AI is not just in the news. It is the news. It dominates tech, of course, but it also dominates finance, medicine, marketing, international politics—and my personal contender for one of the most promising intersections of bits and atoms, physical AI. The term refers to artificial intelligence systems that move beyond digital, software-only environments to perceive, understand, reason about, and interact with the physical world in real-time.

Our advances in physical AI tell a leadership story that's been overlooked in the usual account of our industry's evolution over recent decades.

We’ll get back to that in a minute, but first I’d like to offer a revisionist history of the role the security industry has played in technology leadership and adoption. I feel a correction is needed because, as an industry, we spend too much time beating ourselves up for being behind the bleeding edge of tech. A closer look at our history reveals that the security industry has led in technology adoption more than it's given credit for.

First, there was IoT. So, to rewind, let’s start with the classic central alarm station and its analog receivers awaiting dial-in calls from alarm panels. Crude as it was by today's standards, that architecture was at the forefront of the whole IoT craze. There’s no way around it: monitored alarm systems represent a distributed network of autonomous devices with centralized control, reporting, and data analysis. (I should know: half my college papers were written during night shifts in a central station.)

That’s IoT, before there was an I.

Next in our leadership hall of fame, there’s widespread use of mobile networks for machine-to-machine communications. Once dial-up alarm panels went cellular, they quickly became one of the largest classes of mobile-enabled devices in the world. And it wasn’t long after that we leveraged mobile phones to provide digital credentials for access control to hundreds of millions of people.

And now there’s edge computing—a trend that blends very nicely with physical AI. Here again, we’ve been at it for a while. By the time the dot-com era had come and gone, the security industry was already doing edge computing in a big way. Millions of smart security devices were already running code at the edge on network cameras, access controllers, and smart-home gateways, to name a few.

Today, we find the physical security industry once again reenacting its under-sung technology leadership role—this time in the domain of physical AI. The evidence? There are already 10’s millions of connected security devices streaming data to cloud-native service providers running AI platforms to sift the signal from the noise.

That’s far more than you’ll find in almost any other vertical market. Industrial robots, for example, numbered roughly 3.5 million at the end of 2024, with only a fraction under AI control. As for drones, roughly 1 million have been registered with the FAA, but most of those are still flown with conventional human control. Waymo—as impressive as it is—is only running about 4,000 autonomous AI-controlled vehicles as of this writing (and—side note—Tesla is running exactly zero.)

What does this say about the direction of the industry?

First, we are poised to be one of the truly verifiable ROI stories for physical AI in a landscape where measurable productivity gains have been elusive. A 2025 MIT report, titled "The GenAI Divide" found that 95% of enterprise generative AI pilots failed to deliver measurable return on investment (ROI) or significant productivity gains. Despite $30–$40 billion in spending, most initiatives stalled at the pilot stage due to poor integration, not model capability. As many enterprises struggle to document exactly how AI is improving their P&L statements, security systems are already reducing labor burdens and improving security outcomes.

Second, we stand at the front of the line—perhaps only second to medicine—as a demonstrably life-saving use of the technology. A single school shooting thwarted by an AI-driven gun-detection algorithm is all the proof anyone should need. Uvalde is the perfect case in point. The gunman was visibly brandishing his weapon outside the school long before he entered and began shooting. This type of scenario is well within the detection capability of AI algorithms to alert school officials and law enforcement. When automated and coupled to an access control lockdown capability, it can keep a shooter at bay long enough for law enforcement to arrive and neutralize the threat.

And finally, the sheer physicality of our AI tools positions us well against the “AI is eating software” hysteria. As other purely digital software sectors fret about being hollowed out by vibe coding, the security industry continues to enjoy a defensive moat based on getting physical with AI. Think about it. No matter how long some prompt-lord runs Claude Code, their vibe-coded app is not going to walk out of the room and wire up a security system.

So, as we continue to figure out the best uses of physical AI, let’s recognize and celebrate not only this latest chapter of technology leadership, but also the many that came before it.

Case closed. Thank you.