An AI surveillance system you can run on the cameras already on your wall

A surveillance setup where software analyses the camera feeds automatically and raises an event when something specific happens, rather than only recording for later review. In PGAK's case the intelligence sits in software behind the estate, so it applies to the cameras already installed instead of requiring new ones.

No. That is the most common and most expensive misconception in this category. A camera with AI on board is one clever camera; analytics running behind your recorder upgrades every camera it can reach. Existing cameras are usually the starting point, with a handful repositioned where a specific capability needs a better view.

Accuracy depends on your cameras, lighting and mounting heights, which is why a single headline percentage is close to meaningless. The useful question is how the system behaves on your site — how few false alarms it produces at night, and whether it still catches what matters. That is what a feasibility check on your own cameras answers.

Yes, with obligations. Recording on your own premises is permitted, and signage is expected. Where it gets stricter is workplace monitoring, biometric data such as face templates, and how long you retain footage and who can access it — all of which fall under the DPDP Act. Consent and retention policy are part of a proper deployment, not an optional extra.

Not if it is set up correctly, and this is the failure that kills most deployments. Because objects are classified before rules are applied, and rules are bound to zones and hours, the alert volume stays low enough to remain readable. A system producing alerts nobody opens by month three has failed regardless of how accurate its detection is.