Articles · Training
Classes not filling? The training centre loses people after the trial
12 October 2026
The groups are not filling, enquiries keep coming in all the same, and the receptionists say they cannot keep up. The owner of a training centre puts these together into a picture that makes sense: not enough hands, so people give up waiting for a reply and go elsewhere. A vendor is offering an AI receptionist that does the replying itself, and promises that one person will then be enough for two branches.
The under-filling is real. At Vektor Znaniy training centre, 70% of places in the groups now running are taken and 165 are free. The centre is invented — it is one of the companies we test our audit agent on, and the agent wrote the report on its data. Two branches, one receptionist in each, 1,639 enquiries in four months.
"We can't keep up" has one peculiarity: it is said without dates. Lay it out on a calendar and it falls into four stretches of different lengths, each with its own answer.
A normal week: 71 enquiries, and they get dealt with
A normal week brings the centre 71 enquiries. The receptionist both staffs the desk and answers enquiries; by her account, enquiries and messages take about two hours a day. Of the enquiries from weeks like these, 3.5% are still unprocessed.
Evening and Sunday enquiries — 26% of the total — are dealt with on the morning of the next working day, and no harm from that can be seen: 22% of them end in a contract, against 17% of daytime ones. Round-the-clock replies, the AI receptionist's main selling point, would fix something here that does not hurt. One caveat: response time itself cannot be calculated, because the log of calls and messages has not been exported from the system.
Four weeks of enrolment: 165 enquiries a week, and here they really are swamped
From 17 August to 13 September enquiries ran at 165 a week, 2.3 times the usual. In those weeks, the receptionist says, enquiries sit for two days, dealing with them takes the whole day, and they still cannot keep up. Of the enquiries from those four weeks, 22% — 145 of them — are still marked as new, half of them for 22 days or longer.
So the owner is right — for four weeks. That changes the question: what is needed is not a worker for the year but hands for enrolment. The report names a temporary assistant for the receptionist before the next enrolment: 240,000 KZT for six weeks, and since the surge lasts four, the length is worth discussing. This is a cost, not a saving.
Every day after the trial lesson: of 500 who came, 213 signed a contract
The third stretch is not seasonal; it runs all the time. About 121 people a month come to a trial lesson. Of the 500 who came, 213 signed a contract — 43%. Another 149 were closed as lost with no reason given. And 76 people are still at "trial held": neither a contract nor a loss, and for 61 of them the trial was more than a week ago.
Of the 500 people who came to a trial lesson, 213 signed a contract. The centre has no rule that someone calls them afterwards.
What happens to someone after the trial is left to chance: "we call when we remember". Yet the centre has already spent more on these people than on anyone else. The enquiry was taken, the lesson booked, the teacher taught it, the person came and saw. One action is left, and it is the one nobody has been given.
From the inside this loss makes no sound. An unanswered enquiry is noisy: it sits in the list of new ones, and the receptionist complains about it. A person who has had their trial is silent. They are no longer a new enquiry, and the "trial held" stage looks like work done. The owner hears the noise at the entrance and mends the entrance.
Two more holes of the same sort lie nearby. Of those booked for a trial, 27% do not turn up, even though the system sends a reminder the day before — why, nobody records. And a reason is recorded for only 36% of lost enquiries; of those recorded, the most common is "the timetable didn't suit", at 22%. It may be that some of the 165 free places are at the wrong times, but there is nothing to prove it with yet.
Twelve months: the contract term on offer
The AI receptionist costs 150,000 KZT to set up and 70,000 KZT a month — 990,000 KZT in the first year — and the contract runs for 12 months. It works only in WhatsApp, which accounts for 32% of enquiries. It does not take calls, draw up contracts or staff the desk.
The promise of one receptionist for two branches can be checked against the timetable. Classes run in both branches at once in 15 of the week's 23 teaching slots, and each front desk needs a person. There is nobody to provide cover: the academic coordinator already covers the receptionists' days off, holidays and sick leave. Not one payment the centre makes goes away with this purchase.
What is on the table, then, is a year's contract for a load that lasts four weeks, and one that misses a loss running every day. In that position there is no reason to sign for twelve months before a trial of the software.
The next-day call is free; the agent comes after it
First, a rule in the enquiry-tracking system: the day after a trial lesson, the receptionist gets a task to call. About 5 minutes per call, roughly 10.1 hours a month between the two of them. Start with the 76 people who are already stuck. This adds work, and the report puts no figure on how many clients the call will win back: only a check will show.
Second, make the reason mandatory for a lost enquiry and for a no-show. Without that, there will be just as many questions in six months' time.
There is a place for an agent in the third stretch, not the first. The report describes an agent that every day finds the people who came to a trial and were left with no next step, reminds the receptionist and messages the client using approved wording. The main condition: the system has to release the history of enquiries, and at present it releases only a list. The conversation with a parent after the trial, prices, discounts and contracts stay with people.
The report leaves a way back to the AI receptionist too. It is worth returning to two to three months after the call rule and the mandatory lost reason are in place — and only if it turns out that the time really is going on routine questions in WhatsApp. So far nobody has counted.
Three numbers a training centre can pull today
- What became of the people who came to a trial. In the enquiry-tracking system, take everyone who had a trial lesson in the last three to four months. Sort them into three piles: contract, lost, neither. The warning sign is the third pile, and especially the people in it whose trial was more than a week ago — 61 of 76 in the example. Then ask the receptionist who calls after a trial, and on which day. "When we remember" means there is no rule.
- Enquiries by week, and the share still marked "new". The same system: the date each enquiry was created, counted week by week. In the example, 71 enquiries in a normal week and 165 during enrolment, with 3.5% and 22% unprocessed. If the backlog is bunched into a few weeks, you need hands for those weeks, not software for the year. If enquiries sit unanswered all year round, your case is a different one, and a conversation about help with messages makes sense.
- The share of lost enquiries with a reason recorded. Filter by "lost" status and look at the reason field. In the example, 36%. If fewer than half have a reason, you do not know why the groups are under-filled, and any answer to that question is an opinion. Among the recorded reasons, look at the share of "the timetable didn't suit", and check the hours of the free places against the hours people ask about.
In our example the calendar showed two different troubles: the entrance is overloaded for four weeks, and after the trial lesson people are lost all the time. The order of work follows from that — the next-day call first, then hands for enrolment, and only then a conversation about software.
Want the same look at your own company? The audit is free for now — it is an experiment. Show us how your work runs and get a report: where time and money are lost, and where AI would help and where it would not.
Request an auditSee sample reports
“Audit” here means a review of how the company works, not an audit of financial statements. The companies in the examples are invented; the reports on them were written by our AI agent.