March Networks just made searching through security camera footage feel like magic. Their new AI Smart Search lets you literally speak or type what you're looking for—"show me everyone wearing a red jacket yesterday afternoon"—and it finds it in seconds across thousands of hours of video from hundreds of cameras.
I saw a demo at ISC West last week and immediately thought about all the times I've watched someone scrub through hours of footage looking for one specific moment. This completely changes that game.
How It Actually Works
Instead of storing full video, the system takes regular snapshots from all your cameras and converts them into a searchable database using generative AI. You search using natural language—like you're asking a person, not a computer.
"Find instances where the emergency exit was blocked" or "Show me when the freezer door was left open" or "Who was at the loading dock between 2 and 4 PM on Tuesday?" It understands context and pulls relevant footage.
The demo operator showed banking use cases. A manager could ask "Were all tellers wearing name badges today?" and get a report in seconds. For retail, "Show me displays that weren't facing the correct direction." For restaurants, "Check if employees washed hands after handling raw meat."
Stuff that would take hours of manual review happens instantly.
Why This Matters Beyond Security
The obvious use case is loss prevention and security. Find shoplifters, identify security breaches, check for policy violations. But the operational applications are what really got my attention.
A retail manager checking if marketing displays were set up correctly across 50 locations. An operations director verifying safety protocols are being followed. A quality control person auditing manufacturing processes without watching hours of footage.
Someone I know who manages a chain of restaurants said this would solve their biggest headache—ensuring food safety procedures are being followed consistently. Right now they rely on random spot checks. With this, they could audit systematically without dedicating staff to watching cameras all day.
The healthcare applications are potentially huge too. Verifying that staff are following sanitization protocols, checking that equipment is being properly maintained, ensuring privacy rules aren't being violated. All searchable instead of hoping you happen to notice during manual review.
The Privacy Conversation We Need to Have
Okay, let's talk about the elephant in the room. This technology is powerful, which means it can be misused. The ability to search through footage of people doing their jobs—or customers shopping, or anyone in view of a camera—has obvious privacy implications.
March Networks emphasized they're focused on operational and security use cases, not surveillance for the sake of surveillance. The system is designed for businesses checking compliance and safety, not for tracking individual behavior in creepy ways.
But the technology doesn't care about intentions. A company could use this to monitor bathroom break frequency, track how often employees check their phones, or a dozen other invasive things. The tool is neutral; the ethics depend on how it's used.
I'm conflicted about this. On one hand, businesses have legitimate needs to ensure safety, prevent theft, and verify compliance. On the other hand, this makes constant monitoring frictionless in a way that feels dystopian.
The Technical Side Is Impressive
Using snapshots instead of full video is clever from both a cost and performance standpoint. Processing full video with AI is expensive and slow. Snapshots give you enough information to be useful while being much cheaper to store and process.
The large language model integration means the search understands nuance. "Red jacket" also finds burgundy hoodies and crimson windbreakers. "Emergency exit blocked" recognizes boxes stacked in front of doors even if you never explicitly trained it on that.
The system learns over time too. If you frequently search for certain things, it gets better at understanding what you mean. That's both convenient and slightly concerning—it's learning what your priorities are, which says something about your management style.
Real-World Implementation Challenges
The demo was slick, but implementation isn't trivial. You need decent cameras, good lighting, and clear views of what you're trying to monitor. The AI can't find what the camera can't see.
There's also the garbage-in-garbage-out problem. If your cameras aren't positioned well or the image quality is poor, the AI will struggle. A security director at the demo said his biggest challenge wouldn't be the software—it'd be getting their existing camera infrastructure upgraded.
And then there's training staff to use it effectively. Knowing what questions to ask and how to phrase them to get useful results isn't always obvious. The system is powerful but requires some skill to use well.
Who This Is For
Obviously security companies and large retail operations are the target market. If you've got dozens or hundreds of locations with multiple cameras each, the ROI is clear. The time savings on manual review pay for themselves quickly.
But I could see mid-sized businesses finding value too. Not for constant monitoring, but for periodic audits and specific investigations. Did the cleaning crew actually show up? Was the equipment maintenance checklist really followed? The ability to verify instead of trusting is worth something.
Small businesses probably don't need this unless they have specific compliance requirements. But as the technology gets cheaper and easier to deploy, that could change.
The Future This Hints At
This is one piece of a larger trend: taking unstructured data (in this case, video footage) and making it searchable and analyzable using AI. Today it's security cameras. Tomorrow it could be body cam footage, drone surveillance, traffic cameras, or any other visual data source.
The implications are massive. Law enforcement, traffic management, urban planning, disaster response—anywhere you have lots of video data that needs analysis. The technology enables use cases that weren't practical before because manual review was too time-consuming.
My Take
I'm impressed by the technology and can see genuine value in operational monitoring and security applications. The ability to audit at scale without dedicating massive human resources is legitimately useful.
But I'm also wary of how this could be misused. Any technology that makes surveillance easier needs guardrails. Who has access? What can they search for? How long is data retained? What oversight exists?
March Networks seems to be thinking about these questions, but ultimately it's up to the companies deploying this to use it ethically. And history suggests that when you give people powerful monitoring tools, some will push boundaries on what's acceptable.
If I were a business owner, I'd use this for safety audits and compliance verification. I wouldn't use it to monitor employee productivity at a granular level, even though technically I could.
The technology is fascinating and will probably become ubiquitous in commercial settings. Whether that makes you excited or nervous probably depends on whether you're the one doing the searching or the one being searched for.
Either way, this is where we are now. Security footage isn't just dumb video storage anymore—it's a searchable database of everything that happened. That's powerful. Use that power wisely.