Public payroll audited in real time
An AI-based payroll audit platform — replacing occasional sampling with a continuous second line of internal control, powered by high-performance computing and generative models.
Billion-scale volume, oversight by sampling
Public payroll is a chronic problem
Payroll processing error rates are estimated at 0.4% to 0.7%. At billion-scale volumes, that means hundreds of millions in annual inconsistencies that traditional auditing never catches.
Non-discretionary spending
Personnel costs are mandatory — they can't simply be cut.
External factors
Court rulings and index revisions have a constant impact.
Fiscal pressure
Variable revenue against fixed spending — an explosive combination.
Demographics
Demographic shifts put pressure on public employee pension systems.
Legislation
The complexity of state legislation, especially where it changes over time.
Data quality
An employee's career history sits in silos — which blocks the 360° view.
Five things that set Vert apart
- Experts in state legislation — a tailored approach, with rules organized by business experts.
- A continuous second line of control — rather than an occasional sampling exercise.
- Distributed in-memory HPC — billions of historical records evaluated with fast response times.
- End-to-end integration — from data acquisition to the auditor's screen.
- Generative AI — structuring court cases, official gazettes and personnel records.
Techniques combined into one consolidated score
Classification models
External data
Text mining
Relationship network
Geolocation
Deep learning
Anomaly detection
Business rules
Autonomous decisions
Alerts in production
- Record integrity — are these people alive? Are there material inconsistencies? 1.7% of records show registration inconsistencies.
- Exclusive-dedication allowance — cross-checked against official gazettes and court cases to validate approvals.
- Death, termination and special allowances — automated alerts across every payroll cycle.
- Pay items under continuous control — excused absences, base salary, seniority bonuses and transport allowance.
- LLM for data extraction — 57,090 files processed and 14,489 people correlated in a single pass.
An end-to-end platform
From the sources (official gazette, the e-SAJ court system, personnel records, mainframe and the payroll system) to the data lakehouse (Bronze · Silver · Gold), with a vector database + LLM for extraction, audit AI agents and presentation to the auditor — all with native governance, observability, orchestration and security.
Continuous auditing, not sampling
Find out how much your organization loses to inconsistencies the traditional process never sees.