Fortix AI: Industry-Trained AI Agents on Every UK Camera

Fortix AI's pitch, in their own words: one platform that turns any camera system into an autonomous monitoring agent. For thirty years CCTV has done one job: record what happened so somebody can look at it afterwards. The picture quality improved, the storage got cheaper, the cameras moved to IP — but the workflow stayed identical. Something happens, the recorder fires, somebody trawls through the footage on Monday morning. That gap between the event and the response is where almost every CCTV deployment fails. A fly-tipping incident at 02:00 doesn't get a response until the morning shift starts. A trespass on a construction site at the weekend doesn't get noticed until the foreman walks in. A worker without a hard hat in a live exclusion zone gets recorded — and then gets recorded again, every shift, because nobody's watching every camera every minute. Fortix AI is the technology partner we work with to close that gap. Their platform takes the cameras you already have and turns them into autonomous monitoring agents that detect, understand and alert in real time. Here's why we've put them at the centre of our AI portfolio, and what it looks like in practice on a UK deployment. The core idea: cameras that read behaviour, not just motion The single biggest design decision in Fortix AI is that the detection models are trained on behaviour, not on motion. Conventional motion-trigger CCTV fires whenever pixels change between frames — which is why every wildlife camera in a hedgerow generates thousands of false alerts a week. Fortix is different: their industry-trained agents recognise the pattern of what's happening, not just the fact that something moved. A fly-tipping incident isn't "vehicle pulled up in a field". It's "vehicle pulled up, person got out, opened the back of the vehicle, deposited material, drove off" — with the number plate, the make, model and colour captured automatically through paired ANPR. An anti-social behaviour event isn't "people in frame". It's loitering, confrontation, the specific gestures that precede vandalism. PPE compliance isn't a generic detection — it's hard hat plus hi-vis plus eyewear per zone, with per-shift compliance reporting routed to the safety lead by Monday morning. A live loitering detection from a Fortix AI agent. The model fires on the behavioural pattern, not the motion alone — vastly fewer false alerts than conventional CCTV. Industry-trained agents — the same engine, tuned per vertical Fortix calls their detection models Industry-Trained Agents, and the framing matters. The same underlying inference engine runs across construction sites, council estates, agricultural land, manufacturing lines, logistics yards and environmental monitoring — but the models are tuned to the way each industry actually looks on a CCTV feed. The hi-vis a contractor wears on a national housebuilder site looks nothing like the hi-vis on a council refuse-collection lorry, and a generic "safety vest" model can't tell the difference. The industry-trained agent can. Out of the box, the platform covers people and vehicle detection, fire and smoke (including the early-warning thermal-signature case that beats most point sensors), fly-tipping with paired ANPR, anti-social behaviour, perimeter and access monitoring, PPE compliance, licence plate recognition, and weapons and violence detection. The expanding library adds shoplifting and theft detection, occupancy and people counting, facial recognition for access control, rural crime monitoring and environmental hazard detection — and where the standard models don't cover a specific use case, custom models can be trained in a few weeks. The architecture: edge box on-site, cloud dashboard everywhere The other Fortix design choice that matters for UK deployments is the edge box. A lot of "AI CCTV" runs entirely in the cloud — which means streaming raw video continuously to somewhere external, which is a non-starter on the off-grid solar 4G mobile CCTV cameras we sell. It also tends to be a non-starter for anyone with data-residency concerns, GDPR sensitivities or a council procurement team. Fortix's pattern is a small edge box that lives on your network and handles the inference locally. The cloud dashboard delivers the live feeds, the alerts and the auto-generated incident reports — but only structured event data and short evidence clips leave the site. That means you can run a full Fortix agent stack on a solar CCTV camera with a 20 GB/month SIM without melting the data plan. The deployment time matters too. Fortix's standard pattern is live in under 24 hours: the edge box ships pre-configured, your existing cameras connect to it via IP / RTSP / cloud / FTP, and you log into the dashboard the same working day. There is no rip-and-replace, no VMS swap, no firmware flash for the cameras. What this looks like for our customers We tend to package Fortix AI three ways. The first is straight bolt-on to an existing CCTV estate — same cameras, same NVR, plus an edge box and a subscription. The second is a turnkey solar or 4G mobile CCTV deployment where the AI agent is part of the order from day one — common on construction sites and council fly-tipping hotspots. The third, increasingly, is a defence and critical-infrastructure pattern where the Fortix layer sits in front of ATEX-rated cameras on hazardous-area installations. Across all three, what changes is the operational model. The cameras stop being a passive record-and-review tool and start being an active alerting layer that runs in the background. The safety lead gets a per-shift PPE compliance report instead of a stack of recorded incidents to investigate. The council enforcement officer gets a number plate and a timestamped clip instead of "we know it happened somewhere on Tuesday". The site foreman gets a push notification at the moment of intrusion instead of a phone call from the alarm company. The full Fortix AI solution page Our full breakdown — features, detection…