University admissions bot
Applicants kept asking the same 20 questions by email and phone. The bot answers around the clock and escalates the complex ones to a human.
Outcome
- ~80% of routine questions resolved without staff involvement
- Instant replies instead of «within one business day»
- Role
- End-to-end development
- Time it took
- 2 weeks
The problem
The admissions office handled enquiries through three channels: email, phone and walk-ins. The questions were always the same ones: deadlines, required documents, fees, student housing. A staff member answered them by hand every day, and in the evenings or at weekends applicants got no answer at all.
What I built
A Telegram bot with a tree of common questions plus keyword search. The answers live outside the code, so the office edits them without a developer.
The important decision here was not technical: the bot does not try to answer everything. If a question matches nothing it knows, it does not invent an answer. It hands the enquiry to a human and tells the applicant someone will get back to them. A bot that answers wrongly with confidence is worse than no bot at all.
How it works
Questions and answers live in a database, and so does the enquiry log. After a week you can see what gets asked most, and those answers are the ones worth moving into the first menu.
Stack
- Python
- aiogram 3
- SQLite
Got a similar task?
Describe it in two sentences. I'll tell you if it can be automated and roughly what it costs.