Overview
How might we configure a useful first-run experience from a few goal and role questions instead of a blank composer?
When to use
- Lovable, Manus, and app builders where answers pick templates, stack, or starter project shape.
- ChatGPT custom GPT and multi-mode assistants where role and goal change defaults and examples.
- Enterprise or team products where setup selects connectors, memory, or workspace presets.
- Onboarding where wizard answers must actually change prompts, templates, or feature visibility.
When to skip
- Power users returning to a familiar tool who want skip-to-app immediately.
- Products with one obvious use case where a wizard invents fake choices.
- When answers cannot actually change configuration and the wizard is theater.
States
Design the question flow and what it changes, not only a decorative welcome screen.
- 01
Welcome
First run opens a short wizard. Headline states why questions appear and how long they take.
- 02
Answering
One question at a time: primary goal, audience or team size, experience level. Progress stays visible.
- 03
Reviewing
A summary shows chosen answers before apply. Back navigation fixes mistakes without restarting.
- 04
Configured
Answers map to real changes: templates, starters, connectors, or feature presets land in the product.
- 05
Skipped
Skip bypasses the wizard. Defaults stay neutral; the person can reopen setup later.
- 06
First success
A tailored starter prompt, workspace, or first task opens so the first win matches stated goals.
Key UX elements
The parts that must be present for a wizard to earn its questions.
Questions
Ask three to four high-signal questions.
Goal, role, and familiarity beat long surveys. Each answer should map to a concrete configuration change.
Progress
Show step count and time to finish.
Step 2 of 3 and a short ETA reduce abandonment. People should know the wizard is almost done.
Skip
Offer skip without penalty.
A clear Skip control must be visible on step one. Never trap new users in a mandatory survey.
Apply
Connect answers to real defaults.
Summarize choices, then apply templates, connectors, or presets. If nothing changes, the wizard is theater.
Starter
Drop into a first task that matches answers.
Pre-filled prompt, sample file, or workspace title should reflect what they said they came to do.
Revisit
Let people change setup later.
Store wizard output in settings so role or goal shifts do not require a new account.
Rules
Long multi-page surveys before any AI interaction.
No Skip escape on step one.
Collecting data that is never applied to prompts, templates, or defaults.
Locking the user into a persona they cannot change later in settings.
Evidence
| Product | Implementation |
|---|---|
| ChatGPT | Custom GPT and onboarding flows that ask how people want to use the product. |
| Notion AI / productivity suites | Setup wizards that pick templates from team or role goals. |
| Lovable / app builders | Role and project questions that frame the first generated app. |
| Manus / agent products | Role pickers and use-case choosers during first-run onboarding. |
