■ HIGH RISK ■ Public Service & Government
AI cameras, drones, and patrol robots are replacing the watching part of watching — which was most of the job. Guards who respond, de-escalate, and make judgment calls stay employed; guards whose function was being awake in a chair are being budgeted out now.
“Cameras don't nap. Neither do motion sensors. You, however...”
Our AI replacement risk score — how we score jobs
Security guarding is mostly vigilance with occasional bursts of consequence: monitoring camera feeds and access points, walking patrol rounds, checking credentials at desks and gates, writing incident reports, and — rarely but critically — responding when something actually happens. The industry runs on thin margins and high turnover, staffing everything from corporate lobbies to construction sites, warehouses, hospitals, and events.
AI video analytics attack the job's center of mass. Modern camera systems detect intruders, loitering, weapons, and left objects across hundreds of feeds simultaneously, without fatigue — and human attention over a camera wall was never good after the first hour anyway. Autonomous patrol robots and drones now cover parking lots, warehouses, and perimeters on schedules no human sustains, while remote monitoring centers let one operator verify alarms across dozens of sites that each used to staff an overnight guard. For clients, replacing night posts with sensors plus remote verification plus on-call response is straightforward math, and the industry's own vendors are selling exactly that. Hence our 72: the posts justified by 'someone watching' are dissolving.
What survives is presence with judgment. De-escalating an agitated visitor, handling a trespasser humanely, managing crowds, protecting people rather than perimeters — hospitals, schools, events, and executive protection all need humans, and armed or specialized roles carry legal responsibilities no robot assumes. Ironically, the sensors generate more verified alerts needing a human response, shifting demand from static posts to mobile response and hybrid operator roles. The occupation won't vanish; the bottom of it — the lone overnight guard whose deterrent value was a heartbeat — is the layer being deleted.
Automatability: our editorial assessment of current and near-term AI capability
Already underway and compounding through this decade. AI video analytics and remote monitoring are displacing static and overnight posts right now, with patrol robots scaling in warehouses, lots, and campuses through the late 2020s. People-facing and response-oriented security holds or grows over the same period. If your post exists because a body must be present overnight, the countdown has started; if it exists because humans need protecting, you're fine.
The static ones, yes. AI camera analytics, patrol robots, and remote monitoring centers are actively replacing overnight and perimeter posts, because verifying alarms remotely costs a fraction of staffing a guard. Response, people-facing, and specialized roles are holding steady or growing. The occupation is splitting: watching is automating, protecting is not.
Anywhere the job is managing humans rather than watching property: hospital and school security, event and crowd management, executive protection, and armed or supervisory roles carrying legal accountability. Remote-monitoring operator and security-technology positions are also growing — someone has to run the systems replacing the posts. The exposed jobs are lone static posts whose function was presence.
As an entry point with a plan, yes; as a destination, be careful. Security remains one of the most accessible fields and a common bridge to law enforcement or corporate security careers. But plan your ladder early: stack de-escalation, medical, and technical training, aim for people-facing or supervisory roles, and don't build a decade around night watch posts that analytics are pricing out.
Partially — visible cameras, drones, and patrol robots do deter opportunistic crime, and they never miss a round. What they can't do is exercise judgment: distinguishing a lost visitor from a threat, de-escalating before something becomes an incident, or physically intervening. Sites with real human risk keep human guards; sites protecting empty buildings at 3 a.m. increasingly don't.