Most cloud loss-prevention demos look simple: cameras upload video, a remote server "sees" theft, your phone pings. On an Australian retail floor - pharmacy, cosmetics, liquor, fashion - that path collides with three realities: NBN upload, the seconds you have before exit, and customer faces as personal information.
The alternative is inference at the edge: frames analysed inside your building, alerts in seconds, footage that never leaves for routine detection.
> Quick answer: On-premises retail CCTV AI keeps video and inference in your building - typical alerts in about two seconds, no routine offshore upload, detection that can continue when the internet is poor.
At a glance
When CCTV footage uploads for analytics:
Pharmacy customers expect discretion in health-adjacent retail. Fashion and liquor shoppers still expect their visit not to become training data for a foreign model. Trust is a retail asset either way.
Architecture overview: edge AI vs cloud CCTV.
Edge AI runs models on hardware at the point of capture. A compact appliance joins your camera LAN, ingests streams, and scores behaviour locally.
Detection pipeline in plain stages: how AI detects shoplifting in real time.
| Factor | Cloud analytics | On-premises edge |
|---|---|---|
| Typical alert latency | 5 to 15 seconds | ~2 seconds |
| Internet required for detection | Yes | No |
| Peak-hour NBN contention | Degrades upload | Local LAN only |
| Floor display during outage | Often fails | Continues |
| Bandwidth cost | High continuous upload | Low (updates only) |
Fifteen seconds is enough to conceal a high-value item and reach the door. Two seconds is enough for a trained approach in-aisle. Regional and suburban stores on contended NBN see cloud blindness during peaks - when planned theft prefers crowds. Tactics: organised retail crime.
Themes Australian retailers ask before they buy:
This section is education, not legal advice. Ask your adviser how APP 8 and related principles apply to your cameras and contracts.
Shrink still funds the project: cost of retail shrinkage.
Managed edge services can get quieter and more useful over time without turning your live floor into a cloud training set:
Ask vendors what bytes leave the building, when, and under which contract clauses.
Cloud analytics may suffice when:
For real-time loss prevention where speed, privacy, and offline operation matter, on-premises edge matches how Australian retail actually runs. Adoption discipline: edge AI adoption roadmap.
IntelliGuard was designed on-premises for Australian retail and pharmacy loss prevention. The NeuraIQ appliance runs concealment detection locally. Alerts reach staff in about two seconds with who-to-approach context and aisle location. No video uploads to cloud servers for inference.
$299 to $369 per store per month (excl. GST). Edge appliance included. 30-day money-back guarantee.
Cloud AI works for many jobs. Real-time retail CCTV security is a poor fit when custody, latency, and NBN reality collide. Choose architecture before you choose brand - then prove it on your highest-shrink floor.
See how IntelliGuard works or book a demo.
Sources: OAIC guidance on CCTV and personal information; OAIC Australian Privacy Principle 8 (cross-border disclosure) - confirm current OAIC pages when advising customers.