A large construction site in the UAE or KSA might have three hundred cameras covering every zone, entrance, and crane radius. Those cameras are recording continuously. In most cases, no one is watching the feed in real time, and the recordings are reviewed only when something has already gone wrong.
This is the standard operating model across the industry. Cameras are treated as an insurance mechanism — evidence after the fact — rather than an intelligence tool. The footage loops. The risk signals evaporate with it.
Why manual compliance monitoring does not scale
Safety compliance on a large site is monitored through walkdowns. HSE officers conduct scheduled and unscheduled inspections, record non-compliances on paper or in a mobile app, and escalate to supervisors. This process has not changed meaningfully in two decades.
The coverage problem is structural. A site with five thousand workers operating across twenty active zones cannot be adequately monitored by a team of HSE officers conducting walkdowns. The ratio of monitoring capacity to monitored population is too low. Non-compliances that occur between inspections go unrecorded. Zones that are inconvenient to access get inspected less frequently. The data generated by the process is a sample — not a complete picture.
The consequence is not just regulatory. Every non-compliance that goes undetected and uncorrected is a potential incident. In MENA markets where large giga-projects are operating at scale with workforces of tens of thousands, the gap between monitoring capacity and actual risk exposure is significant.
What the cameras already see
The cameras already mounted on most large construction sites have line-of-sight coverage of the zones where PPE compliance matters most: the areas where heavy equipment operates, where height work is conducted, where confined space entry occurs. The problem is not visibility. The problem is that no one is reading the signal those cameras produce.
Canopy's construction safety model reads the signal. It detects the presence or absence of hard hats, high-visibility vests, and safety harnesses in defined zones. It computes a compliance rate per zone per shift. It generates an alert on the first frame in which a non-compliance is detected — not after a walkdown, not when a supervisor reviews footage, but in near-real time as the camera sees it.
No individual is identified. The output is a zone-level compliance rate and an event timestamp. The alert contains enough information to dispatch an HSE officer to the right zone at the right time. It does not contain a face image, a name, or any identifier.
The operational case for developers
The commercial case for automating construction site safety monitoring has three components.
Incident reduction. Sites with higher PPE compliance rates have lower incident rates. This is not a correlation that requires complicated analysis — it is the mechanism by which PPE works. Continuous monitoring increases compliance rates by making non-compliance immediately visible and therefore immediately correctable. The expected impact on incident frequency is measurable within a single project cycle.
Insurance and liability. Developers operating at giga-project scale carry significant liability exposure. Documentation of continuous safety monitoring — zone-level compliance rates, event logs, alert records — is not currently producible by any manual monitoring process. Canopy produces it as a standard output. That documentation has direct value in insurance negotiations and in any regulatory or legal proceeding that follows an incident.
Labour contractor accountability. On large projects with multiple labour contractors, safety compliance varies significantly by contractor. Zone-level compliance data makes that variation visible. Contractors with persistent compliance problems can be identified before an incident occurs rather than after. That changes the nature of the performance management conversation between developer and contractor in a way that verbal site reports and walkdown summaries cannot.
The cameras are already there. The intelligence they could produce is not being extracted. That is the gap Canopy closes — not by adding infrastructure, but by reading the signal the existing infrastructure already generates.