Three agents for the risks that cost the most: fire, cash, break-ins
A new class of agents that watch for a specific behaviour and its consequence, shipped in July 2026. What each one does, and the small interface change that matters more than it looks.
A cat crosses the frame at 2 a.m. and the alarm fires. That is motion detection, and it is where video analytics started. Most vendors now claim to have moved past it, and the July 2026 release from Spot AI is a useful example of what "past it" concretely means: agents defined by a behaviour and its consequence.
The company described the batch as "a new class of AI Agents" that "watch for specific risk behaviors, reason about what they see, and respond instantly, so incidents are handled in seconds instead of reviewed hours later." Three shipped first.
Possible Fire
Visual flame detection, indoors and out. Smoke detectors are slow in large open spaces and useless outdoors; a camera that recognises flame can raise an alarm while a fire is still a wastebasket. For a manufacturer with a yard full of pallets or batteries, this is the agent most likely to justify the subscription on its own.
Cash Register Theft
Detection of employee cash-handling theft "without touching the POS." That last clause is where the engineering is. Most register-fraud tools work from transaction logs and need integration with the point-of-sale system, which is exactly what small retailers and franchisees cannot arrange. Doing it from video alone means watching hands and drawers. It shipped as a beta, an admission that reading hands and drawers reliably is hard.
Possible Vehicle Break-in
Catching the attempt "before glass breaks," and talking the person down in real time. Vehicle break-ins were among the risks the company cited when it launched its outdoor security agent in early 2025; this moves the intervention earlier, from the sound of a window to the sight of someone working a door handle.
The change that matters more than it looks
The same release rebuilt pan-tilt-zoom controls, with three modes and a patrol feature that rotates through saved positions. It sounds like housekeeping. I think a PTZ camera on tour is the hardest thing in the building for an agent to reason about, because the scene it was configured to watch keeps changing. A retailer's asset-protection team asked Spot AI this exact question during a demo: how do AI zones stay valid when the camera moves? Presets and patrols are the beginning of an answer, because a saved position is a scene the agent can learn.
What to watch
New agents arrive in batches now, and the batches say what the company thinks customers will pay for. This batch covers fire, cash and cars. Each is a loss with a dollar figure attached and a window of a few seconds in which an alert can still change the outcome.