Hanwha Vision's 2026 forecast puts the operator in the commander's chair
In a trends piece for SourceSecurity, Hanwha's John Lutz Boorman argues that AI agents will take over routine watching and initial response, leaving humans with judgement calls and an oversight role the company calls AI governance manager.
Camera makers publish year-ahead forecasts every December, and I stop reading most of them by the second paragraph. Hanwha Vision's for 2026, written by John Lutz Boorman of Hanwha Vision Europe and published on SourceSecurity on December 17, 2025, is more specific than most about what it thinks the operator's job becomes once agents are in the loop.
The headline claim is that autonomous AI agents will "revolutionise video surveillance systems structurally and operationally" in 2026. That is vendor language. The substance is in the division of labour Boorman lays out. The agent will "autonomously conduct complex situational analysis, automatically execute an initial response, and recommend the most effective follow-up actions." The person, in his phrasing, becomes a commander, making the final decisions that need nuanced judgement or carry legal weight, and also takes on a new duty he calls AI governance manager, tracking and supervising every autonomous action the system takes.
Five trends, one theme
The agent prediction is one of five. The others are about making the agent's input trustworthy, which is the part a camera manufacturer can control. Boorman argues that false alarms mostly trace back to bad pixels: low light, fog, sensor noise. His answer is what Hanwha calls a Trusted Data Environment, meaning deep-learning image signal processing that separates objects from noise, larger sensors that gather more light, and design-stage transparency to satisfy Europe's risk-based AI rules.
The remaining trends are energy, digital twins and architecture. On energy, Boorman cites the International Energy Agency's projection that data centre power consumption will more than double by 2030 and argues for low-power chipsets and doing more inference on the camera. On digital twins, he sees camera metadata joining access control and IoT sensor data in a live model of a building or a city. On architecture, the recommendation is hybrid: real-time detection at the edge, large-scale analytics and long-term storage in the cloud, with transmission cost and regulation as the reasons to keep video local.
What to make of it
Read as a product roadmap, the piece says Hanwha intends to compete on image quality and edge silicon while others compete on models, and that it expects to sell into agent architectures built by other people. Read as a forecast, the part I would keep is the governance role. Boorman works for a vendor with an interest in agents being adopted, and he still writes that someone in the security department will have to audit what the software did and why. Few of the agent launches this year have said as much.
Read the original at SourceSecurity.