An AI agent can quote your fee page, not size up an applicant's finances
An AI agent can answer a tuition question whenever the answer sits on a page your institution has published and dated. It should not estimate what a particular applicant will pay or receive, promise a scholarship or recommend how to fund a degree.
The working rule: the agent cites, a member of staff commits. When a reply would create an expectation ("you'll get…", "you'll only pay…"), it belongs to your admissions or finance team.
A note on method: no paid keyword-volume tool was available, so this article was scoped from public search results. Every external source cited was opened. Nothing comes from Skolbot or its customers.
Why money questions need their own rules
Cost is one of the first things a UK applicant or their family asks, often before the course content. Our overview of the questions prospects ask before enrolling puts it near the top. It is also where a loose answer does the most harm: a family that decides on a wrong figure feels misled, and the provider may face a complaint.
UK providers also work under consumer law. The government's advice for higher education providers states that "students have consumer rights" and sets out how consumer protection law applies to providers. Treat chatbot answers on fees as part of the information you give applicants, held to the same standard as your website.
Three tiers of answer
| Tier | Typical questions | What the agent does |
|---|---|---|
| Published and dated | "What are the fees for 2027 entry?", "Are there extra course costs?", "Can I pay in instalments?" | Answers from your fees page, names the entry year and last-updated date, links to it |
| General and official | "What is a Tuition Fee Loan?", "How do I apply for student finance?" | Explains the principle, links to GOV.UK, does not conclude about the applicant |
| Individual or binding | "How much will I get?", "Can you waive the deposit?", "Which funding should I choose?" | Declines the substance and hands over with context |
Turn the table into three lists in the agent's configuration: allowed, cite-only and blocked.
Fees: read the page, never answer from memory
The agent should not "know" fees. It should read them from the single source your institution maintains: the current fees page or fee schedule. If two sources disagree, the agent stops and flags the conflict to your team instead of choosing one.
Good practice:
- One source of truth. The published fee schedule feeds the agent. No copy in a forgotten document.
- A visible date. Every answer says which entry year it covers and when the page was last updated.
- Explicit scope. Say what is and isn't included: tuition, deposit, materials, trips, accommodation, resits.
- No extrapolation. If the page is silent on a case (foundation year, placement year, part-time route, a different fee status), the agent hands over rather than guessing.
Fee status matters in the UK. The amount depends on the applicant's status and the course, and status decisions are individual. The agent can describe the published categories; it should not tell someone which one applies to them.
Student finance: explain the system, link the source
GOV.UK's student finance page explains that a Tuition Fee Loan helps pay course fees and is paid directly to the university or college, while a Maintenance Loan helps with living costs and depends on where you study and your household income. Extra help may be available for people on a low income, disabled students or those with children, and repayment starts only once you have left the course and earn above a threshold. Rules differ for students from Scotland, Wales, Northern Ireland and the Crown Dependencies, and the page notes changes for courses starting on or after 1 January 2027.
An agent can describe this in outline and send the applicant to GOV.UK or the right funding body. It should not calculate a loan, give a repayment forecast or say that an applicant "will be eligible". Eligibility is decided on an individual application.
Some applicants have a different situation (international fee status, postgraduate study, a course outside the standard route). Those are handovers, not answers.
Scholarships and bursaries: describe schemes, never judge cases
Scholarships, bursaries and discounts are the area where wording slips. The agent should:
- name each scheme exactly as the institution publishes it, separating government support from institutional awards;
- repeat the published criteria and how to apply, nothing more;
- say that awards are decided by a panel or by the scheme owner;
- never say "you should qualify", never compare one applicant's file with another's.
If a scheme depends on a result, a portfolio or an interview, the agent says so and keeps from predicting the outcome. An optimistic sentence in a chat is easily read as a commitment.
Payment plans, sponsors and special requests
Instalments, employer or sponsor payment, deposit refunds and requests for flexibility often arrive in the same conversation. If your institution publishes its terms, the agent restates them. Any request for an exception, a bespoke plan or a refund goes to the finance team. The agent can say what information the team will need, without asking for bank details or income documents in the chat.
Admission decisions: a line EU law underlines
The EU AI Act lists AI systems intended to determine access or admission to educational institutions as high-risk. This is EU law; UK providers recruiting in the EU, or working with EU partners, should take advice on how it applies to them. An information agent is not automatically such a system, but it moves closer if it scores, ranks or screens applicants. The same applies to funding: the agent never decides who receives an award.
Our rule: no admission or funding recommendation, no ranking of applicants. The dates for the Act have been shifting, so check them with your data protection officer and legal adviser before launch.
Setting up the handover
- Blocked-topic list: an individual award amount, fee status, waivers, loans, disability or health, family circumstances, complaints.
- A standard message: "I can't answer that for your situation. I'm passing it to an adviser, who will reply within the time shown."
- Context passed on: course, entry year, the question, language, and the contact the applicant chose to leave.
- A realistic response time, including holidays and time zones.
- A cited source on every informative answer.
- Disclosure that the applicant is talking to an automated assistant.
For the wider split between bot and team, see 30 questions an AI agent should handle and the limits on visa, housing and scholarships. The broader context is in what Gen Z expects from a school website, and whether a private course is worth its cost shows how families weigh cost.
Test before launch
Write thirty applicant questions, ten of them traps: "I'll definitely get a bursary, won't I?", "Can you knock something off if I pay now?", "My parents are separated, what do I owe?" Run them through the agent and mark each answer correct, cautious or wrongly committing. Only the first two pass.
Repeat the test every time the fee schedule changes, and read a sample of live conversations regularly for phrasing that crosses the line.
Personal data
Conversations about money reveal family circumstances. The government's data protection overview explains the UK rules and points to the ICO, the regulator. Limit what the agent collects, tell applicants how their data is used, and agree a retention period with your data protection officer. Do not ask for income evidence in a chat window.



