A store that no longer lets theft slip through
VigiNova is an artificial intelligence model that automatically detects theft using computer vision. Codally co-designed the solution with our Montreal partner to turn passive video surveillance into actionable alerts, 24/7.
- Client
- VigiNova · Montreal, Canada
- Sector
- Security · Retail & Commerce
- Solution
- AI theft detection model
The challenge
A retail chain faced recurring losses from shoplifting. Cameras recorded footage, but no one watched the streams continuously — especially during off-peak hours. Teams discovered incidents after the fact, when it was too late to intervene.
- Shoplifting hard to spot in real time
- Costly, fallible human monitoring over long shifts
- Late alerts: losses discovered after the fact
- Rising security budget with no measurable drop in incidents
Our approach
With VigiNova, we designed a computer vision model trained to detect gestures and behaviors associated with theft. The system integrates with existing cameras, analyzes feeds continuously and instantly notifies security teams.
- Connects to existing cameras — no infrastructure overhaul
- AI model calibrated for business scenarios (theft, intrusion, abnormal behavior)
- Instant alerts: mobile, email and monitoring application
- Desktop dashboard to view, track and prioritize incidents
How the tool works
From video capture to on-site intervention, every step is automated so teams act at the right moment.
- 1Feed capture
Existing cameras feed the model in real time or on replay.
- 2AI analysis
The model automatically detects theft and suspicious behavior.
- 3Instant alert
Notification in under one second to the relevant teams.
- 4Targeted response
Agents respond to reliable signals, not blind patrols.
Security team side
Teams use a desktop application to supervise all sites, review alerts and replay flagged video sequences in a few clicks.
- Centralized monitoring
Unified view of all camera feeds and active alerts.
- History & replay
Access to AI-flagged sequences for analysis and evidence.
- Configurable thresholds
Sensitivity and rules adjustable per store.
- Performance indicators
Track incidents, reaction time and trends by site.
The results
Six months after deploying the AI model, security indicators improved measurably across pilot sites.
Before / After VigiNova
| Metric | Before | After |
|---|---|---|
| Detection delay | 4h 20 on average | 38s |
| Theft rate | 2.8 incidents / week | 1.9 incidents / week |
| Reaction time | 12 min on average | 6 min 20 |
| Surveillance coverage | Limited hours | 24/7 automated |
Codally × VigiNova support
Deployment goes beyond technical setup: we support model calibration, team training and continuous optimization.
- On-site calibration
Model tuning based on store layout and priority scenarios.
- Team training
Dashboard onboarding, alert handling and intervention best practices.
- Technical support
Responsive assistance for adjustments, updates and model improvements.
- Continuous improvement
Model refinement from field feedback to reduce false positives.
VigiNova helped us move from passive surveillance to proactive detection. Theft dropped measurably and our teams finally intervene at the right moment.
Security & compliance
Video feeds and alert data are processed with strict encryption protocols. The architecture meets Quebec privacy requirements (Bill 25) and international best practices.
- Encryption
Data encrypted in transit and at rest (TLS 1.3, AES-256).
- Secure hosting
Certified infrastructure with access control and logging.
- Privacy
Data minimization and restricted access for authorized teams.
- Bill 25 compliance
Governance aligned with Quebec personal information protection requirements.
Do you manage a store, warehouse or industrial site?
Request a personalized VigiNova demo and discover how the AI model can protect your spaces.