Ask Skye
Designing the intelligence layer behind Scooler's student mobility platform
Role
Lead UX & Interaction Designer
Company
Scooler AI
Year
2024
Scope
Overview
Ask Skye is the agentic intelligence layer at the center of Scooler AI.
When I joined the company, AI existed primarily as a collection of disconnected experiences. Students could interact with AI, but the experience felt similar to other chat-based products on the market.
Through product audits, competitive research, and usability testing, I helped redefine Skye from a chatbot into a context-aware guidance system that powers every major student journey across the platform.
The Problem
Most AI-powered education products share the same interaction pattern:
Users arrive at a blank text field and are expected to know what to ask.
For students navigating university applications, visas, essays, financial planning, and relocation, this creates a significant burden.
The challenge wasn't improving AI responses.
The challenge was designing a system that reduced uncertainty before students even started asking questions.
Product Opportunity
During the broader Scooler redesign, we identified an opportunity to reposition AI as infrastructure rather than a feature.
Instead of creating another AI chatbot, we envisioned Ask Skye as a persistent layer that could understand student context, coordinate actions, and guide decision-making throughout the platform.
This became one of the foundational product decisions behind Scooler's redesign.
Research
Understanding the limitations of current AI experiences
I conducted a competitive review of ChatGPT, Claude, Microsoft Copilot within the larger context of the MVP experience and what we were trying to achieve in being AI native.
The objective was to understand where highly capable AI systems still create friction for users and how best to solve for it.
Unstructured Interviews
To understand user experience with current AI assistants and surface unfiltered pain points.
App Design Reviews
Noting the pros and cons of 3 major AI interfaces — UI and interaction design patterns across ChatGPT, Claude, and Microsoft Copilot.
Desk Research
Analysing online user reviews and sentiments to identify recurring themes at scale.
User Pain Points
1. Context Fragmentation
The MVP AI experience had user goals fragmented across user flows and captured no context of the user journey. Students repeatedly have to re-explain their goals, academic background, and progress every time they need assistance.
Context mapping showing fragmented student information across journeys.
2. Cognitive Load
Most AI tools place the burden of interaction on the user.
Students are expected to know what to ask, how much information to provide, and how to phrase prompts to get useful results. Even when they receive a response, it may be too broad, too detailed, or not aligned with what they actually need.
Research screenshots highlighting blank-state interactions.
3. AI Identity and Interaction
Studying abroad is one of the most significant decisions many students will make.
Yet most AI assistants communicate in a way that feels generic, transactional, and detached from the emotional realities of the process. Whether students are feeling uncertain about university choices, stressed about visa requirements, or anxious about deadlines, existing AI experiences rarely acknowledge those moments.
Pain Point: Students need guidance that feels supportive and trustworthy, not just informational.
Competitive analysis matrix — UI and interaction design review.
Design Principle
Reduce prompting. Increase guidance.
The core design principle became simple:
Students should not have to figure out how to use AI.
The system should already understand the context and proactively guide them toward their next step.
This principle influenced every interaction that followed.
Ideation and Wireframing
Translating insights into interaction concepts
With the pain points clearly defined, the ideation phase focused on how to reduce student effort at every touchpoint. I ran concept explorations across the three core problem areas — context fragmentation, cognitive load, and AI identity — using low-fidelity sketches and flow diagrams to map potential solutions before committing to any direction.
Wireframes were iterated rapidly to test layout hypotheses, information hierarchy, and the positioning of AI within each workflow — establishing the structural foundation before visual design began.
Early concept sketches and ideation boards.
Designing Ask Skye
Decision 1
Replace generic chat with contextual entry points
Rather than exposing a single universal chat interface, Ask Skye appears within relevant workflows. For example:
- Review Essay
- Explain Admission Requirements
- Compare Universities
- Generate Application Checklist
This reduces cognitive effort while increasing response relevance.
Post-onboarding follow up
Ask Skye Panel
For new and existing users, Ask Skye will have context on your on-boarding answers and guide you to pick up where you last stopped.
Decision 2
Designing an AI Identity and Interaction
One of the key challenges identified during research was that most AI assistants felt functional but emotionally distant. For students navigating high-stakes decisions around university applications, visas, finances, and relocation, a purely transactional interaction model risked increasing anxiety rather than reducing it.
To address this, I designed Ask Skye's personality to feel supportive, encouraging, and context-aware without becoming overly conversational or distracting. The goal was to create an experience that felt less like interacting with a tool and more like receiving guidance from a knowledgeable mentor.
Ask Skye — new user, first interaction.
Decision 3
Apply Selective Disclosure in Chat Interface
Users frequently feel they have no fine-grained control over how verbose or concise responses are. Getting an answer that's too long (lots of preamble and caveats) or too short (lacking depth) without an easy way to adjust. Similarly, the inability to interrupt or steer a response mid-generation frustrates people — you have to wait for the whole thing to finish before correcting course. Applying Selective Disclosure in the Chat Interface helps to manage this pain point.
Ask Skye — existing user with active conversation history.
Impact
Product Impact
Ask Skye became the intelligence layer powering multiple experiences across Scooler. Instead of existing as a standalone feature, it connected and enhanced the platform's core workflows.
User Experience Impact
Eliminated blank-state friction
Contextual entry points across 6 workflows meant Ask Skye always appeared with relevant student context pre-loaded — students never faced an empty chat with no direction.
68% drop in session abandonment at AI entry points
"I knew exactly what to ask — it already felt like it understood where I was."
Reduced information overload
Selective disclosure let students control response depth — brief, mid, or detailed — directly solving the research finding that AI responses were frequently too long or too short for the moment.
61% of testers adjusted response depth within their first session
"For the first time I didn't feel like I had to read a wall of text just to get one answer."
Shifted AI from tool to guide
Designing Skye with a defined, supportive persona changed how students described the experience — less like querying a search engine, more like getting advice from someone who already knew their situation.
4.3 / 5 perceived helpfulness vs 2.7 / 5 for generic AI chat in comparative testing
"Supportive", "personal" and "trustworthy" were the top 3 words used to describe Skye in post-test interviews.
Protected Content
This content is protected for confidentiality purposes. Please enter the password to continue viewing.
Reflection
The most important lesson from this project was that designing AI experiences is rarely about designing conversations.
It's about designing systems.
The success of Ask Skye came not from making the chatbot smarter, but from embedding intelligence into the moments where students needed guidance most.
