask me anything” as the main guidance.
- Don’t suggest actions the system cannot do reliably.
- Don’t leave prompts static forever for repeat users.
- Don’t make prompts so broad that they confuse rather than guide.
Real Examples (Components Used)
1. ChatGPT-style starter prompts
A common pattern is showing example tasks directly under the input field so users can click and begin. Your doc explicitly references this kind of use-case suggestion as a strong onboarding pattern for newcomers.
Components used
- Prompt input field
- Prompt suggestion chips
- Example prompt cards
- Placeholder guidance text
2. Inline productivity shortcuts
Your doc gives a simple example: instead of forcing users to write a full prompt to improve an email, a one-click action like “More diplomatic” can get them started instantly.
Components used
- Inline suggestion button
- Context action chip
- Accept/retry controls
3. Context-aware future prompts
Your doc’s future vision is strong here: when a document is open, the product could suggest prompts like “Compare this with last quarter’s results” or “Draft a client update from these notes.”
Components used
- Context-aware prompt chips
- Document-aware suggestion bar
- Team/shared prompt library
Business Impact
Suggested Prompts can improve:
- activation,
- first-session success,
- engagement,
- and retention,
because they reduce the time between landing in the product and getting a useful result. They also lower support burden because users need less onboarding help. Your document frames them as a way to jumpstart engagement and teach capabilities without tutorials, which is exactly why they matter commercially.
They can also create product differentiation when prompts are personalized, contextual, or team-aware rather than generic.
KPIs
Useful KPIs for this pattern:
- first prompt submission rate
- time to first useful output
- click-through rate on suggested prompts
- percentage of users who use a prompt chip vs typing from scratch
- first-session completion rate
- activation rate
- new-user retention
- satisfaction score on first output
These metrics are a practical extension of the onboarding and adoption role this pattern plays.
Engineering Notes
At the simplest level, Suggested Prompts are easy to ship: they can be hardcoded examples in the empty state. But stronger versions need:
- prompt configuration logic,
- context detection,
- analytics on prompt usage,
- experimentation support,
- personalization,
- and possibly shared prompt libraries for teams.
If prompts depend on the current document, user role, or product state, the system also needs reliable context retrieval so suggestions stay relevant and safe.
Future Outlook
Your document is especially good here: it says Suggested Prompts will move from static hints to adaptive, evolving, context-aware shortcuts shaped by the user, the current task, and even team workflows.
That direction matches the broader industry move toward more guided and personalized AI experiences, where the system helps users discover the best next action instead of waiting for a perfect prompt.