Stop Automating the Wrong Work

Physical security does not have an artificial intelligence (AI) problem—it has a workflow problem.

For the last 20 to 40 years, this industry has mostly iterated. We added better cameras, better access control panels, better sensors, better dashboards, better reports and more integrations. We rarely asked the harder question: Why does this process exist in the first place?

AI makes that question urgent. If we use it only to automate the old way of doing physical security, we may actually make things worse. We will clear alarms faster, generate reports faster and route tickets faster but still be stuck inside a model built around uncontextualized sensors, human fatigue and observe-and-report guard work. Putting a jet engine on a stagecoach is not transformation.

Research from McKinsey backs this up. Better tools are increasingly available to everyone—the companies pulling ahead are the ones applying technology to the right problems and rebuilding how work actually gets done. McKinsey’s study of leading AI transformations found an average 20% earnings before interest, taxes, depreciation and amortization (EBITDA) uplift and breakeven in one to two years, with $3 of incremental operating profit for every $1 invested. (EBITDA is a standard way to look at operating performance before financing decisions blur the picture.)

Why Physical Security Has Always Lagged—and Why That’s Changing

Physical security has always had a different adoption curve than other industries. In many businesses, value can move faster than trust. A company can test a marketing AI tool, a finance automation workflow or a customer service agent and accept some error while the model improves. Physical security does not have that luxury. Trust has to move at parity with value.

If an AI system gives a bad recommendation in a spreadsheet, someone fixes the math. If an AI system makes a bad call in physical security, someone may be put at risk, access may be granted incorrectly, a real incident may be missed or a guard may be sent into a situation without proper context. Physical security has lagged not because the opportunity was smaller, but because the trust threshold was higher.

That is changing fast. Better computer vision models, better data governance, better auditability, better integrations, better privacy controls and better human oversight frameworks are changing the equation. The real work is now not just experimenting with AI, but also reimagining workflows, shifting operating models and deliberately designing where humans stay in the loop and where they move above it.

Physical security has serious ground to cover. Pro-Vigil’s 2026 survey found that while 61% of business leaders now believe AI can improve physical security, only 15% have actually built it into their strategy. A quarter don’t even know whether their current systems use AI at all. The gap between belief and action is the story.

Context Is What Sensors Have Always Been Missing

Start with computer vision: A door contact can tell you a door was opened. A badge reader can tell you a credential was presented. A motion sensor can tell you something moved. None of those things can tell you what actually happened. Was it one person or three? Was there tailgating? Was someone carrying a prohibited object? Did a contractor enter a restricted cage and touch a server rack? Did a guard respond, or did the alarm get acknowledged into oblivion?

Computer vision changes the operating model because it replaces guessing with context: it gives physical security the eyes and the reasoning layer it has always been missing.

Automation says, “Let’s make the old process faster,” while transformation says, “Why are we still running that process at all?”

Those are not the same questions, and the industry has been answering the wrong one.

Here is what transformation actually looks like in practice:

• Instead of sending guards on fixed patrols because that is what the post order says, use computer vision to identify where activity, risk or policy violations are actually happening. Patrol based on reality, not habit.
• Instead of having operators stare at video walls waiting for something to happen, use AI to turn thousands of cameras into a real-time exception engine. Humans should investigate meaningful events, not babysit screens.
• Instead of treating access control alarms as isolated events, combine badge data with video to verify that one credential equals one human entry. That is how physical access starts looking more like zero trust in cybersecurity: trust is not granted once—it is continuously verified.
• Instead of writing after-the-fact reports from memory, let the system capture evidence, sequence, timeline and response automatically as the event unfolds. The human validates, decides and improves the process, rather than spending the night on clerical work.
• Instead of measuring security by how many alarms were cleared, measure it by how much unnecessary work was removed, how fast risk was understood and how often the right action happened without adding more human drag.

Not more software. Not more dashboards. Not more alerts. Fewer unnecessary processes.

The Market Is Moving Whether You Are Ready or Not

MarketsandMarkets projects the global physical security market will grow from $120.79 billion in 2025 to $151.5 billion by 2030. The winners in that market will not be the companies that digitize yesterday’s workflows—they will be the ones that understand physical security is becoming operational intelligence.

The broader enterprise is not waiting: 88% of organizations are already using AI in at least one business function, 62% are experimenting with agentic AI (systems that complete multi-step workflows autonomously, not just answer questions) and 23% are actively scaling it.

McKinsey notes that fewer than 10% have achieved full enterprise-wide AI value, most often because weak data and poor governance blocked progress. That is the specific warning for physical security: Scattered pilots on top of old processes will not get us there. The leverage points that matter are alarm reduction, access verification, guard force optimization, incident response, compliance and operational intelligence. Pick those and rebuild around them.

Computer vision as the sensing layer. AI as the reasoning layer. Humans focused on judgment, trust and action.

The companies that get this will move quickly. The ones that keep automating the old world will fall behind while telling themselves they are modernizing.

This is the AI industrial revolution for physical security. The winners will not be the ones with the most technology—they will be the ones with the courage to remove the work that never needed to exist in the first place.

Where to Start

Audit your workflows before you automate them.

1. Identify the dead weight. Map every alarm, patrol, report, handoff and guard task that only exists because your systems cannot see, understand or act on what is actually happening.

2. Eliminate before you automate. Remove the work that never needed to exist. Add computer vision as the sensing layer and AI as the reasoning layer—not on top of broken processes, but in place of them.

3. Put humans where judgment matters. Reserve human attention for decisions that require trust, context and accountability—not screen watching, alarm clearing or after-the-fact paperwork.

The views and opinions expressed in guest posts and/or profiles are those of the authors or sources and do not necessarily reflect the official policy or position of the Security Industry Association.

This article originally appeared in All Things AI, a newsletter presented by the SIA AI Advisory Board.