Video AgentsAll tags
Engineering Corner

Attention is a better word than surveillance

Surveillance records people so that someone can assign blame later. The safety directors running video agents describe something else: a camera that notices a forklift near-miss now, so the next one never happens.

By Daniel Reyes · 2 min read
Close-up of a CNC machine tool over a chuck, coolant lines behind
A CNC machine tool. Staccato, a firearms manufacturer on an 800-acre campus, runs no guards and no metal detectors. "We trust our employees," its technology director says. Photo: Daniel Smyth / Pexels

Surveillance is the word that ends most conversations about cameras and AI, and I think it does a lot of damage. It means watching people in order to catch them: a suspect, and a consequence after the fact. That is what the recorder era was built for, and it is why a camera has so rarely been welcome on a factory floor.

The systems being deployed now do something different, and the people running them describe it differently.

What a near-miss alert is for

At Unique Industries, a party-goods manufacturer with 450 employees across three shifts, the director of safety gets a notification when a forklift and a pedestrian come too close. "When we get an alert about a near-miss between a forklift and pedestrian, it's always legitimate," he told Spot AI. On his telling, the alert exists so that the next near-miss does not become a collision. His summary: "Tech goes with safety. It's a help, not a hindrance."

At Primex Farms, the safety lead described the shift this way: "Having cameras that talk back to you changes everything." A camera that talks back sounds like a co-worker tapping you on the shoulder.

Trust is a design choice

Staccato builds firearms on an 800-acre campus in Texas. Its director of technology, Mike Tiller, puts the culture this way: "We don't have security guards or metal detectors - we trust our employees." What he wanted from AI was context. "What impressed me was the system's ability to understand context," he said, and then a line vendors could print on the box: "The technology has to serve a purpose beyond just having AI for AI's sake."

Surveillance cannot give him that. It is indifferent to context by design; it records everything so that someone can decide later. Attention decides now, and mostly in favour of the person in the frame.

The more employees see me tackling issues like theft, the less likely they are to try it themselves.

That is Mike Wimpy, who runs loss prevention across twelve McKeever grocery stores. The value he describes is being visibly present, "everywhere," as he puts it, so that the incident never happens. Deterrence, in that description, is a form of attention.

Why the word matters

Words set defaults. A system built as surveillance retains everything and reviews it only after harm, with the people in the frame treated as subjects. A system built as attention keeps short retention, makes fast decisions, sends the alert to the person who can act, and keeps a record to prove what was prevented.

Spot AI's own head of customer success titled a recent essay "Passive surveillance is dead." He is right, and I would rather the replacement had a plain name than a euphemism. Attention is what a good supervisor gives a floor and what a good guard gives a parking lot, and there has never been enough of it to go round. The safety director at Unique Industries gets his as a forklift alert that, he says, is always legitimate.