--- title: "An AI Agent in Education: Student Support and Course Administration" description: "Most questions reaching a student office repeat, and the answer sits in a document nobody reads. Where an agent helps and where it has no business." canonical: "https://interfase.sk/en/news/an-ai-agent-in-education-student-support-and-course-administration" lang: "en" last_updated: "2026-11-02" --- # An AI Agent in Education: Student Support and Course Administration > Most questions reaching a student office repeat, and the answer sits in a document nobody reads. Where an agent helps and where it has no business. Published: 2026-11-02 - Most enquiries to a student office are repeated questions about deadlines, conditions and documents — exactly what an agent handles. - The agent must not assess work or decide on exceptions; those decisions affect a person and belong to the teaching staff. - Answers have to come from the institution's valid documents, not the model's free reasoning — otherwise you get binding but untrue statements. - The fastest return is in deadline reminders and course enrolment, not in the teaching content itself. Before the semester starts, the student office receives hundreds of emails. When enrolment opens, when it closes, what to submit, where to find the timetable. The answers sit in a policy nobody reads, because it runs to forty pages and is hard to search. This is one of the few cases where the benefit of an AI agent is measurable almost immediately — provided the line is drawn correctly. ## What the agent does well **Repeated questions about how study is organised.** Enrolment and withdrawal deadlines, conditions for completion, what to submit when suspending study, how credits are recognised. The answers are in the institution's documents and change once a semester. **Enrolment on courses and exam dates.** Finding free capacity, registering, withdrawing, offering an alternative. An ordinary administrative operation with no judgement involved. **Reminders.** An approaching submission deadline, the end of the withdrawal period, a missing document. This is the cheapest feature with the most visible effect — it cuts both the number of enquiries and the number of missed deadlines. **Navigating materials.** "Which lecture covered X" is a question the agent answers from the course material faster than a student can page through it. ## Where the line sits The agent **must not assess submitted work** and must not decide on exceptions, credit recognition or resit dates. Those decisions affect a person directly and belong to the teaching staff or the student office. The second line is subtler: the agent must not **predict** how something will turn out. "Will I pass if I get 12 points on the test" either has an unambiguous answer from the published conditions, or none at all. ## Answers need a source This matters more in education than in commercial operations, because an answer about study conditions is taken as binding. If the model invents a deadline that does not exist, the harm falls on the student rather than the institution. So: **the agent answers only from valid documents**, states what each answer is based on, and when it has no source it says so and hands the enquiry to a person. How to prepare the material so this works is covered in [how to prepare company data for AI](/blog/ako-pripravit-firemne-data-pre-ai). ## Student data has its own regime Applications, grades, reasons for suspending study and medical documentation are personal data, and some of it falls into a special category. The agent should see only what a given enquiry requires, and access gets filtered at the data layer rather than by instructing the model. We described the basic principles in the piece on [AI agents and GDPR](/blog/ai-agenti-a-gdpr-na-co-si-dat-pozor-pri-spracovani-osobnych-udajov). ## It does not replace a teacher Worth saying out loud, because expectations tend to be inflated. The agent will not explain the material better than the person teaching it, and that should not be its job. Its contribution is **taking away the administration** that eats the time teaching staff and the student office need for real work. A good sign of a correct deployment: organisational questions have dropped while subject-matter consultations have stayed level or gone up. ## What to measure - **Share of enquiries handled without** the student office. - **Number of missed deadlines** before and after reminders were introduced. - **Share of answers with no source** — how often the agent correctly said it did not know. - **Time to answer** during enrolment periods, when the load peaks. ## Where to start With the one area where questions repeat most — usually enrolment and the organisation of the semester. The material there is bounded, correctness can be verified, and the effect is visible within two weeks. If you are working on this, [get in touch](/kontakt). It builds on our [AI and automation](/riesenia/ai) work and on [AI and automation](/riesenia/ai); examples are in our [case studies](/referencie). --- Web version: https://interfase.sk/en/news/an-ai-agent-in-education-student-support-and-course-administration llms.txt: https://interfase.sk/llms.txt