Hakimo.

It is 2:47 a.m. on a Wednesday. A figure hops a fence at a construction yard along Interstate 405 in Bothell, Washington. Within three seconds, a synthetic voice cuts through the dark: "You are trespassing on a monitored site. Leave immediately. Authorities have been notified." The figure bolts. No guard was on duty. No alarm panel blinked in a distant office. The entire episode - detection, assessment, deterrence - was handled by a piece of software running on a rack somewhere in San Francisco.

That software is Hakimo. And the fact that an AI just chased someone off a highway construction site tells you something about where physical security is headed.

A $300 Billion Industry Running on Instinct

Physical security is one of those markets that got big without getting smart. The global industry clears hundreds of billions in revenue each year, but the core workflow hasn't changed much since the 1980s: install cameras, hire guards, hope for the best. Guards watch screens until their eyes glaze. Alarm systems scream wolf until operators stop listening. By some industry estimates, over 90% of security alarms are false. The real threats slip through while everyone is busy chasing phantoms.

Meanwhile, cybersecurity raced ahead. Automated threat detection, real-time response, machine learning models that learn what "normal" looks like and flag anything that doesn't fit. Physical security - the part where actual humans get hurt and actual property gets stolen - stayed analog.

That gap is what caught the attention of two Stanford researchers in early 2020.

Two Stanford Researchers Walk Into a Security Operations Center

Sam Joseph and Sagar Honnungar met while both were looking for their next thing. Joseph had been working in enterprise software. Honnungar had just left Rubrik, the cloud data management company, where he'd filed two patents and built their first cloud-native Office 365 protection product. He'd studied electrical engineering at IIT Madras - ranking second in his department - then earned a master's at Stanford, researching machine learning under professors Srijan Kumar and Jure Leskovec.

They started talking to security professionals. The conversations followed a pattern: everyone had cameras, nobody trusted what the cameras were telling them. False alarms buried real incidents. Tailgating - someone slipping through a badge-secured door behind an authorized person - went undetected everywhere. Guards rotated too frequently to learn a building's rhythms.

The two founders saw a computer vision problem hiding inside a staffing problem. If you could build an AI model that actually understood what it was looking at - not just motion detection, but contextual understanding of scenes - you could replace the weakest link in the security chain. Not the cameras. The attention span.

They raised a $4 million seed round from Neotribe Ventures, Defy Ventures, and Firebolt Ventures in October 2020, and started building.

An AI Operator That Speaks, Sees, and Decides

The core product is the Hakimo AI Operator - an autonomous agent that combines computer vision with generative AI to do what a human security operator does, except continuously, across every camera feed simultaneously, at a fraction of the cost.

Here's how it works in practice. The AI Operator plugs into a customer's existing security infrastructure - any ONVIF-compliant camera, NVR, badge reader, or access control system. It watches the feeds. When it detects something unusual - and "unusual" can be defined in natural language, like "person near the fence after midnight" or "vehicle in the loading dock with no appointment" - it assesses the threat, delivers a customized audio talk-down through connected speakers, and escalates to a human operator if needed.

The system can detect any anomaly that can be described in words. That single sentence is essentially the whole product pitch, and it's more powerful than it sounds. Traditional alarm systems are hardcoded: motion in zone A triggers alert B. Hakimo's approach lets customers define threats the way they'd describe them to a new guard on their first day.

Building the Machine, Year by Year

From Highway Projects to Student Dorms

The Skanska story is illustrative. Skanska, one of the world's largest construction firms, had a problem on their 4.5-mile I-405 highway reconstruction project in Washington state: valuable equipment and materials spread across multiple work zones, with traditional security methods failing to keep up. They deployed five Hakimo camera systems - one stationary in the yard, four mobile units repositioned to active work areas.

The result: threats detected and responded to within seconds. Automated talk-downs stopped theft attempts before they became thefts. Scott Turner, Skanska's General Superintendent on the project, put it plainly: the AI functions are "key to the success of not having things stolen from our project." Skanska is now considering company-wide expansion.

At AMP, a student housing provider, Hakimo delivered a 66% cost reduction and approximately 90% decrease in crime rates. Curio Storage saw significant reductions in break-ins. A Fortune 500 manufacturer integrated Hakimo with their access control systems to catch tailgating in real time, cutting false alarms while flagging actual compliance violations.

Customers report annual savings of up to $125,000 compared to traditional guard-based security. The median response time across all sites is six seconds. Half of all incidents Hakimo catches happen between 11 p.m. and 4 a.m. - hours when human guards are statistically least alert.

Who's Betting on Hakimo

The investor roster reads like a deliberate strategy, not just a funding stack. Vertex Ventures and Zigg Capital led the Series A. Rocketship.vc led the earlier round. But the strategic investors tell the bigger story: Honeywell Ventures brings deep integration with building management systems. Carrier Ventures - the corporate venture arm of Carrier Global, a major building technology company - connects Hakimo to existing security infrastructure at scale. RXR Arden Digital Ventures adds real estate expertise. Angel investor Gokul Rajaram, known for leading products at Google, Square, and DoorDash, brings product sensibility.

The $20.5 million in total funding isn't enormous by Silicon Valley standards, but for a company that tripled revenue in 2025 while protecting over 400 properties, it suggests efficient capital deployment. The team has grown to 80+ employees split between the US and an R&D center in India.

Cybersecurity-Grade Intelligence for the Physical World

Hakimo's stated mission is to "create a future where security is proactive, intelligent, and seamlessly integrated." Strip away the mission-statement polish and the core idea is more specific: physical security should work like cybersecurity. Automated detection. Continuous monitoring without fatigue. Machine learning models that improve over time. Response times measured in seconds, not the 15-to-45-minute windows that define traditional guard dispatches.

The company's competitive edge isn't just the AI - it's the architecture. Hakimo integrates with whatever hardware a customer already has. ONVIF-compliant cameras, Axis speakers, existing VMS platforms, access control systems. They connect to operational tools like ServiceNow, Slack, and SIEM platforms. This "bring your own cameras" approach means customers don't rip and replace. They upgrade their eyes.

The Guard Station Is Getting Smaller

The physical security market is consolidating around a question: what happens when AI can do what a trained guard can do, but across every camera, every hour, without bathroom breaks or shift changes? Hakimo's answer is not that guards disappear - their model keeps humans in the loop for escalation and judgment calls. The answer is that the guard-to-camera ratio changes dramatically. One human operator, backstopped by AI, can monitor what previously required a team.

The competitors in this space - Deep Sentinel on the residential side, Rhombus with its cloud cameras, Arcules with its surveillance platform - are each working a different angle. Hakimo's differentiator is the AI Operator concept: an autonomous agent that doesn't just detect, but assesses, responds, and learns. It's closer to a digital colleague than a smart alarm.

With attacks on critical infrastructure rising (up 71% in the energy and utilities sector alone, according to Hakimo's own reporting), the demand for always-on, intelligent security isn't theoretical. It's urgent.

2:48 a.m. The Site Is Quiet Again

Back at the I-405 construction yard, the intruder is gone. The AI logged the incident, timestamped the footage, and filed the report before any human knew there was a situation. A supervisor in Seattle will review it over coffee in the morning, scrubbing through the footage with a natural language search - maybe typing "person near south fence after midnight" - and finding the clip in seconds.

Nothing was stolen. No guard was dispatched. No false alarm clogged the system. The whole event lasted about forty seconds.

Physical security used to be a problem of attention - not enough people watching not enough screens. Hakimo turned it into a problem of software. And software, unlike guards, scales without getting tired.

The fence is still there. The cameras are still there. The difference is what's behind them.

See Hakimo in Action