Education Centre Chatbot: From Enquiry to Enrolment

The seasonality problem

Enquiry volume at an education centre is not a straight line. For two or three weeks a year it runs several times higher than an ordinary week, and those are precisely the weeks in which decisions are made. No staffing rota can absorb that shape.

The consequence is predictable: during peak weeks replies slow down, and the parent or student writes to three centres the same evening. Whichever answers substantively first frames the comparison, so response speed becomes part of the decision alongside answer quality.

The second factor is timing. Admissions questions arrive in the evening and at weekends, because parents look after work and students look after class. The hours the centre is open and the hours the questions arrive do not overlap.

Questions that can safely be answered

These are most of the volume and all of them are factual — not one requires a judgement about a student.

• The course list, duration, sessions per week and lesson length.
• Price, payment methods, instalment options and discount conditions — as published, without negotiation.
• Start dates and when the next group opens.
• Schedule options — morning, evening, weekend — and which groups still have places.
• Age requirements, entry conditions and the documents needed.
• Branch addresses, transport, parking and whether online attendance is possible.
• Whether a certificate is issued and what it attests to, without inflating its status.

Qualification: four fields are enough

1. 1Who it is forThemselves, their child, or an employee. This changes the tone of the entire conversation and every subsequent question, so it should be asked first because everything else depends on it.
2. 2The goalAn exam, a job, university, or simply the skill. The goal determines course length and urgency — 'preparing for the June exam' settles the timeline immediately.
3. 3Current level, self-assessedNot the bot assessing a level, but the person's own words: 'complete beginner', 'intermediate', 'I can speak but not write'. That is information rather than a teaching decision.
4. 4Schedule constraintsWhich days and hours are possible. This is the most practical filter and it lets the counsellor arrive at the call prepared.

Level assessment: the boundary not to cross

This is where most implementations go wrong. The bot must not assess a student's level, decide which group they join, or say 'this course is right for you'.

• The placement test is the centre's own process — the bot can describe it, book it and say when the result will be known, but it cannot conduct it.
• The right course is chosen by a counsellor or teacher, not by the bot. The bot lists options.
• Set an earlier handover threshold on questions about children; a parent's decision usually turns on factors that never appear in the conversation.
• Do not disclose personal records — results, attendance, assessments — in chat, because identity cannot be verified.
• Never make outcome promises such as 'after this course you will get a job'.

What the counsellor should receive

• The four fields, structured, so the queue can be sorted.
• The specific course or group of interest.
• A one-sentence summary in the person's own words: 'mother, for a twelve-year-old, June exam, needs weekend schedule'.
• The time and channel of the conversation — an enquiry arriving at peak needs a different priority.
• Any pricing stated in the conversation, so the counsellor does not quote a different figure.

Handling the seasonal peak

Peak weeks are when the bot adds the most value and also when it breaks most often. Three preparations make the difference.

• Update the material before the peak: new dates, new prices, new group schedules. A bot quoting last term's start date does the most damage at exactly the wrong moment.
• Take remaining places from a live system rather than from an index. A student enrolled into a full group is the most expensive error of the season.
• Set a separate escalation rule for peak: some questions that normally go straight to a counsellor should be queued with a concrete stated response time instead.
• Read the abandoned conversations after the peak. They are the most accurate improvement list for next season.

What to measure

1. 1Enrolment conversionEnquiries that began in a conversation and ended in an enrolment. This is the headline number in education.
2. 2Counsellor reach and response timeHow many qualified enquiries a counsellor reached, and how long after. During the season this figure maps directly onto lost students.
3. 3Out-of-hours enquiry shareUsually more than half, and it justifies the project on its own.
4. 4Peak-week response timeMeasured separately from the overall average. Averages hide the season, and the season is where the whole difference is made.
5. 5The ten most asked questionsThey change every season, and they are the agenda for the next material update.

How Vexvon works in an education centre

The bot answers from the knowledge base — your site crawled to as many as 4,500 pages, PDFs up to 25 MB and bulk entries at up to 200 per request. Entries are typed into thirteen categories including pricing, opening hours, branch and campaign, which is what allows seasonal date and price changes to be held separately and updated before the peak.

Remaining places should not be answered from an index: your own system can be registered as an API tool the agent calls mid-conversation. When the lookup fails, the defined behaviour is to say it could not be checked — the mechanism that prevents a student being enrolled into a full group.

Name, phone and a one-sentence summary are extracted from the conversation, and a buyer-intent filter removes conversations that are not genuine enquiries before a lead is created. The lead reaches the counsellor group as a Telegram notification with a Start button, is assigned to whoever presses it and moves through a nine-stage pipeline, with reminders checked every minute.

Language is detected per message and the company can restrict the list, so a centre with a multilingual intake serves three languages without maintaining three bots. Every channel writes to one customer record.

Frequently asked questions

1. 1What should an education centre chatbot answer?Course lists, duration, price, start dates, schedule options, age and entry requirements, branch and document questions. These are most of the volume and none requires a teaching judgement.
2. 2Can the bot assess a student's level?No. Placement is the centre's own process. The bot can describe the test, book it and say when the result is known, but it must neither conduct it nor assign a course.
3. 3What should qualification ask?Who it is for, the goal, current level in the person's own words, and schedule constraints. Four fields are enough for most centres.
4. 4Can the bot say whether places remain?Only from a live lookup. A student enrolled into a full group is the most expensive error of the season, and an indexed page goes stale hourly during a peak week.
5. 5How do you prepare for the seasonal peak?Update dates and prices beforehand, tie availability to a live lookup, set a separate escalation rule for peak, and read the abandoned conversations afterwards.
6. 6How should success be measured?Enrolment conversion, time for a counsellor to reach the enquiry, out-of-hours enquiry share, and peak-week response time measured separately from the average.

Start with last season's questions

Export the questions that arrived during last enrolment season and sort them by frequency. The top twenty are your bot's specification. Then look at the hours they arrived in — that second figure is usually enough to justify the project on its own.