All Case Studies
Ab
Abstrabit Technologies
Case Study
🎓 MACU · Higher Education · USA

Prospective Students Chose
Faster Universities.
Now MACU Answers in Seconds.

We built an AI that reads any transcript format, cross-references 8,000+ equivalency codes, and delivers instant credit estimates — with self-healing institutional memory.

Client
MACU
Industry
Higher Education
Location
USA
Org Size
500+ Employees
Users
Registrar + Students
30 min → sec
Per transcript evaluation
8,000+
Course codes cross-referenced
99%+
Extraction accuracy
Self-heals
Learns from every correction
The Problem

Students Don't Wait 4 Weeks.
They Commit to Whoever Answers First.

Time to credit estimate — from transcript submission
MACU (after AI)
Seconds
✅ Instant
Competitor A
3–5 days
⏳ Waiting
MACU (before)
2–4 weeks
❌ Lost
⏱️
30+ Min per Transcript
Staff manually read each course, searched equivalency databases, and drafted evaluations line by line.
30–45 min per evaluation
🎲
Inconsistent Decisions
Two staff members evaluating the same transcript reached different conclusions. No consistency guarantee.
Same transcript, different answers
🚪
Knowledge Walked Out
Edge-case decisions lived in people's heads. When staff left, all nuanced judgment calls disappeared.
Zero institutional memory
The Solution

3-Tier AI Decision Engine
That Gets Smarter With Every Upload.

30+ min
Manual — per transcript
→
Seconds
AI — any format
①
State Regents CEP Matrix
8,000+ officially approved course equivalencies. Direct match = instant resolution.
✓ match → resolved ✕ no match → tier 2
↓ course not in state matrix
②
MACU Internal Rules
University-specific mappings maintained by staff in Google Sheets. No developer needed.
✓ match → resolved ✕ no match → tier 3
↓ no internal rule exists
③
Learnt Historical Decisions
Every staff correction is saved. The system never makes the same mistake twice — institutional memory that survives turnover.
✓ prior decision → applied ⚠ novel → flag for staff
The Proof

Messy PDF In. Structured Data Out.
Decision Trace — Course by Course.

oklahoma-state-transcript.pdf ✓ EXTRACTED
Raw PDF — mixed format, floating header
Course NameGradeCrDept
ENGL 1113 — Composition IA3ENG-101
MATH 1513 — College AlgebraB+3MAT-112
⚠ HEADER: Spring 2023 ↓——floating
HIST 1483 — US History to 1865A-3HIS-201
PSYC 1113 — Intro to PsychologyB3PSY-101
BIOL 1114 — Biology w/ LabA4BIO-105
👁️ Claude Vision parsed in 3.8s · 5 courses · 1 floating header resolved · 0 errors
Structured output → 3-tier engine
{
  "student_id": "TRN-2024-04471",
  "source": "Oklahoma State University",
  "confidence": 0.99,
  "courses": [
    { "code": "ENGL 1113", "grade": "A", "cr": 3 },
    { "code": "MATH 1513", "grade": "B+", "cr": 3 },
    { "code": "HIST 1483", "grade": "A-", "cr": 3 },
    // ... 2 more courses
  ],
  "ready": true
}
✓ 5 courses · grading normalised · sent to engine 3.8 sec
📋 Course: PSYC 1113 — Intro to Psychology
3 credits · Grade B · Oklahoma State University
1
State Regents CEP Matrix
Not found — OSU psychology not standardised at state level
✕ No match — falling through
2
MACU Internal Rules
Rule found: Intro psychology → MACU PSY-101 (3 cr, Gen Ed)
✓ Match — resolved at tier 2
✅ PSYC 1113 → MACU PSY-101 · 3 credits · Gen Ed tier 2
🧠 Learnt Database — Growing Institutional Memory
ART 2213 (UCO) — Digital Art → 3 cr · ART-EL staff
CS 2334 (OU) — Prog Structures → 3 cr · CS-201 auto
COMM 1113 (TCC) — Speech → 3 cr · COM-101 staff
NURS 2412 (OCCC) — Pharma → 4 cr · NUR-EL auto
Staff Dashboard
Pending review4 flagged
Auto-resolved38 today
Learnt DB247 total
Student Portal
Your credits~47 transfer
Gen Ed12 of 18 cr
EstimateInstant
Results

Backlog Eliminated. Students Answered Instantly.
Memory That Never Leaves.

~99%
Faster per evaluation
(30+ min → seconds)
Zero
Knowledge lost to
staff turnover
Real-time
Backlog eliminated —
was 2–4 weeks
Evaluation Time
Before
30+ minutes
After
Seconds
Enrollment Backlog
Before
2–4 weeks
After
Near real-time
Decision Consistency
Before
Staff-dependent
After
Standardised 3-tier
Tech Stack
Claude Sonnet Vision React + Vite Python + Flask MongoDB Zustand Material UI Google Sheets API Google Drive API AWS EC2 Nginx + Gunicorn
Metric
Before
After
Evaluation time
30+ minutes
Seconds
Enrollment backlog
2–4 weeks
Near real-time
Decision consistency
Variable by staff
Standardised 3-tier
Institutional memory
Lost with turnover
Persisted in learnt DB
Student self-service
Not available
Instant portal
The Compounding Advantage
Most automation delivers a one-time gain. This system gets better with every upload. Every staff correction becomes permanent institutional memory — knowledge that never walks out the door.
Abstrabit Technologies