SquintPRO · AI cameras
AI cameras: replace your CCTV, or add AI to the cameras you already have
When a site asks for AI cameras, it usually wants results, not hardware: an alert when someone enters a restricted zone, a record of who was missing a hard hat, a vehicle found in last night's footage. There are two routes to those results. You can buy new cameras that run analytics inside the camera, or you can run AI software on the CCTV already installed. This guide explains both honestly, with a decision table for Saudi sites, then shows where SquintPRO, Interkey's own video analytics platform, fits and where it does not.
Published 9 min read
Orientation
At a glance
The whole page in one table, before the detail.
- The question behind "AI cameras"
- Which events the site needs detected, on which views, and who acts on the alert
- Route 1
- New cameras with analytics running inside each camera
- Route 2
- AI software that analyses the streams of the cameras you already own
- What usually decides it
- How many cameras you have, how many use cases each view needs, and how much central control you want
- Where SquintPRO fits
- Route 2: existing CCTV over RTSP or ONVIF, licensed per camera stream, unlimited use cases on a licensed stream
- Where it runs
- On-premise, so footage stays on your site, or in the cloud with 3, 6 or 12 months' retention
In short, for decision makers
- You rarely have to replace every camera. Most AI security camera projects can start on the CCTV already installed; a survey shows which views qualify.
- The two routes differ in where the intelligence lives. Inside each new camera, or in software that serves every camera from one place.
- Cost follows different drivers. Hardware and installation per camera on one route; licences per camera stream, processing and integration on the other.
- Our view: buy new cameras only for the views that genuinely need them, such as plate reading at a gate, and put the analytics in software you control.
What people mean when they ask for AI cameras
"AI camera", "AI CCTV camera", "smart cameras for business" and "AI vision camera" are used for very different things. Some suppliers mean a camera with a processor inside that runs a fixed set of detections. Others mean an ordinary camera whose stream is analysed by software on a server or in the cloud.
The words matter less than the outcome. Before talking to any supplier, write down three things:
Read more
- The events. A person without protective equipment in a work zone, a vehicle in a pedestrian area, entry after hours, a crowd building up at a gate.
- The views. Which cameras look at the places where those events happen, and at what distance and angle.
- The response. Who receives the alert, in which system, and what they are expected to do with it.
Then the camera question becomes concrete: does this view show the event clearly enough for software to detect it, or does it need a different camera?
Why the question matters for sites in Saudi Arabia
Industrial plants, logistics yards, ports, hospitals, campuses and public venues in the Kingdom often run hundreds of cameras that were installed for recording and human review. Replacing them is a capital project with procurement, cabling and installation windows; adding analytics to them is a software project. That difference shapes budgets, timelines and approvals.
Read more
Two further points are particular to Saudi sites. First, where video is processed is often set by policy. Government entities and regulated sectors work under the cybersecurity controls of the National Cybersecurity Authority (NCA), and footage that shows identifiable people is personal data under the Personal Data Protection Law, overseen by the Saudi Data and AI Authority (SDAIA). Whichever route you choose, settle where the stream is analysed and how long it is kept before go-live. Second, dust, midday glare, heat haze and poorly lit night areas affect detection on new and existing cameras alike, so test on your own site either way.
RelatedNational Cybersecurity Authority (NCA)Saudi Data and AI Authority (SDAIA)
Two routes to AI cameras
Route 1: analytics inside new cameras
The camera carries a processor that runs detections on its own view. The detections available are those the camera's maker supplies for that model, sometimes extended by add-on applications from the same maker. It can suit a small site needing one or two fixed functions on a handful of views, with no server room. Ask how many detections run at once on one camera, how new ones are added, how long the model is updated, and whether events from different makers' cameras can be managed centrally.
Read more
Route 2: AI software on the cameras you already have
The cameras stay as they are and send their video stream to analytics software running on servers at your site or in the cloud. One platform serves every connected camera, so a new use case is added in software rather than by changing hardware. It suits a large or mixed estate, several use cases per view, or a need to manage rules, alerts and evidence from one place. Its limits are real too: it needs processing capacity and network paths to the cameras, and it cannot fix a view the camera does not have.
Most sites end up combining the two in practice: software on the existing estate, plus a few new cameras where a view must change, such as plate reading at entry lanes.
Decision table: new AI cameras or AI on existing cameras
| What to compare | New cameras with built-in analytics | AI software on existing cameras |
|---|---|---|
| Main cost drivers | Camera hardware, cabling and installation for every view, plus removal of the old cameras | Software licences per camera stream, processing on site or in the cloud, and integration work |
| Use cases per camera | Limited to what the camera model supports and its processor can run at once | Set by the software; with SquintPRO, unlimited on a licensed stream |
| Adding a use case later | Depends on the maker releasing it for that model | Configured in software and tested on your own footage |
| Upgrades | Firmware and applications per camera; older models may stop receiving updates | The platform is upgraded once for every connected camera |
| Vendor lock-in | Analytics tied to one camera maker's range | Standard streams (RTSP, ONVIF) keep you free to choose cameras |
| Central management | Per camera, or through the maker's own management tool | One place for rules, alerts, evidence and search across the estate |
| Mixed camera estate | Usually means replacing what does not match | Works across makers once the survey qualifies each view |
| Usually fits when | A few views, one or two fixed functions, no server room | Many cameras, several use cases, alerts into an existing control room |
Neither column wins on every row. Use the table to decide view by view, not for the whole site at once.
Questions to settle before you buy AI security cameras
Put the same questions to every supplier, ourselves included, and ask for answers about your own site:
- Which of our existing cameras can be used as they are, which need a new angle or lighting, and which must be replaced, and why?
- How many use cases can run on one view, and what does adding one cost?
- Where is the video analysed, and does any footage leave our site?
- How long are footage and event records kept, and who can search them?
- Which system receives the alerts: our video management system, the control room or the incident system?
- How will false alerts and missed detections be measured on our footage before rollout?
- What happens to the analytics if we change camera supplier later?
Interkey has written up how to scope a pilot that proves something, including the success criteria worth agreeing in writing.
Where SquintPRO fits: AI on existing cameras
SquintPRO is Interkey's AI video analytics platform, and it takes the second route. It connects to the cameras you already own over standard RTSP or ONVIF streams; the camera compatibility guide explains what a camera needs to qualify.
- Licensed per camera stream. Only the streams connected to SquintPRO count, so you can start with the cameras that cover your highest-risk areas.
- Unlimited use cases on a licensed stream. Protective equipment, restricted zones, vehicles and footage search can share one view without another licence.
- New events without retraining. Events can be defined by text or by a reference image; test each definition on your own recordings first.
- On-premise or cloud. On-premise, footage stays on your site. In the cloud, recordings are kept for 3, 6 or 12 months.
- Alerts where your team already looks. Each alert carries video evidence and goes into your video management system, control room or incident system, not a new screen.
- No face recognition. SquintPRO detects people, vehicles and conditions; it does not identify who a person is.
Read more
Not every view will qualify. Plate reading at gates, for example, usually needs a camera positioned for it. The survey says which views need new hardware, and only those add cameras to the quote. See also computer vision solutions, from pilot to production.
Relatedcamera compatibility guidecomputer vision solutions, from pilot to production
Where SquintPRO is not the answer
If a small site needs a single fixed function on two or three cameras, and has neither a server room nor a network link able to carry streams to the cloud, cameras with built-in analytics may be the simpler choice, and we will say so. SquintPRO is also not the answer where the cameras cannot see the event: a view that is too distant, too steep or blinded by the sun needs a new position or a new camera first. And it does not identify people, so a project that requires face recognition needs a different scope.
How Interkey runs an AI camera project
-
Use cases and views.
The events that matter, the cameras that cover them, and who acts on each alert.
-
Camera survey.
Each camera is sorted into usable as is, usable after adjustment, or needing new hardware, with the reason for each.
-
Pilot.
One site, a defined set of cameras, one or two use cases and success criteria written before anything is switched on.
-
Measurement.
Detections, false alerts and missed events are measured on your own footage against those criteria.
-
Rollout and operation.
More cameras, use cases or sites are added once the pilot passes, with tuning against real site conditions.
Cost, data residency and local support in Saudi Arabia
SquintPRO has no published price list. Each quote is built from the site, the number of camera streams and the deployment model, together with any camera work and the depth of integration; what determines SquintPRO pricing explains each factor.
For data residency, the on-premise deployment keeps footage inside your own site, which suits entities whose policy fixes where video may be held. Interkey builds, deploys and supports SquintPRO from Riyadh, with Arabic-speaking engineers who carry out the survey, configuration, integration and tuning on site.
Questions buyers ask
Do we have to buy AI cameras to get AI video analytics, or can our current CCTV be used?
Usually the current CCTV can be used. SquintPRO analyses the streams of the cameras you already own, over RTSP or ONVIF, and a survey shows which views need a new angle, lighting or camera.
What is the difference between an AI camera and AI software on a normal camera?
An AI camera runs detections on a processor inside the camera, limited to what its maker supplies for that model. AI software runs on a server at your site or in the cloud and serves every connected camera, so use cases are added and managed in one place.
How is AI on existing cameras priced for a site in Saudi Arabia?
SquintPRO is licensed per camera stream, with unlimited use cases on a licensed stream, and is quoted per site. The quote also reflects the deployment model, any camera work and the integration depth. There is no published price list.
Can the video stay inside the Kingdom, or on our own site?
Yes. On-premise, SquintPRO runs on servers at your site and footage does not leave it. In the cloud, recordings are kept for 3, 6 or 12 months, chosen at the start. The choice usually follows your organisation's data policy.
How long does it take to add AI to an existing camera estate?
It depends on the number of cameras, what the survey finds and how many systems the alerts must reach. We estimate it as a range after the survey, and suggest a one-site pilot first.
Do AI security cameras recognise faces?
Some products do; SquintPRO does not. It detects people, vehicles and conditions such as missing protective equipment or entry into a restricted zone, without identifying who a person is.
Next
Related pages
Next step
Check which of your cameras can run AI
Tell us the site, roughly how many cameras it has and the events you want detected. We will tell you which views can carry analytics as they are, which need changes, and what a pilot could measure.
Or directly
+966-11-2180999 [email protected]Tawuniya Towers, North Tower, 7th Floor, King Fahad Highway, Olaya, P.O. Box 56835, Riyadh 11564, Saudi Arabia