Your cameras already saw it. Nobody was watching.
Theft detection for restaurants and supermarkets in the UAE. A-Eye watches the till, the floor and the back door continuously, and raises the handful of moments a manager should actually look at.
Most losses in food service and grocery are not dramatic. They are small, repeated and ordinary looking: a void with no customer present, an item passed around the scanner, stock leaving through a back door at a quiet hour. Cameras record all of it and nobody has time to watch. A-Eye flags the moments worth reviewing and attaches the picture, so a manager spends minutes on evidence instead of hours on footage.
Losses hide in the ordinary, not the obvious
Owners tend to picture theft as a dramatic event. In practice the expensive pattern is small, routine and repeated hundreds of times before anyone notices.
It looks like normal work
A refund, a void, a discount, a bin trip to the back. Each is a legitimate action that happens all day, which is exactly why the pattern hides inside it.
Nobody reviews the footage
Recording is not watching. Cameras get looked at after a loss is discovered, which is weeks late and only if someone knows roughly when to look.
The count only shows the total
A stock count tells you what is missing, never when or how. By then the trail is cold and the conversation becomes an accusation without evidence.
In food service the loss is doubly invisible, because product leaves as portions rather than units. Nothing in the till or the stock file separates a generous portion from a giveaway from a sale that never got rung.
The patterns worth a manager’s attention
A-Eye raises events, not verdicts. Each of these is a moment a person should look at, not a conclusion about anybody.
| Pattern | Where it shows up | Why it matters |
|---|---|---|
| Void or refund with no customer | Restaurant and grocery tills | A cancelled sale while nobody is at the counter is the single most common till pattern worth reviewing. |
| Drawer opened without a sale | Any till | No-sale drawer opens are normal occasionally and revealing when they cluster around one shift. |
| Item passed around the scanner | Supermarket checkouts | Goods that cross the counter without being scanned leave with a customer and stay on your stock file. |
| Items leaving with waste | Kitchens and back of house | Bins and waste runs are a routine path to the back door, and a routine one to check. |
| Back door activity at quiet hours | Restaurants and stores | Movement through goods-out when the site is closed or nearly empty is worth knowing about the same night. |
| Goods-in shortfalls | Delivery bays | Deliveries signed for in full but short on arrival are a loss that looks like a supplier problem for months. |
| Repeated pattern by shift | Across the week | One event is noise. The same event at the same hour every week is the thing that actually needs addressing. |
That last row is the one that changes outcomes. Individual incidents are usually explainable and often innocent. A pattern across weeks is what turns a suspicion into something a manager can act on fairly.
Pairing what the camera saw with what the till recorded
Run on the cameras you have
A-Eye works with existing CCTV wherever coverage allows. Whether current angles cover the till and the back door usefully is assessed per site rather than assumed.
Connect the transaction record
Till events and camera events are matched in time, so a void or refund can be seen next to what was happening at the counter when it was rung.
Flag the exceptions
Only the moments meeting agreed conditions are raised. The goal is a short daily list a manager can genuinely review, not an alert stream everyone learns to ignore.
Attach the evidence
Each flagged event stores a reference image of the moment, so a review starts from something concrete rather than a memory or a stock variance.
Review by a person, always
A manager decides what an event means. The system never concludes that theft occurred and never names anyone as responsible.
Feed the result back into Odoo
Confirmed losses become recorded adjustments with a reason attached, so your stock figures reflect reality rather than absorbing the gap silently.
What this does not do
Loss prevention is a subject where overselling causes real damage, so it is worth being exact about the limits.
It does not identify people
A-Eye flags events at a location and a time. It does not put a name to anyone, and we do not offer facial identification for this use case.
It does not decide guilt
A flagged event means something is worth looking at. Many turn out to be entirely legitimate. Treating a flag as proof would be both wrong and unfair to your staff.
It does not replace controls
Cameras do not fix weak till permissions, unlimited void rights or an unmonitored back door. Detection tells you where those gaps are; closing them is process work.
It does not catch everything
Any system covers what its cameras can see, in the conditions they see it. We would rather scope that honestly than let you assume total coverage you do not have.
Compliance and the conversation with your team
Two obligations sit with you as the operator, and both are better handled before a system goes in than after.
Surveillance approvals are yours
Camera permissions in the UAE are processed by whoever operates the premises, and that responsibility stays with you. We will build to the requirements; we cannot hold the approval on your behalf.
Employment decisions are yours
Anything arising from a review is a management and HR matter, handled under your own policies and local employment law. We supply an observation, never a recommendation about a person.
Built for these operations
Restaurants and cafes
Till voids, comped items and back of house movement, where product leaves as portions and never appears as a missing unit.
Supermarkets and grocery
Checkout scanning, high value lines and goods-in, across a range wide enough that manual review is impossible.
Multi-site operators
Comparison across branches, which is usually where a pattern becomes obvious that looked like bad luck at one site.
Quick service and delivery
High transaction volume and fast turnover, where small repeated losses add up long before anyone notices a variance.
Cloud and central kitchens
Production and dispatch areas with few customers present and a lot of product moving.
Franchise operations
Consistent oversight across sites you do not staff directly, using the same measures everywhere.
Theft detection FAQs
Do we need to install new cameras?
Often not. A-Eye is built to run on CCTV that is already in place. What matters is whether existing angles actually cover the till, the checkout and the goods door, which we assess per site before anything is committed.
Does it work with our POS?
Pairing camera events with till events is what makes this useful, so connecting the transaction record is part of the work. Where you run Odoo, the loop closes fully and confirmed losses become recorded adjustments.
Will this create hundreds of alerts nobody reads?
That is the usual failure mode and we design against it. Conditions are tuned to your site so the daily list is short enough that a manager genuinely reviews it. A list nobody opens is worse than no list.
Can it tell us who did it?
No. It flags an event at a place and a time, with an image of that moment. Identifying a person is a management judgement made by people who know the site, and it should stay that way.
Is continuous footage stored?
The system works from live video and keeps reference images of flagged moments rather than recording everything. Processing runs on site, so video is not sent to a cloud service.
Do we have to tell staff?
That is your call under your own policies and local law, and we would encourage it. Operators who explain it as protecting honest staff from blanket suspicion get better cooperation than those who deploy it quietly.
How long before we see anything useful?
Patterns tend to surface within the first weeks because the same conditions recur. What happens next depends on the process changes you make, which is the part that actually reduces the loss.
What if it turns out we do not have a theft problem?
Then you will have learned that, which is worth knowing. Plenty of losses are process failures rather than theft: bad goods-in checks, unrecorded waste, portion drift. Those show up the same way and are usually cheaper to fix.
Find out what your cameras are already recording
Tell us how many sites you run, what your current camera coverage looks like and which till system you use. We will tell you honestly what could be detected today and what would need changing first.