Turning a fragmented learning journey into one connected product system
GRAFIX Education is a SaaS learning platform where planning, lessons, assignments, progress and mentor feedback previously lived in separate experiences. I connected them into one system and introduced AI-assisted checks with clear escalation to mentors.
Summary
Objectives
- Connect planning, lessons, assignments, feedback and progress into one learning journey.
- Reduce repetitive mentor review work through AI pre-checks and clear escalation.
- Surface Pro value when learners need continued feedback and support.
Role
Senior Product Designer
Period
2022–2026 · 12 major releases
Team
Senior Product Designer · Product Manager · Tech Lead
2 Front-end · 2 Back-end · QA Engineers
Key Skills
- Product Strategy & Prioritization
- Product Design
- Information Architecture
- Interaction Design
- Interactive Prototyping
- User Research & Product Analytics
- Design Systems & Design QA
- AI Interaction Design
Impact
↑ 21%
Plan upgrades
↓ 25%
Drop-off in core learning flows
↓ 20%
Manual review and support effort
Three problems shaped the redesign
Learning continuity: planning, lessons, assignments and progress felt disconnected and dated. Funnel data showed drop-off between key steps, so keeping context and the next action visible became the first hypothesis.
Feedback loop: repeatable checks depended on manual mentor review, and mentors kept repeating checklist-level issues. AI pre-checks with mentor escalation could speed up revisions and reduce manual effort.
Pro value: upgrade prompts appeared outside moments of need, while trial feedback linked Pro value to feedback access — showing Pro after its value became visible should increase plan upgrades.
A clearer entry into learning
An explicit empty state explains why no course is active, then turns programme discovery into one next action: browse, compare and start.
Course page: one source of truth
The course page combines access windows, pace, schedule, modules and assignments around one clear continuation point — the active lesson.
Focused lesson workspace
The player, chapter timeline and lesson list stay in one workspace, so learners keep course position while moving through the lesson.
Contextual upgrade and learning momentum
Course progress and pace establish momentum before a contextual Pro prompt explains exactly what continues: AI reviews and mentor checks.
Assignment submission without ambiguity
Task requirements, assessment criteria, attempt limits and the review sequence are visible before upload, reducing submission errors and uncertainty about what happens next.
AI-assisted review with human accountability
AI pre-checks surface concrete checklist fixes before the next attempt; mentor discussion stays in the same review thread for judgement and context.
Design system as shared infrastructure
Reusable component states and documented usage turned the design system into practical implementation guidance rather than a visual inventory.
Post-release feedback supported the direction
Validation inputs: post-release learner and mentor interviews, support feedback, upgrade behaviour and review workload.
The strongest improvements came from connecting the journey: making next actions clear, showing Pro when its value became visible and shortening the wait for useful feedback.
“I upgraded as soon as I watched the feedback video during the trial. That was when I understood why the plan was worth it. Honestly, I probably should’ve bought it from the start.”
“The new version is just much easier to use. I always know where I am, what I’ve finished and what I’m supposed to do next. And honestly, it looks much better too.”
“I was pretty sceptical about the AI feedback at first, but it’s actually useful. It catches things I missed and helps me improve the work while I’m waiting to hear back from my mentor.”
“I’m seeing fewer submissions with the same basic issues. I can spend more time on the idea and the student’s reasoning instead of repeating the same checklist feedback.”
What I learned
From screens to systems
The hardest part was not redesigning individual screens; it was keeping the learning journey coherent while Product, Engineering and mentors solved different parts of it. I learned to treat the next user decision — not the screen — as the unit of design.
Making collaboration concrete
I moved between research, product metrics, interaction design and technical constraints, using prototypes to turn abstract discussions into decisions. Bringing engineers in early improved feasibility and making trade-offs explicit helped the team move faster.
What I would improve
I would instrument behaviour and AI quality earlier. We validated the main flows, but earlier signals around confidence, false positives and mentor escalation would have reduced post-release assumptions.