Skip to content
Camzify
Use case · Stock and cash

AI theft prevention

Theft prevention through video analytics means detecting unauthorized presence and movement in the areas where stock and cash are kept, while it is happening, rather than reviewing footage after the loss is counted. Camzify puts zones on stockrooms, cages, cash offices and docks, notifies in the hours each should be empty, flags entry with a snapshot, and runs a patrol round that checks the controlled areas are clear and records it.

  • Zones on stock and cash areas
  • Entry in set hours flagged with a snapshot
  • Controlled areas checked on the round
A warehouse operator reviewing an alert on a tablet while a ceiling camera covers a caged storage area
The problem

Shrinkage is counted, not caught

Stock goes missing and the stocktake finds out. The cameras over the stockroom recorded every entry for the last month and nobody has an hour to spare per camera to find the one that mattered. Loss prevention reviews footage after the count, identifies a pattern, and is too late for that pattern by definition.

The areas that matter are few and well defined: the stockroom, the cage, the cash office, the dock. What they need is a rule about who should be there and when, and a record that the rule was checked.

How Camzify handles it

Zones on the places stock is kept, notified when they should be empty

Zone intrusion detection draws each controlled area on its camera, and the camera's notification window is set to the hours the area should be empty. A tracked person inside the cage during that window, or in the cash office at any hour it should be empty, raises an alert with the snapshot. Abandoned object detection catches stock staged by a fire exit or a bag left in a corridor, which is how goods leave a building that has a guard on the front door.

A patrol round checks the same areas on a schedule, clear, closed, nothing staged, and records each answer with a frame. When something does go missing, AI suspect search takes a description in plain language and returns matching appearances across the cameras and the time window, instead of a week of scrubbing. The same tracking gives occupancy per zone, which is how a store learns when the floor is too busy to watch.

The patrol round

What a stock-and-cash round checks

A theft-prevention sequence is the cameras on the controlled areas in order, and the checklist is the state each should be in: empty, locked, closed, nothing staged. Run at closing and through the night, it produces a report per round that shows the controls held, and once staff know the round exists and is recorded, it changes behavior on its own.

Include the dock in the sequence. Stock leaves through it, and a dock door found up at 22:00 is a checklist failure with a snapshot rather than a discovery at the next count. The closing round itself is covered under lock-up and closing checks.

A check found Not Compliant captures a snapshot and messages the guard designated for that camera; on a manual round the operator chooses to send it, on an automated round it goes on its own. The item stays Pending until it is marked Fixed, which captures the after frame, and the report shows both.

What it will not do

It will not see concealment, and it will not identify people. It works on presence, movement and objects against rules you define, and describes what it saw; it does not recognize faces or name anyone. It will not cover an area without a camera on it. And it will not stop the person in the cage; it puts the snapshot in front of whoever you designate.

We do not publish shrinkage reductions or detection rates. Those depend on your store, your staff and your cameras, and the trust page explains why we will not estimate them.

FAQ

Frequently asked questions

It does not detect the act of concealing merchandise, and we will not claim it does. What it detects is presence and movement against a rule: a person in the stockroom in the hours it should be empty, a person in the cage who did not come through the door, an object left where it should not be. For a retail floor that is awareness, not a shoplifting alarm.

Each camera carries a notification window. The cash office, the stockroom and the high-value cage can each be a zone, notified in the hours the area should be empty, and entry during them raises an alert with the snapshot. A patrol round checks those areas are clear on schedule and records it, which is a deterrent in itself once staff know the round exists and is recorded.

Multi-object tracking follows a subject across frames on one camera, and the cross-camera journey map links appearances across cameras on the same site. Attribute extraction adds a description, clothing color, bag, direction of travel. None of it is facial recognition, and none of it identifies a named person.

The dock is where stock leaves, so most theft-prevention rounds include it. Dock doors closed outside the delivery window, no vehicle at the bay after hours, staging area clear. The loading dock monitoring page covers that part in detail.

The channels are set per alert category and the severity per camera per detection, so a zone alert in the cash office can be critical and reach the manager by SMS while a stockroom alert during trading hours is informational. A failed checklist item on a round messages the guard designated for that camera.

Every detection is logged with camera, time and snapshot, and every round produces a report with the result of each item and its frame. When something is missing, AI suspect search takes a plain-language description and returns matching appearances across the indexed cameras and time window, which shortens the review considerably.

Ready to patrol your site 24/7?

Book a 15-minute demo and see a live patrol run on your own cameras.

This site is being updated

We are rebuilding pages as you read this, so an image, a link or a section may look unfinished for a while. The product itself is unaffected. If something important is broken, tell us at the contact page and we will fix it.