project overview
Designing AI Students Can Actually Trust
An AI-powered academic experience that makes complex tasks easier — without asking users to blindly trust the AI.
concept video
See the idea in motion
concept walkthrough
solution highlight (1)
One-Step Intelligent Course Planning
Natural-language interaction
Students describe what they need in plain language — "plan next semester around no Friday 8ams and more data electives" — and the system reads their credits, conflicts, and ratings automatically.
5 steps reduced to 1
Check handbook, verify credits, match schedules, pick courses, submit for approval: five steps collapsed into one real-time generate-and-confirm step.
- 1. Check handbook
- 2. Verify credits
- 3. Match schedule
- 4. Pick courses
- 5. Submit for approval
solution highlight (2)
Resource Provenance & Trust Mechanism
AI Assistant · Answer
Internal
Student Handbook, course database, financial systems
External
Industry reports, public academic sites
How to actually build trust into an AI product?
project context
Where this started
Deloitte came to SCADpro with an open brief: explore how AI will reshape higher education. Over the course of one summer, an 11-person, 7-major student team transformed that brief into research, a point of view on trust, and a three-part product system under the guidance of Professor David Meyers.
research process
A 4-phase research strategy
To fully understand the relationship between AI and higher education, I helped plan and drive a 4-dimensional research strategy.
Primary Research
Interviews and contextual inquiries with students, faculty, and university leadership on AI attitudes, adoption, and ethics.
Secondary Research
Industry trend reports, Deloitte Insights, and policy documents.
AI Survey Field
A competitive audit of existing AI tools and development activity.
SME Inquiry
In-depth interviews with experts on data security, privacy, and policy.
research insights
5 insights that shaped everything after
Secondary research alone surfaced five insights that ended up steering the rest of the project.
Privacy & distrust
Users don't trust AI because they can't see how it handles their data — this is the seed for the Resource Provenance mechanism.
AI-powered student experience
AI shows real value across four everyday touchpoints which freeing up real human capacity.
Digital divide & employment risk
Students without exposure to AI tools risk being disadvantaged in a job market that increasingly expects AI fluency.
Policy fragmentation
Schools are torn between protecting core skills (prohibition) and embracing what's coming (integration).
Automation vs. augmentation
AI doesn't simply replace people; however, it raises efficiency through human-AI collaboration. -- Per McKinsey
personas
2 very different comfort levels with AI
Christina
Art history student
"AI helps me save a lot of time on my research, but I still can't fully trust it. I always end up double-checking the facts and examples before I actually use them..."
About
Motivation
Christina is motivated by AI's ability to save time and streamline information gathering, allowing her to focus more on her studies despite needing to verify the information.
Pain Points
AI is time-saving but struggles with trust due to inaccuracies, requiring verification of information and examples.
Core Need
- Accuracy: A reliable AI tool that provides trustworthy information without the need for constant verification.
- Efficiency: A solution that significantly reduces the time spent on research and information gathering.
- Ease of Use: An intuitive AI tool that simplifies the process of finding and summarizing relevant materials.
Lynn Patterson
University professor
"AI is just a tool and does not inherently make people lazy; how it is used depends on the person using it..."
About
Motivation
Lynn is motivated by the potential of AI to assist students while ensuring they remain engaged and pursue their interests independently.
Pain Points
Concerning that AI's current limitations in handling complex tasks might prevent it from being a fully effective educational tool.
Core Need
- Effective AI Integration: Tools that enhance learning without hindering student motivation.
- Advanced Capabilities: AI that can handle more complex tasks to better support educational goals.
- Balanced Use: Guidelines for using AI in a way that maintains high educational standards.
problem statement
Even the smartest AI can't enter a high-stakes decision if no one trusts it.
How might we solve the trust relationship between users and AI?
brainstorming session
From "what can AI do" to "why won't anyone let it"
We started out exploring what AI could do for higher education. The deeper we went, the clearer it became that the real blocker wasn't efficiency, it was trust.
The full 11-person team put every idea on the board, from chatbots to AI teaching assistants, then dot-voted the ones worth developing further.
From there it was rounds of narrowing together to grouping ideas by who they served (students, faculty, staff) and what problem they actually solved.
Once the direction felt right, we storyboarded the concept walkthrough frame by frame to pressure-test the flow before building anything.
design exploration
Before converging on the final system, the team explored three different directions for tackling the efficiency-and-trust problem.
AI Assistant: mobile app
Concept
A smart campus companion app for students, covering real-time campus life: finances, transportation, dining, and student organizations.
Solves for
Better communication efficiency and stronger management of the student lifecycle.
CLM: a university's own language model
Concept
A custom AI model trained on the university's own data and knowledge base, supporting the full workflow across students, faculty, and leadership.
Solves for
Data security and privacy, with accurate, traceable decision support for both teaching and administration.
A smart device: hardware interaction terminal
Concept
A hardware device combining natural language processing and IoT, similar to an Amazon Echo or a smart desk scanner, embedded directly into classrooms and research spaces.
Solves for
A visible, always-present source of trust that people in the room can actually point to.
design decision moment
How the three directions became one system
Cut: Smart Device & AI Assistant as standalone products
- Hardware: expensive and slow to build, didn't solve the core trust problem on its own
- Generic app: pulled focus away from the high-value teaching workflows Deloitte and SCADpro actually cared about
Commit: a university-specific model (CLM)
- Trained on the university's own data, fixing accuracy and privacy at the root
- Set the exact direction for the EduAI Dashboard
As a design leader, I pushed the team to commit to one direction instead of shipping three half-finished ones. The dashboard won out because it was the direction the research itself was already pointing to.
"Our own library and course records are more accurate than anything a generic AI pulls from the open web, students just don't know to trust them yet."
SCAD Librarian, faculty interviewdesign mocks share
Why this system, and what it looks like
dashboard interactive interface
The turning point was a conversation with a SCAD faculty member during research: the university already runs its own internal databases: course records, financial aid, the official Student Handbook. That single insight reframed the whole system. If trustworthy internal data already exists, the real design problem isn't generating answers, it's showing people which of an AI's answers come from that trusted internal source versus the open web.
1. SSO Login & Identity Verification
2. Centralized AI Dashboard
3. Conversational AI & Query Workspace
4. Citation & Provenance Modal / Verification
How the four screens connect — sign in securely (1), land on one dashboard that gathers your tools into a single view (2), ask questions directly through the AI chat (3), and trace any AI answer back to its source before acting on it (4). Each screen hands off to the next, turning a handful of separate tools into one guided path for the decisions that matter most.
Modular Dashboard Architecture & Progressive AI Exposure
Problem — Traditional academic portals scatter schedule management, advisor booking, and degree planning across isolated sub-pages, causing cognitive overload and high bounce rates.
Core decision — We paired a task-oriented grid with an ambient AI anchor.
Primary Focus
High-frequency tasks: next class, office hours booking, schedule which sit as visual cards on the main canvas for one-click access.
AI Integration
Instead of burying AI inside a standalone chat tab, we embedded the AI Virtual Tutor and Degree Planner directly into the grid, so students can invoke AI help without leaving their workflow.
Tradeoff considered — We chose a slightly denser grid over an ultra-minimalist blank search bar. Research has shown that students need immediate visual anchors such as upcoming classes or a booking list; rather than having to guess what to ask the AI from a blank slate.
A dashboard was the clearest way to make that distinction visible. Unlike a voice device or a wearable, it has room to visually separate internal and external sources right on the interface. That's exactly what the Resource Provenance mechanism needed: to turn trust into something you can see, not something you have to assume.
project results
What we can measure so far
Collapsed the traditional manual flow that “check handbook → verify credits → match schedule → select courses → submit for review” into a single real-time generate-and-confirm step, sharply cutting the cognitive and operational load on students.
A transparent, visual provenance mechanism lets students and faculty judge the credibility of information at a glance which cutting the time spent verifying AI-generated answers by roughly 60%.
At our final presentation on campus, Deloitte's leadership team responded with strong, positive feedback on both the depth of the research and the final concept, specifically calling out the trust-first framing as a standout part of the work.
Exceptional Insights & Strategic Vision
“It demonstrated profound user research and sharp observational insights. I like the research direction about how we think about academic workflows with a visionary direction.”
Deloitte UX AnalystExpanding the Campus Ecosystem
“Besides a stellar student-facing portal, this project paves the way for holistic campus integration that connecting faculty, advisors, and students in a unified AI ecosystem.”
Deloitte Engineering SpecialistUnlocking AI Transparency
“Addressing the ‘black-box’ trust issue through source-provenance mechanisms and interactive data simulation is a brilliant touch. It transforms complex AI outputs into verifiable, actionable decisions.”
Deloitte Product Principalwhat's next
Looking back, a few months later
Static bar chart showing AI-generated learning plans.
Interactive dynamic timeline with real-time GPA simulation and drag-and-drop course testing.
Opaque recommendation outputs — a “black-box” of uncertainty.
Inline citation & provenance UI providing 1-click access to source documents (e.g., degree manuals, grade history).
Extend the web dashboard into physical touchpoints using the AI Desktop Device (voice UI) and Secure AI Fob (ambient nudges) for hands-free campus navigation.
AI Desktop Device
Secure AI Fob
self reflection
What this project taught me
I learned a lot of practical experience in user research, UI/UX, and product design through this collaboration with Deloitte, under the guidance of our faculty advisor.
-
Gratitude for the Deloitte Team
Deep appreciation for the support from our Deloitte team members, whose presence and insights at our final presentation were invaluable.
-
AI Exploration
Excitement about sharing findings from our exploration of AI's role in academic environments.
-
Design Thinking Insights
Gained valuable knowledge on design thinking and future expectations through interviews with Deloitte staff.
-
Remote Contributions
Heartfelt thanks to our Deloitte contact for their impactful questions and perspectives during remote Zoom sessions, which shaped our research and conclusions.
want the full process
Research decks, interview notes, and the full design process live in the processbook.
open the processbook ↗