Handbook / Proof / OPIT
OPIT: AI tutoring at scale. −60% professor support. +40% student progression.
Cohorte's founder architected the AI tutoring system at OPIT, the LLM-native accredited online university. Production system, presented at Microsoft Milan in May 2025. The existence proof that the methodology installs AI inside an accredited institution and moves outcome metrics.
The story in one paragraph
OPIT is the LLM-native accredited online university serving students across Europe and North Africa. Their pedagogical model treats AI as a first-class teaching assistant rather than a tool students must avoid. Cohorte's founder architected the system that turned that pedagogical commitment into a production reality: a tutoring agent that answers student questions with citation-grounded retrieval, faculty-in-the-loop verification, curriculum-bounded answers, and observability. Outcomes are documented: −60% professor support load on routine questions, +40% student progression rate, −80% exam-related support tickets.
The four layers of the architecture
Curriculum-bounded retrieval
The tutoring agent answers from a curated, curriculum-bounded knowledge base, not from the open web. Every answer is sourced from course material, faculty-approved references, and verified primary sources.
Citation-grounded answers
Every answer includes the source citation. Students see where the answer came from. The agent cannot generate uncited content; it returns "I don't have a verified source for that" rather than confabulating.
Faculty-in-the-loop queue
Faculty review a sample of agent responses weekly. Flagged responses (low confidence, contested topic, novel question) are escalated for faculty review before reaching the student. The agent learns from corrections.
Production monitoring
Continuous monitoring on response quality, escalation rates, faculty satisfaction, student outcomes. The observability layer is what made the −60% / +40% / −80% metrics auditable.
The numbers in context
| Metric | Value | What it means |
|---|---|---|
| Professor support load | −60% | Faculty time spent on routine student questions (definitions, syllabus clarifications, foundational concepts) dropped by 60%. Faculty time redeployed to higher-judgment activities. |
| Student progression rate | +40% | The percentage of students completing modules within the expected window, measured course-by-course. The agent's 24/7 availability removed a friction point in async learning. |
| Exam-related support tickets | −80% | Tickets to student services regarding exam preparation, format, scheduling. The agent absorbed the policy-and-process questions that previously required a human. |
| Citation coverage | ~98% | Of generated responses, ~98% include verifiable citations to course material. The remainder are flagged for faculty review before delivery. |
| Faculty satisfaction | Documented | Faculty surveys at month 6 and month 12 are positive: workload reduction is real, pedagogical concerns addressed with the curriculum-bounded design. |
Methodological note. All metrics measured against the pre-deployment baseline. The methodology for measurement was reviewed by OPIT's institutional research function and is auditable.
What this means for a peer institutional buyer
The OPIT system is the existence proof that AI tutoring can be done inside an accredited institution without breaking academic integrity, faculty trust, or pedagogical quality. The architecture and the operating discipline are the methodology Cohorte teaches in our Higher Education programs.
Directly relevant
The architecture (citation-grounded retrieval, faculty-in-the-loop, curriculum-bounded answers, observability) carries directly. The HE catalogue walks through how to apply it.
User-facing AI analogue
OPIT is the closest analogue for AI talking directly to end users at scale. The verification primitives carry to customer-service AI in FS, audit-research AI in PS.
Brand-voice analogue
OPIT's "answer in the voice of the curriculum" pattern is the closest existing analogue to hospitality's "answer in the brand standard" pattern.
The reference
Riccardo Ocleppo is the founder and CEO of OPIT. He is the engagement sponsor for the AI tutoring system and is the named reference contact for prospective Cohorte enterprise customers in higher education.
Reference calls are scheduled within 5 business days of NDA signature. 30 minutes, run by Riccardo directly, no Cohorte representative on the call. Each named reference is capped at 1 reference call per quarter.
The public reference. OPIT and Charafeddine presented the AI tutoring system at Microsoft Milan in May 2025 as a reference deployment. The public talk covers the architecture, the outcomes, and the operating discipline.
The OPIT architecture, applied to your institution.
Start with the Pilot. Or scope AI Readiness for HE. Reference call with Riccardo Ocleppo available within 5 business days of NDA.