Reconstruct hours
Hours arrived late or incomplete and had to be reconstructed at the end of the month.
Case 01 · WorkHours
A Telegram bot and back office for field teams: working time, receipts and approvals reach accounting reliably, even with a poor connection.
Time exceptionTeam West · review
Receipt recognisedOCR · Approval missing
8 recordsready for payroll
7:42 hof 8:00 h planned
The starting point
The actual work happened in the field, but its data lived in messages, notes, photos and several spreadsheets.
Hours arrived late or incomplete and had to be reconstructed at the end of the month.
A receipt photo in chat was not yet a correctly classified and approved accounting record.
Working time, payroll and accounting had to be reconciled manually.
The new workflow
As you scroll, the same flow is shown from three perspectives: employee, office and system data.
The bot already knows the planned job. Standard time is one button, while deviations remain possible.
Hours, receipts and open cases are in one place. Every correction leaves an audit trail.
A local transaction comes first. Delivery follows in the background with retries and deduplication.
7:42 hof 8:00 h planned
7:15 → 7:30 hCorrection logged
Receipt #0421OCR · Approval pending
Payroll run8 records ready
Local transaction wh_1842 savedOK
Network unavailable · retryingRETRY
Delivery confirmed · Key recognisedSYNC
Duplicate record preventedSAFE
System views
All examples are deliberately anonymised and show the real scope of the product without personal data.
Behind the scenes
Every connection is isolated. An external service can fail without taking the field team’s end-of-day workflow down with it.
Control remains human
The model never decides alone, especially for bookings and corrections.
A recognised receipt is booked only after explicit confirmation. Any change automatically removes the approval.
Corrections and deletions write their audit entry before the actual change. History remains history.