Education
AI tutors are going national. Assessment, retention and child safety have to keep up.
The lines that matter Enrolment conversion, retention, cost per student and staff time
Governments are funding AI tutors at national scale, while AI agents can now complete whole online courses on a student's behalf. The evidence shows that tools designed for learning help, and general chatbots can leave students worse off once the AI is removed.
Regulation centred on children is already in force in Australia, the UK and the US, and the EU's rules for education AI follow in December 2027. Schools, edtech providers and care organisations need to decide where AI may act, for whom, and under whose supervision.
- Enrolment and retention pressure on revenue
- Staff workload in teaching and administration
- Assessment integrity when agents can complete coursework
- Child safety and privacy obligations
- Student data spread across learning and administrative systems
- Staff capacity to redesign teaching and assessment
- Vendor contracts without clear data and age-assurance terms
- Equity of access to good AI provision
Where AI can move the numbers
Opportunities grouped by Agent Value Ledger category. They are possibilities to test, not promised results.
Revenue
Enquiry-to-enrolment conversion, retention, new learning products
Operating costs
Student services, marking support, course and timetable administration
Capital
Better use of facilities and timetabled space
Risk
Safeguarding records, assessment integrity, privacy compliance
The lines this sector is judged on
Universities and higher education
- Tuition revenue
- Enquiry response, application processing, offer conversion
- Student retention
- Early warning on disengagement and timely support
- Cost per student
- Student services, enrolment and course administration
- Academic staff time
- Marking support, course admin and research administration
Schools and school groups
- Enrolment and fee revenue
- Admissions handling for independent schools
- Teacher workload
- Planning, reporting and differentiated materials
- Administration cost
- Finance, timetabling and parent communications
- Safeguarding and compliance
- Record-keeping and policy reporting
Edtech and learning providers
- Customer acquisition cost
- Lead qualification and onboarding
- Retention and churn
- Learner engagement and proactive support
- Content cost
- Course production and localisation
- Support cost
- First-line learner and customer service
A student-retention agent at a private university
Assumptions: 12,000 students; average tuition US$15,000 a year; student services staff cost US$40 an hour.
| Line | How the agent moves it | Measured by | Working | Per year |
|---|---|---|---|---|
| Revenue | Disengagement spotted early, with advisers prompted to act | Retention against a matched cohort | 1 point higher retention × 12,000 × US$15,000, one year of tuition only | +US$1.80m |
| Operating cost | Routine student enquiries and case notes handled by the agent | Hours per enquiry; cost per student | 15,000 hours × US$40, counted only if redeployed or hiring is avoided | +US$0.60m |
| Admissions revenue | Faster responses to prospective students | Enquiry-to-enrolment conversion | Not counted until a full intake cycle | Excluded |
| Cost of the agent | Model usage, integration, adviser oversight, privacy and assurance | Total cost of ownership | Annual run cost | −US$0.50m |
| Net annual value | ≈ +US$1.9m, before any admissions benefit | |||
Hypothetical illustration of the method. The figures are assumptions, not a client result or a forecast.
- Decisions about learners stay with staff, with clear override
- Vendor standards for age assurance, data use and training on student data
- Assessment redesigned around supervised checks
Executive AI Strategy & Board Advisory
Start with a council or board briefing on AI strategy, then a retention and student-services diagnostic.
About this serviceEducation & Families in the Age of Agency
AI that acts can widen access to good teaching and care, but only if institutions decide where it may act, for whom, and under whose supervision.
Read the brief →AI tutors and assessment that still mean something
Situation. A school network wants an AI tutor for every student, while teachers are finding coursework completed by agents.
Our approach. An age-staged delegation map, vendor standards for age assurance and data use, and assessment redesigned around supervised checks.
What good looks like. Every student has a tutor built for learning, and the network's qualifications still prove what students can do.
Scenarios and worked ledgers are hypothetical illustrations. They do not describe specific clients.
Start with an executive conversation
A confidential discussion of your priorities, and where AI could move the numbers in your business.