Why We Built AI Bot Eye
If fire is already visible, it should not remain unnoticed.
CCTV was already watching. We wanted it to speak up.
A fire can already be visible through a suitable CCTV camera while the people responsible still do not know. The camera may record everything, but recording alone does not create timely awareness. AI Bot Eye was built to help turn that visible moment into a warning with the event image, site and camera or area — through visual fire early warning on compatible existing CCTV.
- Rajkot
- Hometown
- A problem we could not ignore
Knowledge created a responsibility to act.
A devastating fire tragedy in our hometown of Rajkot made the cost of delayed awareness impossible for us to ignore. The loss of lives, including children, affected us deeply.
We could not undo what had happened, and we would never claim that one technology alone could have prevented it. But AI and computer vision were within our ability. That gave us a responsibility to explore whether existing CCTV could become an additional visual early-warning layer.
We could not undo what happened. We could choose what to build next.
Same site · Suitable existing cameras · A different response
The first possibility became real.
The first version proved the possibility. Making it useful meant earning the trust of the people responsible — in real operating environments, not only in a demo.
Same visible fire · two outcomes
Visible. Only recorded. If nobody is watching that feed, the cue may be found later — in review.
Visible. Evaluated. Alerted. A configured path can bring image, site and camera to the responsible team.
The responsible leader is the hero.
Safety, security, plant, facility and operations leaders already carry the duty to protect people, property and continuity. AI Bot Eye was built to guide that work — with empathy for the pressure, field knowledge from real sites, and a practical path from a suitable camera view to a configured response.
We do not replace the leader’s judgment. We help bring visible fire to their attention sooner.
Detection was only the beginning.
What real sites taught us.
A controlled demonstration can prove that detection is possible. Trust has to survive real factories, warehouses, spinning mills, hotels, offices and other environments — for the people who will live with the alerts.
- Every camera has its own environment
Reflections, lights, vehicle lamps, machinery, dust, steam and recurring activity can resemble fire or smoke. Before full live alerting, AI Bot Eye learns the normal visual environment and recurring fire-like patterns of each selected camera. Smoke remains camera- and scene-dependent.
- Known false events should create learning
When a confirmed false AI Bot Eye event is marked, the system retains camera-specific learning so the same known condition does not keep producing the same alarm. This is not a zero-false-alarm promise — it is how repeated noise is reduced on known conditions.
- An alert needs useful evidence
Leaders act faster when they know what was visible and where. Depending on the deployment, an event can include the image, site, meaningful camera or area name and event time — routed through channels the organisation already trusts.
- Every organisation responds differently
A manufacturing site, hotel, office, warehouse or multi-site operation may require a different response process. AI Bot Eye is configured around the customer’s approved process. It does not independently decide emergency action.
Field reliability
Trust begins before the fire.
AI can only monitor what the camera can actually see. During deployments, we repeatedly found selected cameras that were offline, frozen, obstructed, producing blank or corrupted frames, delivering unusable views or suffering repeated stream instability.
Camera Health became core
That experience led Camera Health to become a core reliability capability. It helps the organisation understand when selected visual coverage became unavailable, when it recovered and how long it was affected.
Feed readiness
Offline periods, frozen or non-updating feeds, blank or corrupted frames, and repeated stream instability.
View usability
Lens obstruction, significant view shift or unusable views, blur, low contrast, overexposure or abnormal colour changes.
Know if coverage is ready
If a camera is expected to contribute to early warning, the organisation should know whether that camera is actually ready.
The system matured where it had to work.
Built through real environments.
Early industrial environments, including spinning mills, helped the product learn from dust, large operating areas, changing light and recurring activity. Named deployments on the site show where that learning continued.
Manufacturing
Repeated visits, demonstrations and work in real manufacturing conditions helped the product mature where cameras already watch critical zones.
Read the Rolex Rings deployment →
Hospitality
A different operating environment — control-room assessment and back-of-house camera views on existing CCTV.
Read the Leela Palace deployment →Spinning mills near Rajkot
Spinning mills in and around Rajkot also became important early learning environments — dust, scale and recurring activity that a controlled demonstration alone cannot teach.
Product evolution from trust
As customers gained confidence in the underlying camera intelligence, some asked whether the same capability could address other meaningful visual problems. This led to site-specific work such as Stack Light Monitoring. Fire remains AI Bot Eye’s primary commercial focus.
Real operating environments produced real visual events.
Beyond controlled demonstrations.
Outside controlled fire tests, live deployments have also seen flame-related activity through selected site cameras. We do not treat every cue as an unwanted fire — and we do not publish prevention claims from these events.
What cameras have shown
Visible flame-related activity identified outside controlled demonstrations.
- Hot work Cutting or grinding sparks during planned site work.
- Burning trash Open burning near areas already covered by selected cameras.
- Outdoor flame Flame activity visible in yard or open industrial views.
- Unexpected cue Other flame-related activity that still needs classification.
How those events are judged
Same visual cue. Different meaning. Evidence decides the label.
- Legitimate operations Known hot work or operational flame that still deserves awareness.
- Needs review Unclear activity — image, site and camera go to the responsible team.
- May be unwanted fire Described only when the available evidence supports that reading.
- Not a demo claim Controlled tests stay labelled separately from field events.
These events are not all the same. Some are legitimate operational activities, some require review and some may represent unwanted fire. Each public example is described only according to the evidence available.
Core video intelligence can remain inside the facility.
On-site by design.
AI Bot Eye connects to compatible CCTV streams and processes video on an on-site edge device. It does not continuously record CCTV video as its core function. The customer’s existing NVR recording remains separate.
Core processing, configured event evidence, camera configuration and camera-specific learning can remain on the on-site system. Continuous CCTV video is not uploaded to a cloud platform.
Local processing and detection can operate without internet. External services such as WhatsApp, authorised remote support and online updates require connectivity. Remote support is optional and customer-controlled. Alert paths depend on deployment design.
A practical path for the people responsible
We guide. The site decides.
Customers may begin with selected critical areas and expand after confidence. Timelines vary by site — because the responsible team owns the process.
You check the view
Priority risk areas, camera visibility, angles, image quality and stream availability — with our review alongside yours.
Together we learn the site
Selected cameras are adapted to their normal environment and recurring non-fire patterns.
We prove it with you
Controlled demonstrations test detection and the full camera-to-response path on your views.
You go live
Approved cameras and configured alert routes — under the organisation’s own response process.
Your judgment stays yours. AI Bot Eye brings the plan and the field method. The safety, security, plant or facility leader remains responsible for how the site acts.
An additional visual early-warning layer.
What AI Bot Eye is — and is not
AI Bot Eye works alongside certified fire-detection systems, alarm panels, suppression systems and the organisation’s approved emergency procedures.
It does not replace those systems. It does not currently integrate with, verify or suppress alarms from the customer’s fire panel.
Its role is to help bring visible fire within suitable CCTV views to the attention of the responsible team.
From finding out later to knowing sooner
From finding out later to knowing sooner.
We began with a simple belief: if danger is already visible, the people responsible should not be the last to know.
The years since then have taught us that responsible early warning requires more than an AI model. It requires suitable camera views, site-specific learning, useful evidence, reliable camera feeds, practical alert routes and a response process the organisation trusts.
AI Bot Eye exists to guide responsible safety, security, plant, facility and operations leaders — helping turn visible fire into timely warning while there is still time to act. The responsibility remains with the people protecting the site.
See AI Bot Eye work on your own cameras
Request an on-site fire-detection demonstration using suitable existing CCTV views. We will review the relevant cameras, conduct a controlled demonstration of the camera-to-response path and recommend a practical pilot or deployment approach.