How does the system know what is normal and what is an anomaly?

During the first 4 weeks the system establishes a baseline for each employee — typical clock-in, clock-out, and break times. It then flags statistical deviations.

What if an employee works rotating shifts with varying hours?

AI monitors attendance relative to the planned shift. Deviations are assessed against the schedule, not against an absolute time standard.

Does AI monitoring work for remote / home office employees?

For remote employees AI monitors system login times (if Microsoft 365 integration is enabled), approved home office days, and task completion.

AI attendance anomaly detection: payroll under control

Labour costs account for 30–60 % of total costs at most companies. Small inaccuracies in attendance — systematic lateness, unrecorded early departures, unconfirmed overtime — can add up to thousands of euros per month.

What AI monitors in attendance

Individual patterns

Consistent lateness: employee X arrives every Monday 20–30 minutes later than their clocked-in time
Shortened working days: systematically leaving early without recording a shorter shift
Irregular breaks: breaks longer than prescribed
“Ghost” attendance: badge clock-in without physical presence (card borrowed by a colleague)

Team patterns

The whole team leaves early when the manager is not on site
Unusual overtime spikes before the end of the accounting month
Employees clocked in at multiple cost centres simultaneously

Payroll anomalies

Employees A and B have identical job content, but A consistently reports 15 % more hours
Attendance records modified shortly before payroll processing
Overtime approved without a work order

Output: monthly attendance report

ATTENDANCE — ANOMALIES — April 2026

🔴 Requires verification:
- Novák J. — 8 days in the month, clocked out 15–40 min early.
  Records not updated (total difference: 4.2 h = €63)
- Mléč P. — logged in at 2 cost centres simultaneously (12 Apr, 18 Apr)

🟡 To monitor:
- Warehouse team — average Friday departure 18 min early
  (past 3 months, total impact: 2 h/person/month)

✅ No anomalies: 28 employees

Legal considerations

AI attendance analysis complies with GDPR when:

Employees are informed about working time monitoring (mandatory)
Data is used exclusively for employment administration purposes
Records are retained for the period required by law (3 years)

Modulario generates the required documentation to fulfil the employer’s duty to inform.

AI attendance anomaly detection — uncover payroll fraud and inefficiencies