Most manufacturers already have cameras and sensors on site. Few use them to understand operations. Here’s what is changing, and where to start.
Walk through any modern plant and you will find cameras on the walls. Ask what they are for, and the answer is almost always security. Footage is recorded, stored, and consulted after something has gone wrong.
That is a narrow return on a sizeable investment. The same devices watch the production line, the loading bay, and the robotic cell. Yet manufacturers that have instrumented every digital process, from ERP to MES, still treat the physical floor as something to review rather than something to measure. Every unplanned stoppage, quality escape, or safety incident has a cost, and the evidence to prevent it is often already captured.
Why this is changing now
Three shifts are closing that gap. First, AI can interpret what a camera sees as it happens, rather than leaving someone to scroll through hours of footage. Second, cameras no longer work alone: environmental sensors, access control, and licence plate recognition add context that video cannot. Third, open APIs mean a manufacturer can describe an operational question and have a custom dashboard built to answer it in days, not months, without waiting for a vendor’s roadmap.
The results are measurable. Chassis manufacturer Metalsa installed cameras inside more than 40 robotic welding cells, areas closed to staff during operation. Engineers now see why a robot deviated from its programmed path without entering a hazardous zone. “For every camera we’ve had on one of the robots, it saves us thousands of dollars each month in terms of diagnosing issues pre-emptively,” says Chris Lewis, IT Support Specialist. At a quality inspection station, frame-by-frame review saves the Quality Assurance team 80 labour-hours a week, and real incident footage has become more than 350 training clips.
Outdoor cooking equipment manufacturer Middleby Outdoor cut investigation times from two days to ten minutes. Combining access control with video helped it tackle time-clock fraud, and license plate recognition removed discrepancies in collection and delivery records.
Naming the shift
We describe this emerging discipline as Physical Operations Intelligence: the use of connected, multi-sensor networks and AI to continuously understand what is happening in physical spaces, and to turn that understanding into decisions and actions the organisation defines for itself.
Security asks, “What happened?” Physical Operations Intelligence asks, “What is happening, and what should we do about it?” One mulch producer uses cameras as productivity meters, counting bags across four lines and flagging slowdowns as they occur. Another business pairs dock footage with deduction claims to settle shipping disputes. A blocked fire exit or a temperature drift in storage becomes an operational signal, not footage to find later.
The focus is on spaces, assets, conditions, and processes. For UK and European manufacturers, and for workforces wary of being watched, that matters.
A maturity path, not a switch
Organisations move through five stages: securing sites; monitoring them in real time; understanding them, as cameras and sensors combine to give operational context; operating, with use cases defined by the business itself; and finally autonomy, where conditions in a space trigger actions directly.
Metalsa shows the path. It began with nine low-resolution security cameras. Today it is piloting computer vision to track rail positions and production progress in real time.
What the guide covers
The accompanying guide sets out how manufacturers are applying this across safety, quality, downtime, and compliance, including traceability for regulators, accreditors, and customers. It also explores environmental monitoring where cameras cannot go, bringing existing cameras onto a modern platform without replacing them, and the full Metalsa case study. It is written for operations, quality, and health and safety leaders, not only security teams.
To download the guide, fill in the form below.
