// Python Dev
Robot collector: automatic debtor calls
Published on 2026-03-26
Client task
An IT company supporting accounting software called debtors by hand every month. This pulled an employee into repetitive work: reading the same script, recording responses, and calling back people who did not answer.
What was delivered
- • Built an automated voice-calling system that loads debtor data from Google Sheets.
- • The system generates the message, voices it through TTS, places calls through telephony, records the customer response, and writes the result back to the spreadsheet.
- • It supports both a strict question-and-answer scenario and a free-form monologue based on arbitrary text from the database.
- • Implemented automatic language detection and up to 10 retry attempts with a configurable interval.
- • Voice generation, calls, and speech recognition run in parallel, while each call status is tracked in real time.
Result
Monthly calling of dozens of clients no longer requires employee involvement. The operator receives a ready-made table showing who answered, what was said, and who could not be reached.
This project addressed a simple but expensive business problem: repeated manual calls to debtors by an employee who should have been doing higher-value work instead of reading scripts and logging responses.
The real challenge was not only to automate phone dialing, but to turn the whole process into a controlled pipeline: fetch the data, generate the right message, reach the customer, handle the response, and return the result to the place where the team already works.
The result was a practical collection tool that could be used for regular debt recovery without constant operator involvement.