FitWhen to use it — and when not
Use it when
- First session of any AI product
- New workspaces, projects, agents
- Features that need data before they are useful
Skip it when
- Returning users — get out of their way
AnatomyThe parts of the pattern
- Honest headlineWhat this space will become.
- Starter templates2–3 concrete jobs with setup time.
- Sample runPreview on clearly labelled sample data.
- Primary actionConnect real data or start from a template.
- Escape hatchStart blank for experts.
GuidelinesDo & don’t
Do
- Lead with the gentlest, highest-value first task.
- Label sample data as sample.
- Show what success looks like before asking for setup.
Don’t
- Fake activity or metrics to look busy.
- Open on a blank chat box with "How can I help?" only.
- Ask for five integrations before the first result.
In productionHow it looks in a shipped product
In the wildReal-world examples
GammaLovablev0Notion AI
Products named for reference only — no affiliation, and the demo above is an original illustration, not a copy of their UI.
For engineersImplementation notes
- Ship templates as data with pre-baked sample inputs and cached sample outputs — instant and free to show.
- Instrument time-to-first-good-result as the activation metric.
- Seed sample data in a separate, clearly flagged namespace that is easy to delete.