EDU.01
Tutoring Agents
AI tutors grounded in your own course material, with citations and a teacher-set hint policy so they guide rather than hand over answers. Welfare concerns go to designated staff.
Open pageIndustry
AI in education,
with the teacher in charge.
Tutors, courses, admissions. A person decides. AI in education means software that tutors from your own course material, drafts courses for teachers to edit, and answers the administrative questions students and applicants ask. Grades, admissions and progression stay with a named person.
Last reviewed 30 October 2026 · Engineering guidance, not legal or regulatory advice.
In context
AI in education means software that tutors from your own course material, drafts courses for teachers to edit, and answers the administrative questions students and applicants ask. Grades, admissions and progression stay with a named person.
Where it breaks today
01
General chatbots do the homework
A general-purpose assistant answers the question a student types, which is often the assignment itself. That is the cognitive off-loading teachers worry about, and it is a design choice, not an inevitability: a tutor can be built to ask before it tells.
02
Teachers write the same material twice
Lesson plans, differentiated worksheets, quiz banks and accessible versions of the same unit are rebuilt by hand each term, inside an LMS that holds the source but cannot reshape it.
03
Admissions offices answer one question a thousand times
Deadlines, documents, fees policy, visa letters, accommodation. The answers exist in a handbook; the queue exists because nobody can search it at eleven at night during the application window.
OUR APPROACH
Support for learning, not a substitute for it
The benefits of AI in education are real where the system is narrow: a tutor that only knows this course, an authoring tool that only rewrites what the teacher wrote, an admissions assistant that only quotes the published policy. The pros and cons debate is mostly about broad tools used without that boundary.
Regulation draws the same line. Under the EU AI Act, systems that decide admission, evaluate learning outcomes, steer a student to a level of education or monitor students during tests are listed as high-risk in Annex III, and emotion recognition in educational institutions is prohibited outright. Children's data carries extra weight under GDPR and, in the UK, the ICO Children's Code; in the US, FERPA and COPPA apply. So we build the assistance layer and keep the determination with staff. Whether a specific system falls into a high-risk category is for your institution to decide, and we help document the reasoning.
HOW IT DEPLOYS
PHASE 01
Choose one course or one queue
We start with a single module, a single programme's admissions questions or a single authoring workflow, and collect the material it will be grounded in.
PHASE 02
Connect to the LMS and SIS
Moodle, Canvas, Blackboard, Google Classroom or Teams for learning; PowerSchool, Ellucian Banner, Tribal SITS or Slate for records. Single sign-on, role-scoped access, nothing copied that does not need to be.
PHASE 03
Teachers review, then widen
Staff test it against the questions students really ask, including the attempts to get answers handed over, before a second course or queue is added.
What we build
01
Grounded tutoring agents
An AI tutor that answers from the course material with a citation, asks a guiding question before giving a worked answer, and shows teachers where a class is getting stuck.
SEE TUTORING AGENTS →02
Course and curriculum authoring
Drafts of lesson plans, quiz banks and accessible variants from a teacher's own outline, exported as SCORM, xAPI or QTI so they land back in the LMS.
SEE COURSE AUTHORING →03
Admissions and student operations
An assistant that answers applicants and students from published policy, collects documents and routes anything about welfare, fees hardship or an individual decision to staff.
SEE ADMISSIONS →04
Retrieval over your own material
The common foundation: a retrieval system over handbooks, syllabi and course content, so every answer can be traced to a page a member of staff wrote.
SEE RAG SYSTEMS →Controls we implement
NO AUTOMATED GRADING DECISIONS
NO EMOTION RECOGNITION
NO TRAINING ON STUDENT DATA
ROLE-SCOPED ACCESS
CITATION ON EVERY ANSWER
AI DISCLOSED TO STUDENTS
These are engineering controls we build into the architecture, not certifications. QQuantum.ai holds no security certifications, and the assessments your institution requires remain yours to run — see our EU AI Act notes.
Education use cases
-
-
EDU.02
Course Authoring
AI course authoring from a teacher's own outline and sources: lessons, quiz banks and accessible variants exported as SCORM, xAPI or QTI. Nothing reaches students without a named approver.
Open page -
EDU.03
Admissions & Student Operations
An admissions and student-services assistant that answers from published policy and chases documents in the SIS or CRM. It never scores or ranks applicants; decisions and welfare go to staff.
Open page
Questions
Questions about Education & EdTech.
How is AI used in education?
The useful uses are narrow ones: tutoring that answers from a specific course with citations, drafting lesson materials and quiz banks for a teacher to edit, translating and adapting material for accessibility, and answering administrative questions from students and applicants. The contested uses are broad ones: general chatbots that write assignments, automated grading of open work, and monitoring students. We build the first group and keep decisions about students with staff.
Can AI replace teachers?
No, and the systems we build are not designed to try. A tutor can explain a step at midnight and repeat it patiently, but it cannot judge whether a pupil is struggling for reasons outside the lesson, set expectations for a class, or take responsibility for an assessment. What it can do is take repeat explanation and material preparation off a teacher, so more of the working day goes to the parts only a person can do.
Is AI good or bad for education?
It depends on what the tool is allowed to do. Research and reporting cited in the current debate point to the same risk: students who use a general assistant as a shortcut can perform well on homework and worse on closed-book tests. A tutor built to ask before it tells, grounded in the course and visible to the teacher, is a different product from an answer engine. Design choices decide which side a deployment lands on.
Does the EU AI Act apply to AI in schools and universities?
Yes. Annex III lists as high-risk the AI systems used to decide access or admission, evaluate learning outcomes, assign a level of education or monitor prohibited behaviour during tests, and Article 5 prohibits emotion recognition in educational institutions. Tutoring and authoring tools that leave those decisions with staff usually sit outside the high-risk list, but the classification is your institution's call. We document the design so the reasoning can be reviewed.
How do you protect student data?
Student records stay in the systems that already hold them, read through role-scoped access rather than copied into a new store. Models run on infrastructure you control or behind processor terms that forbid training on your data. Under-18 users get the stricter treatment GDPR and the UK Children's Code expect, and in the US FERPA and COPPA obligations stay with the institution, which signs the data agreements.
