How to Design AI Onboarding & First-Value UX

AI products fail the blank state more often than they fail the model. Users do not know what to ask, what the system can do, or why setup is interrupting them. First-value design is the craft of getting someone to a real success before you ask for identity, permissions, or payment.

This guide is a binder for that job. It stitches onboarding patterns like first success flow and example prompts, and the live Notion AI vs Lovable vs Manus comparison into one build playbook. Each pattern below shows a shipped product shot, then a referral to try the interactive demo on the pattern page.

What AI onboarding and first-value actually mean

First-value UX is the path from landing or empty state to the first outcome users recognize as success. Onboarding is everything that stands between those moments: education, preference capture, auth, permissions, and monetization.

The win condition is not "completed the tour." It is "produced something I would keep." Progressive disclosure and example prompts exist so users do not invent the product from scratch.

If you only read one distinction: setup that changes the first output can come early. Setup that only changes analytics should wait.

The core pattern: reach a first success fast

Start by defining the first success in product terms, then remove every gate that does not enable it. Notion AI skips a wizard entirely and activates at the cursor with categorized prompt starters.

Real-world example

Notion AI · No wizard. AI activates at the point of work with categorized starters. Full teardown

Pattern: First Success Flow, try the interactive demo on the pattern page.

Embedded products can often win with zero setup. Greenfield builders may need a short wizard so the first generation is not random.

Pattern spine: examples, empty states, and progressive unlock

First success is a system, not a single CTA. Shipped products compose empty states, starters, and progressive unlock so capability does not dump all at once.

1. Example prompts library

Show concrete jobs users can tap. Starters should run, not educate abstractly.

Real-world example

Notion AI · Categorized starters at the cursor. Jobs, not capability adjectives. Full teardown

Pattern: Example Prompts Library, show tappable jobs that produce a real first output.

2. Empty state that teaches by doing

Replace blank anxiety with one recommended action. Manus opens with outcome-framed role cards so the next step is obvious.

Real-world example

Manus · Outcome-framed role cards. Identity capture that points at a first task. Full teardown

Pattern: Prompt Starters, replace blank anxiety with one recommended next action.

3. Segmentation wizard when it changes generation

Lovable front-loads theme, identity, and company size before the builder because those answers change the first app. One question per screen with progress dots keeps the wizard tolerable.

Real-world example

Lovable · Theme first. Wizard steps that alter the first build, not vanity demographics. Full teardown

Pattern: Onboarding Progress Tracking, show progress when a short wizard truly changes first output.

4. Progressive feature unlock

Teach advanced power after first success. Interactive tours and locked features belong after the user has tasted value.

Real-world example

Lovable · Connectors after identity. Advanced surfaces wait until the stub exists. Full teardown

Pattern: Progressive Feature Unlock, unlock advanced power after the first successful outcome.

Three product bets: when setup happens

The same activation job produces three interfaces. Steal the funnel that matches how embedded your AI already is. Full table and steal rules live in the onboarding comparison.

Notion AI: zero-setup at the point of work

No wizard. AI activates inside the doc at the cursor. Steal this when your AI is embedded in a surface users already trust.

Real-world example

Notion AI · First value needs no setup when the host app already has context. Full teardown

Lovable: segmentation wizard before the builder

Four-step full-screen wizard for theme, name, role, and company size. Steal this when first generation quality depends on those answers.

Real-world example

Lovable · Role capture that changes the builder, not a survey for its own sake. Full teardown

Manus: outcome cards with mid-funnel pricing

Role cards frame outcomes, then Pro pricing appears before a task-ready composer. Steal the outcome framing. Treat mid-funnel paywalls carefully: users need enough taste of value to judge price.

Real-world example

Manus · Pricing after role framing. Monetize mid-funnel only if outcome is already vivid. Full teardown

Wizard, in-context activate, or paywall?

This is the highest-leverage product decision in first-value design.

• Activate in context when AI lives inside existing work. Notion AI is the reference.
• Run a short wizard when answers change the first generation. Lovable is the reference.
• Delay paywalls until users have tasted a success, unless your free tier still delivers a complete first job.

If onboarding asks for role, company size, and payment before a single useful output, you are selling trust debt.

Decision checklist

Prefer the shortest path to a keepable output. Friction that personalizes generation is good. Friction that only fills a CRM is not.

Anti-patterns to refuse

• Long tours before any AI output
• Empty composer with no example jobs
• Mandatory demographics that do not change generation
• Paywall before the user has tasted a complete success
• Feature dumps that teach vocabulary instead of a task
• No skip path on optional wizard steps

What to ship next

Pick a posture from the comparison, then define one first success and instrument time-to-that moment. Cut any step that does not move the metric.

1. Read the onboarding comparison and steal one activation posture.
2. Spec starters with example prompts and first success flow.
3. Add a wizard only for fields that change first output quality.
4. Cross-check the older AI onboarding pattern essay for empty-state and progressive unlock details.

Explore related reference

• AI onboarding UX compared
• Notion AI onboarding teardown
• Lovable onboarding teardown
• Manus onboarding teardown
• First success flow pattern
• Example prompts library pattern
• Onboarding Users to AI