[My Role]
UX Designer
[Team]
CLO, Director, POs, Manager, UX Designer
[Platform]
Web · Mobile · Voice
[Project Overview]
Stakeholders asked us to rethink learning for 300k+ employees. Not a better LMS, but a new answer to where, when and how learning happens. We brought ambient AI ideas to a product team that had never worked with them, designed under real pressure with a backend team waiting on our specs, and shipped.
[My Role]
We were a team of two designers, and I worked on every part of the product with my partner, from early flows to final screens. We discussed and shaped each flow together.
A few ideas started with me:
Interview agent: modelled on a real interview call, with a time bar at the top so learners keep track of time while speaking.
Dynamic content cards: cards that adapt to the learner instead of staying fixed.
Timely nudges: a nudge when the learner goes idle, bringing them back at the right moment.
Chat dashboard: a view of everything a learner has done in a conversation, so they don't have to scroll back through the chat.
I also built the design system, which was outside the original scope.
[Problem Statement]
[The core tension]
“Employees are expected to grow continuously, but the systems that exist for learning interrupt the very work they’re supposed to support.”
The result: Training gets postponed, certifications get rushed, and nobody can tell if any of it works.
[Problem Space]
From the CLO brief, org data and desk research, four failure modes kept showing up, and they fed each other.
No time, no space
Back-to-back meetings make learning feel like a second job. “I’ll do it later” becomes never.
Low psychological safety
Fear of judgment stops people from asking basic questions. Gaps stay hidden and widen quietly.
One size for everyone
Same module for a new grad and a senior manager. Same format, same pace, same irrelevance.
Pushed, not pulled
Learning assigned top-down has no internal momentum. Without visible impact, it competes with real work, and loses.
“The real problem isn’t time. It’s that learning never feels like it belongs inside the workday.”
Internal synthesis: team brainstorm, from CLO brief and desk research
[Archetypes]
The Overloaded Deliverer
Mid-level, client-facing · 5–8 years exp.
“I want to grow but I genuinely can’t find a free hour. And then the course expires.”
Delivery pressure
Client-facing
Time-poor
High performer
[Goals]
Learn skills that immediately apply to current projects
Grow without sacrificing delivery quality
Feel recognised for development effort
[Frustrations]
Learning slots conflict with client commitments
Content is too generic to feel immediately useful
Progress resets when a module expires
The Silent Doubter
Junior · 1–3 years exp.
“Everyone else seems to know this already. I don’t want to be the one who asks.”
Fear of judgment
Junior
Eager to grow
Self-sufficient learner
[Goals]
Build confidence without exposing gaps publicly
Get answers without needing to ask people
Learn in a space that feels safe and private
[Frustrations]
Group sessions where basic questions feel embarrassing
No one-on-one way to check understanding
Imposter syndrome amplified by public assessments
[Process]
No time for user interviews. A backend team waiting on our specs. Every decision counted.
[01 Stakeholder briefing]
Aligned with the CLO on goals and what success looks like.
[02 Desk research]
Studied Zero UI, ambient AI and learning in the flow of work.
[03 Team synthesis]
Mapped archetypes against learning moments and picked 5 priority scenarios.
[04 Education & alignment]
Ran sessions so the team understood AI in learning before any screen existed.
[The Zero UI Concept]
This wasn’t a brief we received. It was a concept we introduced, first to ourselves, then to a product team that had never thought about AI in learning this way before.
[The original vision: before any screens existed]
“A developer goes quiet for 20 minutes. The system detects the pattern and asks: ‘Looks like you might be stuck, want a hint?’ No app. No login. No interruption.”
We called it Zero UI. Not no interface, just zero effort.
🎯
Contextual detection
🔔
Ambient nudge
“Client call in 30 min. Want a quick briefing?”
💬
Conversational help
Coding help, explanations & intv. prep, all by talking.
[The Pragmatic Pivot]
The full vision hit real constraints: budget, technology, timeline
The full vision needed real time behaviour detection and deep integrations we had no budget or time for. So we asked: what’s the closest thing we can build now?
The answer was a voice portal. Not invisible, but frictionless.
[Our Framework]
Context
Triggered by real work
moments, not calendar
slots.
Continuity
Builds across sessions
without starting over.
Closure
Clear endpoints that
create real sense of
progress.
Critique
Safe feedback without
fear of peer judgment.
[Design Decisions]
Each transformation maps to the 4Cs. These aren’t feature additions. They’re architectural decisions
about where learning lives in an employee’s day.
One-size-fits-all learning paths
To
A hyper-personal tutor that adapts to you in
real time
Generic interview preparation
To
Role-specific simulations with actionable
feedback
Standard assessments
To
Personalised evaluations with clear
improvement paths
Unguided practice sessions
To
AI-assisted labs with real-time coding support
Idle waiting time
To
Micro-learning moments that maintain
momentum
Loss of motivation mid-course
To
Smart, timely nudges that re-engage at the right
moment
[Outcome]
The voice portal launched and is currently live for users.
5000
Beta
users
25
Scenario
Pathways
∞
Moments of
Learning
[Version 2: My Proposal]
v1 was built under constraints. This is what it looks like without them.
What changed in v2
Side-by-side workspace for multitasking
Centralized access to all learning windows
More intuitive and contextual captions
Embedded mentor assistance during learning
What stayed the same
The 4Cs as the core design filter
Zero UI as the long-term north star
[Key Learnings]
Constraints are a design tool
Working under real time pressure and dependency produced sharper decisions. Not having time to second- guess forced clarity, and the voice portal pivot was better for it.
Design education is part of the job
Getting the product team to understand Zero UI before any screen existed was as important as the screens themselves. Without it, the concept wouldn’t have shipped at all.
Frameworks outlast the project
The 4Cs are still being used as a filter six months later, by a team making decisions I’m not in the room for. That’s the most durable output of the whole project.
