People were leaving before seeing value.
Too many unfamiliar choices arrived before the first useful result.
Proficon Labs · Simpo AI
Making an AI website builder feel less like a black box.
Simpo AI helps people create a website with AI. The early onboarding asked people to make unfamiliar choices before they understood what the system would do for them. I redesigned that first session around clearer progress, smaller decisions and visible feedback.
- Role
- UX Designer
- Scope
- Research, journey design and UI
- Timeline
- Oct 2024 - Sept 2025
- Context
- AI website creation
- Team
- Product and engineering
- Read time
- 6 minutes
Mini case study
From “where do I start?” to a first useful website.
A quick, scannable view of the design work behind the Simpo AI onboarding reset.
Mapped the first session end to end.
Aligned product and engineering around a clearer path to the first website.
Found where progress disappeared.
Reviewed abandonment, traced decisions and benchmarked staged-creation patterns.
Made each step smaller and explainable.
One question at a time, visible progress, editable choices and earlier feedback.
Onboarding drop-off: ~65% → 13%.
The first session became easier to understand and finish.
AI earns trust by showing its intent.
Make the next decision clear, then let the person stay in control.
At a glance90-second summary
- The problem
People were leaving before they could see a useful first result.
- The constraint
The flow needed to make AI guidance clearer without taking control away from the person creating the site.
- My ownership
I investigated abandonment, mapped the first session and designed the new progression.
- The outcome
Directional onboarding drop-off moved from about 65% to 13%.
The original problem
People had to make decisions before they had enough context to make them.
The existing flow exposed several choices at once, including choices about the site and choices about what the AI should generate. That made the first session feel like setup work instead of a path to an outcome.
Design questionHow might the flow explain the next useful choice, show what the AI is doing and keep the person moving towards a first website?
What I investigated
I traced abandonment back to the moments where progress disappeared.
I reviewed the existing website-creation journey, looked for the decisions people met before receiving value and benchmarked familiar staged-creation patterns. The aim was to find where the interaction became unclear, not just where the interface looked busy.
What does the AI need from me?
Early prompts needed a clearer purpose so people could understand why each answer mattered.
What has happened so far?
The journey needed a visible sense of movement and an understandable next step.
Can I change this later?
People need reassurance that a first choice is a starting point, not an irreversible commitment.
When will I see something useful?
The flow needed to bring the first visible result closer to the decisions that created it.
What changed
I turned a broad setup flow into a sequence of small, explainable decisions.
The redesign focused each step on one question, kept the next action visible and used progress cues to show where the person was in the journey. The language shifted from system-first instructions to outcome-first prompts, with feedback that made the AI's contribution more legible.
The design decisions
- Break initial setup into a task-focused sequence
- Explain why each input affects the generated website
- Show progress and preserve a clear route to edit earlier choices
- Bring an early generated result into the journey sooner
~65% → 13%Onboarding drop-off
25%Faster task completion
ClearerTask progression
EarlierVisible AI feedback
The first session became easier to understand and finish.
The reported onboarding drop-off moved from nearly 65% to 13% after the journey was rebuilt around simpler choices and clearer task progression. This was a directional product metric from the team's existing analytics, not a controlled A/B test.
Reflection
AI needs to make its reasoning visible at the moment a person has to decide.
The useful shift was treating onboarding as a conversation with a clear destination, not a form that happens to have AI behind it. Every question had to answer a practical need: what am I choosing, why does it matter and what will happen next?
That became the standard I carried into later AI interaction work: show enough of the system's intent to build confidence, then get out of the person's way.
Explore the case studies behind the outcomes.
Customer journeys, partner operations, digital health and AI-assisted product flows.