The faculty-governed AI teaching platform

AI that faculty govern.

Students already use AI. Ungoverned, the strongest evidence says it makes them measurably worse. Tutela gives every course an AI tutor the professor governs and grounds in their own knowledge — one that teaches with hints, not answers.

Request early access See the evidence
Scroll
Ungoverned
−17%

Give ~1,000 students raw chatbot access and exam scores fall. It felt like help. It functioned as a crutch — answering the questions students were meant to learn to answer.

Bastani et al. · PNAS · 2025
Governed & grounded
2.0× learning

A professor-built tutor, grounded in the course's own material and giving hints instead of answers, more than doubled learning — in less time. The same design erased the harm.

Kestin et al. · Harvard · 2025
The variable was never the intelligence of the model. It was governance.
Whose knowledge it teaches from — and whether it teaches, or merely tells.
02 The evidence

The difference between AI that harms learning and AI that doubles it isn't the model. It's who governs it.

Ungoverned
−17% exams

Exam scores fell when students were given raw chatbot access, in a randomized trial of ~1,000 students.

Bastani et al. · PNAS · 2025
Governed
2.0× learning

Students learned more than twice as much, in less time, from a professor-built tutor grounded in the course's material.

Kestin et al. · Harvard · 2025
At scale
2yrin 6 weeks

A guided, guardrailed tutoring program produced roughly two years of typical progress in six weeks.

World Bank · WP 11125 · 2025
Sources / PNAS · Harvard / Scientific Reports · Stanford Tutor CoPilot · World Bank Policy Research
03 What makes it work

Governed intelligence, grounded in your knowledge.

i.

Governed by faculty

The professor governs the AI through a dedicated console — approving its knowledge, reviewing its answers, setting its policies. The AI never replaces the professor. Ownership turns resistance into leadership.

ii.

Grounded in approved knowledge

The tutor answers only from the professor's approved sources, cites the lesson it drew from, and declines what it hasn't been given. No open internet, no hallucinated authority.

iii.

Guardrails are the product

Hint-first Socratic teaching, assessment-safe modes, usage caps — all professor-configurable. Guardrails aren't friction added to AI. They're the reason it teaches instead of tells.

iv.

Engineered to teach

The course becomes a surface built for retention — animated processes where textbooks show static figures, with retrieval practice woven in. Bloom's tutor, made affordable, under faculty control.

04 The platform

The course students learn on. The AI faculty control.

Institution-owned

The living course

Lessons on an enhanced, scrollable surface — animated explanations, segmented video, and retrieval practice built into the reading itself.

  • Animated processes where textbooks show static figures
  • Spaced retrieval and self-checks woven into the flow
  • Owned by the institution, aligned to its curriculum
Faculty-owned

The governed AI tutor

A per-course assistant grounded in the professor's approved knowledge, taught to guide rather than answer, improved through faculty review.

  • Answers only from approved sources, and cites them
  • Hints-first; assessment-safe modes before exams
  • Named and governed by the faculty who teach it
· The differentiator

Faculty don't get around the AI. They govern it.

Every course's tutor answers to a professor's console: they approve its knowledge base, review and correct its answers, and set the rules it teaches by. When the AI layer is theirs, the loudest skeptics become the platform's champions.

Professor console · PHRM 240
Knowledge baseApproved · 42 sources
Answer policyHints-first
Pre-exam windowAssessment-safe
Flagged for review3 answers
Off-syllabus requestDeclined · logged
05 For deans, provosts & program leads

A defensible AI program you can stand behind in public.

Bans have not held, and generic "AI features" reproduce the exact chatbot that caused the harm. Tutela is the alternative institutions can adopt on the record.

A. Built to be measured

Designed from day one to produce controlled, publishable outcomes from your own cohorts. The institution generates its own evidence rather than importing a vendor's slide.

B. Faculty-adopted, not imposed

Because faculty own and govern their AI layer, adoption comes from the department, not a mandate from above. Ownership is the adoption strategy.

C. Grounded and assessment-safe

Grounding gives a defensible answer to "where did the AI get that," and assessment-safe modes protect the integrity of the work the coursework is graded on.

D. The standard, authored by educators

Deploy the program that lets your institution set the regional standard for AI in education — rather than consume whatever the chatbots decide to become.

AI that faculty govern.

Tutela is being built with founding university partners. If you lead a school, a program, or a course and want the standard for AI in education set by educators — let's talk.

Request early access
Hints, not answers.