Olga Gushchina
All work

AI sales rep for inbound

The bot booked 120 students. Half of them while the team slept.

Online English school, students worldwide

Live since February 2026

What it was like

  • The sale starts with a free trial lesson: someone leaves a request, a manager reaches out on WhatsApp and books them in.
  • Students are spread across the world, almost half of them in the US. The team is in another hemisphere.
  • Requests came in around the clock, and at night nobody could answer: by the time a manager woke up, the lead from Dallas had forgotten they ever asked.

Where it got difficult

  • Teacher availability lives in a Google Sheet and changes constantly.
  • "Tomorrow at six" means two different moments for a student in California and for the teacher.
  • Off-the-shelf bots break here: they confuse dates and zones and offer slots that are already taken.

What was built

  • An AI manager on WhatsApp replies instantly and carries the conversation from request to a booked trial lesson.
  • It checks the live schedule and offers only slots that are genuinely free.
  • A dedicated time-zone and date engine understands "tomorrow" and "on Wednesday" and converts the student's time into the teacher's.
  • The team kept the spreadsheet they were used to: they work in it exactly as before, the bot reads the same sheet.
  • Follow-ups: the bot chases both people who never replied and those who went quiet mid-conversation.
  • Hard cases go to a human manager with the reason attached.
  • Bot-or-human mode per contact: the team takes over any conversation with one click.
  • Every booking is logged to the team Telegram group.

How it looks

Click a frame to take a closer look

A request from another hemisphere: the bot answers at once, asks for the city and handles the time difference itself.
The same conversation to the end: the bot checks the schedule, offers a free slot and locks the booking in.
Every booking lands in the school team chat: date, time, the student’s city.
The live teacher schedule. The team keeps it exactly as before, the bot reads the same sheet.

Result

120

students booked into trial lessons with no human involved

52%

of bookings arrive outside working hours, every third one deep at night

14

time zones, students from 40+ cities, 42% from the US

53

complex conversations where a human manager stepped in

40–50 h

of manual messaging taken off the team

~₽5,000

a month to run it all: the model, the WhatsApp gateway, the server

A lost request costs you the average deal multiplied by your conversion to payment. At a ₽30,000 deal and 30% conversion, every person who never booked is ₽9,000 gone. Ten a month is ₽90,000, and over a year more than a million.

Similar problem?

Let's go through it in 20–30 minutes: I'll show what's solvable here and do the math in your numbers.

Next one

The founder answers students around the clock. He is asleep.