- Land on a dashboard
- Work out which module fits
- Fill in a long profile
- Maybe find something useful
← All work
Case study / 05
HeiVelo
Designing talent intelligence as a calm career companion—not another dashboard.
RoleLead UX Designer
TypeAI product, in development
Timeline2025—now
The problem
One AI platform had to serve individuals, mentors, institutions, industries and community projects—without turning into a dashboard nobody understands.
What I did
Mapped five role models onto one product, replaced feature menus with intent-led entry, and made every AI recommendation explain itself.
Outcome
A unified product direction, goal-led onboarding and a design system for the build. The product is still in development.
A companion that starts with your goal.
Instead of a wall of features, HeiVelo opens with one question and a short list of what you can do today.
Five mental models, one product.
Each role arrives with a different first question. I kept one shared product language underneath, and gave each role its own entry words, priorities and first action.
- “What brought you here?”
- Understand the goal
- Show the next useful action
- Build the profile along the way
Value first, profile later.
The first useful outcome arrives before the profile is complete. Each step asks for one thing, and anything non-essential can wait.
- 01
Welcome
Sign in with one tap or email.
- 02
Intent first
“What do you want to do on HeiVelo?”—jobs, skills, guidance, talent, mentoring or community.
- 03
Role from goals
More than one goal? Choose which role to set up first.
- 04
Profile your way
Upload a résumé to fill it in, or build it step by step.
- 05
About you
Only the basics, with a clear progress bar.
- 06
Education
Current studies, in a few fields.
- 07
Verify identity
Needed to apply or connect—and it can wait.
- 08
Ready
Three next actions: courses, jobs, mentors.
Scores need an explanation.
The trust loop
Recommendations show their evidence and stay editable, so the AI never behaves like a black box.
- AI suggests
- Shows its evidence
- Person corrects or confirms
- Better next suggestion
Conversation with visible control
Voice and natural-language input reduce form fatigue; clickable choices, previews and undo keep it trustworthy.
Reflection, not ranking
A 360° profile supports reflection and fit. Drill-downs show the evidence and how to improve, rather than ranking people without context.
What prototype walkthroughs showed
People needed to understand why the AI suggested something.
→Show the evidence and make every recommendation editable.
Dense setup delayed the first useful moment.
→Reveal value early; build the profile progressively.
Five roles created inconsistent starting expectations.
→Role-specific entry paths on one shared product model.
One direction where there had been five.
Where it stands
HeiVelo is still being built. So far: five role models unified on one product, goal-led onboarding and return journeys, and a design system with detailed product screens.