UX Case Study
AI-Powered Face Recognition for Arivoo SMS
1 min vs 5 mins manual check-in
Manual attendance is costing teachers time, accuracy, and trust in the data they rely on.
Teachers spend 5–10 minutes per class taking attendance manually, disrupting learning flow.
Students mark attendance for absent friends, creating inaccurate records.
Large classes make it difficult for teachers to recognize all students, enabling proxies.
Manual logs are hard to analyze, making it difficult to identify attendance patterns.
Every class period, teachers lose up to 10 minutes and schools lose confidence in their own records.
Origin Story
The solution came from listening, not from a product roadmap.
Our engineering team was experimenting with face recognition technology for security applications.
During a school town hall, teachers shared frustrations about manual attendance taking up class time.
“What if we could use face recognition to automate attendance and give teachers their time back?”
The best product ideas come from the intersection of existing capability and a real, voiced user frustration.
Research Goals
Four research goals shaped the entire discovery phase.
Map current attendance processes and identify time-consuming steps in classroom routines.
Discover specific challenges with proxy marking, student recognition, and data accuracy.
Test if teachers and students are comfortable with camera-based attendance systems.
Assess technical accuracy requirements and address privacy concerns from all stakeholders.
Defining clear research questions upfront kept the team focused and prevented scope creep during a tight 4-week sprint.
Research Approach
A comprehensive multi-method strategy to validate the concept over a 4-week research sprint.
across 3 schools
observed live
school administrators
40 students + 6 teachers
Combining interviews, observation, and prototype testing gave us both the "why" behind the problem and early signal on whether the solution would actually be adopted.
Two distinct users, two distinct definitions of success.
Teacher Persona
Admin Persona
Designing for both personas meant the solution had to be fast enough for teachers and insightful enough for admins, two very different bars to clear.
Research Findings
Six findings that directly shaped every design decision.
Teachers value every minute. Saving even 3–4 minutes is considered highly valuable.
Teachers reported awareness of proxy attendance, with larger classes having higher rates.
Students and teachers were comfortable with photo-based attendance if privacy was ensured.
Admins want attendance trends, not just records. Pattern recognition is key for intervention.
Data storage location and access controls were the top concern across all stakeholder groups.
Must complete attendance in under 60 seconds to be adopted. Any longer feels like too much tech.
Privacy and speed were non-negotiable constraints, not nice-to-haves. The design had to prove both before users would trust the system.
How It Works
Teacher takes a single classroom photo.
Face recognition identifies all students.
System automatically marks present students.
Data syncs to the admin dashboard in real time.
Designed around one research constraint: done in under 60 seconds, or teachers won’t use it.
Designed for speed and simplicity.
One-tap capture for instant photo-based attendance.
Clear confidence indicators for each identified student.
Fast review-and-confirm step, minimal teacher effort.
Real-time data with a live sync indicator for admins.
Every screen reflects a research insight: minimal steps for teachers, transparent results for trust, and instant data for admins.
13min total
55sec total
What used to be a 13-minute administrative burden is now a 55-second step that doesn’t interrupt teaching.
Measurable Results
The tangible impact of AI-powered attendance.
The system didn’t just save time, it eliminated proxy fraud entirely and gave admins real-time insight they never had before.
AI-Powered Attendance
Transforming classroom efficiency through intelligent automation and thoughtful design.
From 13 minutes to under 1 minute per class
Face recognition makes proxy attendance structurally impossible
Near-perfect identification in real classroom settings
Every design decision traces back to a teacher interview, a shadowed classroom, or a student prototype session
This is what happens when AI capability meets genuine user need: a solution teachers actually use, admins actually trust, and students can't game.